- Email: [email protected]
An integrated reconfigurable control and self-organizing communication framework for community resilience microgrids
The Electricity Journal xxx (xxxx) xxx–xxx
Contents lists available at ScienceDirect
The Electricity Journal journal homepage: www.elsevier.com/locate/tej
Special Issue: Contemporary Strategies for Microgrid Operation & Control
An integrated reconfigurable control and self-organizing communication framework for community resilience microgrids ⁎
Lei Wua, , Jie Lia, Melike Erol-Kantarcib, Burak Kantarcib a b
Electrical and Computer Engineering Department, Clarkson University, Potsdam, New York, United States School of Electrical Engineering and Computer Science, University of Ottawa, Ottawa, Canada
A R T I C L E I N F O
A B S T R A C T
Keywords: Community resilience micrgrid distributed energy resource resilience integrated control and communication
CRMs can effectively share distributed energy resources of multiple owners and enhance resilient electricity supply in communities during disruptions. However, in contrast to single-entity microgrids, CRMs present a unique structure and bring new challenges for operations and control, which can often be interrupted due to cascaded failures in interconnected electrical/communication components. An installation in Potsdam, New York, incorporates some control/communication solutions for resilient and economic operation of CRMs.
1. Introduction of community resilience microgrids There is an increasing frequency of catastrophic weather events in the United States and globally, which inflict serious social and economic impacts. The critical issue associated with such catastrophes is the availability of electricity for recovery efforts. In fact, according to a report by the National Oceanic and Atmospheric Administration (NOAA), the U.S. has sustained 188 weather and climate disasters since 1980, more than 80% of them bringing damages to the nation’s electricity infrastructure (NOAA, 2014). The most notable example is Hurricane Sandy in 2012, which cost some 8.5 million customers their power during the storm and its aftermath (NOAA, 2013). Similar incidents, mostly due to ice storms, were also observed in Canada (Canadian Disaster Database, 2017) and other places around the globe. Fortunately, as an emerging technology, community resilience microgrids (CRM) can enhance resilient electricity supply to critical loads in the community during such disruption events. A CRM includes multiple distributed energy resources (DER) and critical loads that are owned and controlled by individual entities within a clearly defined electrical boundary, which are connected via primary distribution lines owned by a local regulated power company. Indeed, as shown in Fig. 1, CRMs extend benefits of traditional single-entity, behind-the-meter microgrid systems by effectively sharing DERs among multiple partners within a community, which could enhance energy resiliency by reducing the overall impact of and improving response and recovery to critical events for the entire community. However, a CRM presents a unique structure and brings new challenges for its operation and control. Specifically, DERs owned and controlled by individual partners can either represent individual ⁎
stakeholders’ financial interests when the CRM is operated in gridconnected mode, or collaborate as a single controllable entity for enhancing resilient electricity supply to critical loads when CRM is islanded. That is, in grid-connected mode, each CRM partner acts as a self-interested entity and operates DERs/flexible loads according to its specific objectives; during disruption events, on the other hand, CRM partners can coordinate with each other by sharing onsite generation capacities and flexible load curtailment/shifting capabilities, in order to supply community critical loads based on certain agreements. Thus, in different CRM operation modes, partners in a CRM would act as either self-interested competitors or as altruistic collaborators following the guidance of the CRM operator. Furthermore, CRMs are complex networked systems, which not only interconnect DERs and loads of multiple owners through distribution lines but also are overlaid with a communication and control system that gathers inputs and sends out control signals to multiple owners for enabling resilient and economic operation. Indeed, CRM operations can often be interrupted or halted due to the cascaded growth of failures in interconnected electrical/communication components or unwillingness/inability of individual CRM partners to respond. Thus, in order to enable the full functionality of CRMs, an integrated control and communication framework is needed for facilitating CRM operations while complying with system control and communication requirements under different operation modes. This article discusses the unique operational characteristics of CRMs across multiple timescales in different operation modes, and presents coordinated system control and communication solutions for enhancing resilient and economic operation of CRMs. Section 2 discusses a practical CRM project that is underway for serving the city of Potsdam, N.Y.
Corresponding author. E-mail address: [email protected] (L. Wu).
http://dx.doi.org/10.1016/j.tej.2017.03.011
1040-6190/ © 2017 Elsevier Inc. All rights reserved.
Please cite this article as: Wu, L., The Electricity Journal (2017), http://dx.doi.org/10.1016/j.tej.2017.03.011
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
Fig. 1. Single-entity microgrid versus CRM.
Catastrophic weather events in Upstate New York, such as ice storms, major snow events, and micro-burst wind events, have historically caused widespread damages. The risks posed to residents and the coordination of emergency services are extremely challenging. One of the most devastating events was the ice storm of 1998, which affected most of northern New York State, southern Canada, and northern New England, and caused over $1.4 billion in damages/costs and 16 deaths in the U.S. (Ross and Lott, 2016). In fact, the availability of a functioning CRM after the disruption of the main grid during these events can enable operations of critical services and mobilization of first responders. Such a CRM can also furnish a functional staging area for teams to perform disaster recovery tasks in the region and address the immediate needs of the most affected population. The Potsdam CRM project is led by Clarkson University, partners with GE Energy Consulting, Nova Energy Specialists, and National Grid. Subsequently, during 2014 to 2017, four synergistic projects have been funded by U.S. DOE, NSF, and New York State Department of Public Service (NYDPS). The DOE project led by GE Global Research funds the development and testing of a microgrid controller, in which Clarkson is subcontractor. One NSF project led by Clarkson University focuses on the human-machine operational impacts of the CRM during normal operations and disaster response, while the other NSF project led by
Section 3 evaluates different operation strategies of a CRM in order to optimally manage dispatches of DERs in grid-connected mode and enhance resiliency in islanded mode. Section 4 presents a self-organizing, small cell-based infrastructure for ensuring flexible, fast, and reliable communication requirements of CRM operations. Section 5 describes an integrated reconfigurable control and self-organizing communication framework to study the interdependency and interaction between control strategies and communication requirements. The paper is concluded in Section 6.
2. Current efforts of the CRM Project in Potsdam The development of CRM technology to address disaster response is currently underway through multiple projects across the U.S. One of such projects is the “Design of a Resilient Underground Microgrid in Potsdam, NY,” (Clarkson University prime contractor, 2014) jointly funded by the New York State Energy Research and Development Authority (NYSERDA) and National Grid. This project is unique in that it involves four electrical generation owners and as many as six additional entities served by the CRM, with interconnections owned by National Grid, the regulated utility serving the area. A preliminary plan for the Potsdam CRM is shown in Fig. 2.
Fig. 2. Conceptual diagram of the Potsdam CRM.
2
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
Fig. 3. CRM market-based operation in grid-connected mode.
objective, a market-based operation and hierarchical control architecture can be deployed, with the CRM operator acting as the mini-ISO in the community market in charge of the CRM level electricity trading while maintaining system reliability. Furthermore, such a market operation and hierarchical control architecture enables flexible plug-andplay participation of other community partners interested in joining the CRM. The market-based operation of the CRM is motivated by the facts that CRM partners prefer extensive autonomy in scheduling their local resources during normal grid-connected operation status, while requiring an effectively coordinated energy trading mechanism. A market-based solution could ensure the fairness and transparency among CRM partners through a well-defined bidding structure, which grants participants options on whether and how much to bid into the CRM market under unified pricing criteria. In addition, such a marketbased operation needs to present a multiple-time-scale architecture to be seamlessly coordinated with the current wholesale market practice with both forward and real-time markets. In turn, CRM partners with quick-start generation or fast demand response flexibilities may adjust their bidding strategies in forward and real-time markets to pursue greater profits. Fig. 3 shows the proposed CRM market structure, which starts by collecting electricity sales/purchase bids from all CRM partners. The CRM operator will execute a market-clearing tool to clear the CRM market and determine optimal electricity sales/purchases quantities and prices for individual partners during each scheduling period (i.e., hourly or by the minute) for maximizing the community social benefits. Such a market-clearing tool should consider the facts of an unbalanced distribution network inside the CRM with unbalanced three-phase loads, untransposed line segments, and single- or two-phase laterals, as well as distinct operational characteristics associated with different distributed generation (DG) facilities owned by CRM partners, such as gas turbines, fuel cells, energy storage devices, and renewable resources (Liu et al., 2017). Important physical constraints such as minimum on/ off time limits of dispatchable DG units, exclusive charging/discharging/idle statuses of energy storage devices, and reconfigurable distribution feeders need to be modeled by introducing binary variables, which lead to a three-phase unbalanced AC-constrained unit commitment (ACUC) problem for clearing the CRM energy market. After the CRM market is cleared, per-phase locational market prices (PLMPs) are calculated and sent back to individual CRM partners. Optimal operation of the CRM under grid-connected mode is conducted through a multiple time-scale market architecture, which sets the most economical operation points for individual generation and demand response facilities of different CRM partners. Such information usually is exchanged on an hourly or several-minute basis in order to regulate power flows among multiple CRM partners as well as between the CRM and the upstream main grid. However, in recognizing the fact that CRM exhibits a variety of dynamical behaviors ranging from minutes and hours to milliseconds, such as real-time uncertainties of community customer loads and renewable energy resource outputs,
Clarkson University investigates an integrated control and communication framework for enhancing operations of CRM in both gridconnected and islanded modes. The NYDPS project led by GE Energy Consulting and partnered with Clarkson University intends to conduct a detailed engineering design and a business plan assessment of the Potsdam CRM. On the completion of those projects, funds will be sought to build the Potsdam CRM from a variety of sources, including the $40 million dollar New York Prize program (NYS, 2014). 3. CRM operation and hierarchical control strategies in gridconnected and islanded modes In recognizing the distinct structure and unique operating goals of CRMs with multiple owners of generators and critical loads, a marketbased operation and hierarchical control strategy is proposed for operating CRMs in grid-connected mode, in which each CRM partner acts as a self-interested entity and operates DERs/flexible loads according to its specific objectives. While in islanded mode during disruption events, a centralized energy management strategy is sought to effectively coordinate CRM partners and guarantee energy resiliency with limited generation resources and competing needs from multiple owners’ critical loads (Li et al., 2016). 3.1. Market-based operation and hierarchical control strategies in gridconnected mode When a CRM is operated in grid-connected mode, i.e., during bluesky days, CRM partners act as self-interested entities to operate their own generation and load resources, and interact with the CRM and/or the main grid to pursue distinct operating objectives, such as minimum electricity cost, maximum generation sales profits, or maximum power supply reliability. For instance, it is common practice for small-capacity partners purchase or sell electricity from or to the main grid through time-of-use (TOU) prices and power purchase agreements (PPA). Indeed, as emphasized in the New York REV Reforming the Energy Vision (REV) Initiative (REV, 2015), distribution-level participants can interact directly with New York Intendent System Operator (NYISO) programs, each CRM partner can make proper market participation decisions by interacting with the CRM operator, the local regulated utility company, and the bulk power system operator by optimally controlling their generation resources and flexible loads in multiple timescales. Specifically, from the CRM operator’s point of view, multiple CRM partners’ participation in the community-level market for energy sales/purchases, voltage regulation, and frequency stabilization can be economically coordinated, while preserving secure operation of the system and maximum community social benefits. From the utility system and bulk power system’s point of view, the CRM operator as a single entity represents all community resources devoted into the community markets, and such resources could be optimally used by the CRM operator to participate in the utility or bulk system’s energy/ancillary service markets. In order to fulfill such a CRM system operation 3
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
Fig. 4. Key functional blocks of the three-level hierarchical control.
critical load interruption using local resources while maintaining system reliability. Considering the fact that, during an emergency, not all community loads could be sufficiently supplied by onsite generation capacities, a critical load prioritization should be systematically identified and coordinated among all CRM partners. In islanded operation mode, the three-level hierarchical control architecture in Fig. 4 also applies for efficiently coordinating the power exchange among multiple DER owners and minimizing critical load interruption of the community. Specifically, (1) in the tertiary control layer, sufficient DERs and adjustable loads will be brought online by the strong cooperation among multiple parties, in order to enable a seamless transfer to islanded mode and enhance the resilient operation under disruption events; (2) in the secondary control layer, the synchronization control will be performed to seamlessly island and resynchronize the CRM system with the main grid; and (3) in the primary control layer, DERs will exhibit fast response to instantaneously match critical demands and stabilize CRM voltage/frequency in real time, when the CRM distribution network loses its voltage and/or frequency stability due to power mismatches subsequent to islanding events.
maintaining the CRM’s distribution system reliability in terms of voltage and frequency stability requires greater real-time response from dispatchable generation resources, flexible demand response assets, as well as the upstream utility grid support. Thus, a three-level hierarchical control architecture, as shown in Fig. 4, is deployed to realize secure and cost-effective operation and coordinate multiple partners in grid-connected mode. 3.1.1. Tertiary control Tertiary control ensures the economic operation of CRM by determining optimal set points of DERs and loads, and regulates power flows among multiple partners as well as between the community microgrid and the main grid at a temporal granularity level of hours and minutes. Specifically, with information from loads and DERs of multiple partners as well as the topological status of the distribution network, the tertiary control would execute unbalanced ACUC and unbalanced AC-constrained economical dispatch (ACED) for procuring cost-efficient scheduling of DERs, adjustable loads, and the main grid power transfer.Secondary control Secondary control is implemented to restore frequency and voltage levels of the CRM distribution network, and to compensate for deviations induced by the primary control at a temporal granularity level of seconds. Specifically, with a sudden change in demand, dispatches of DERs will be adjusted to balance demands. Once balanced, if frequency/voltage deviates from the rated value, the secondary control would generate a frequency/voltage compensation signal to restore the rated frequency/voltage with respect to secondary control parameters.
4. Self-organizing small cell based communication network for CRM operation and control The success of CRM heavily relies on the availability and efficiency of the underlying communication infrastructure. In CRMs, the combination of fast reaction time requirement in response to extreme events and the heterogeneity of data sources, together with their various operation/control constraints, call for a flexible, fast, and reliable communication network. Heterogeneous wireless networks (HetNet) provide broad coverage and ubiquity, which make them an ideal candidate for connecting CRM partners and physical assets. Yet, the co-existence of macro and small cells as well as the organization of small cells including cell identification, interference, and mobility management are challenging issues that need to be addressed even for the mobile user traffic. Indeed, the CRM traffic has tighter quality-of-service (QoS) expectations, which calls for urgently addressing these challenges. In particular, during an emergency, when the CRM operates in islanded mode, control decisions rely more on real-time information relaying capability of the communication network. The dynamic nature of the CRM calls for dynamic management of small cells. Indeed, by recognizing that manually configuring small cells is not practical, self-organization of small cells emerges as a primary requirement for providing a reliable and fast communications medium for the CRM. Self-organization, and more generally network management, has been studied in the context of long term evolution (LTE) and LTE-A networks. Network densification and difficulties related to manually configuring overlapping cells have been discussed in (Madueno et al., 2016; Zhang et al., 2011). The number of small cells is expected to grow significantly in the near future. Indeed, Nokia estimates that by 2020, wireless networks will be dominated by small cells immensely (Peng et al., 2013). In turn, as the number of cells increases, manually configuring, optimizing, and healing of the network become a tedious task and self-organization becomes essential. However, traditional self-organization techniques suffer from instantaneous network management decisions that do not consider predictability of certain
3.1.2. Primary control Primary control realizes active and reactive load sharing among parallel-connected DERs with plug-and-play capabilities, and stabilizes CRM distribution network voltage and frequency in real time. In recognizing that DERs are parallel-connected and most of them have limited capacities while being interfaced via power electronic converters, the most widely used primary control strategy is droop control. In the droop control scheme, dispatch levels of individual DERs are calculated based on droop characteristics as well as local frequency and voltage measurements. 3.2. Centralized operation and hierarchical control strategies in islanded mode When a CRM is operated in islanded mode during major disruption events, multiple CRM partners need to collaborate following CRM operator’s guidance, by sharing onsite generation capacities and flexible load curtailment/shifting capabilities to guarantee resilient energy supply to critical loads of all community partners, according to predefined agreements. Under such an emergency operation mode, a centralized operation and control architecture becomes necessary to effectively utilize multiple CRM partners’ resources for system resiliency and recovery, as shown in Fig. 5. All subscribed CRM partners need to grant the control of their generation and demand response resources to the CRM operator, and identify their respective critical and non-critical loads. From the CRM operator’s point of view, the sole operation objective in islanded mode is to minimize the community 4
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
Fig. 5. CRM centralized operation in islanded mode.
cell deployments for a CRM due to multiple ownerships. Therefore, small cells are expected to self-organize. Meanwhile, for meters, devices, and control equipment, cell association translates into reconfiguration of the topology. Self-organizing techniques, learning-based small cell pilot adjustment, and spectrum assignment in particular will allow reliable collection of information from multiple partners of CRM and reliable delivery of control commands. We aim to take advantage of the predictability of events and network traffic in CRM and effectively integrate them into the decision-making process. In particular, a predictive Q-learning-based pilot adjustment and spectrum assignment technique will be used. Q-learning is a reinforcement learning technique which is used to identify optimal action-selection policies for Markov Reward Processes (Morozs Grace and Clarke, 2013). In Qlearning, the algorithm for most of the time selects an action that maximizes Q, while random action is taken in certain steps. The Q function is updated accordingly and each successful step is rewarded. Conventional Q-learning is concerned with current states and actions, while neglecting the prediction of the future. In turn, we will take learning one step further and integrate prediction capability into learning for addressing two network planning problems: (1) optimal pilot adjustment and (2) optimal spectrum assignment for a small-cellenabled CRM. Besides organizing small cells based on prediction-integrated learning, data aggregation plays an important role in the manageability and the recurring costs of the network (Alorainy et al., 2016; ErRahmadi et al., 2016). Most control actions in a CRM, especially in islanded mode, need near-real-time communications. Yet collecting a large amount of data from DERs/loads and delivering control
traffic types, in particular the traffic generated by CRM devices. The increasing density of small cells has been recognized in (Zhang et al., 2016), and a dynamic stochastic game has been formulated to optimally control the transmit power of each base station. Distributed resource management for small cell wireless networks has also been studied in (Rahmati et al., 2017; Samarakoon et al., 2016) using mathematical optimization approaches as well as a matching game strategy, whereas metaheuristics have been studied for self-organized orthogonal resource allocation challenges in small cell networks (Ahmed and Tirkkonen, 2016). Furthermore, inter- and intra-cell interference, multi-cell cooperative communication, cooperating cell selection, and improved resource block utilization have been identified as the crucial challenges in small-cells-based communications (Chang and Liou, 2017). Recently, learning approaches have been considered for load balancing and interference management (Xu et al., 2014; Bennis et al., 2013). The communication infrastructure of CRMs must be flexibly selforganizing in response to real-time dynamic changes in the network topology and operational status of CRMs, such as joining or leaving the status of plug-and-play DERs corresponding to cell association as well as adjustments on data sampling and control signal rates based on the communication network traffic. Thus, a novel prediction-integrated learning approach is explored in order to enhance self-organization of small cells. The conceptual diagram of the small-cell-enabled Potsdam CRM is illustrated in Fig. 6. As shown in this figure, multiple DERs and consumers in the CRM are interconnected through a mix of small cell base stations and macro cell base stations based on LTE, in which small cells can improve spectrum efficiency, capacity, and coverage of traditional wireless networks. In practice, it would be difficult to pre-specify small
Fig. 6. Conceptual diagram of the Potsdam CRM with overlaid communication network.
5
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
commands using the same communication infrastructure could undermine the real-time requirement. Thus, we will leverage cell association in small cells and integrate multi-scale data aggregation to form multiple paths for data streams with varying QoS requirements. Indeed, in a CRM, packets destined to controllers will have the highest priority, while data collected from meters and loads allow aggregation and may also be treated as relatively delay-tolerant data streams. Adaptive algorithms are needed to aim at optimal aggregation of data at the ingress point. In turn, the adaptive algorithms will observe the self-organized small cell and the backhaul to compute the optimal path for each data stream. This can be done periodically at the aggregation intervals, as well as when the underlying CRM topology changes due to scheduled or unexpected events. Alternatively, computing the optimal path for data streams can be adapted through predictive analysis of the network status, which would periodically predict data rates, service coverage, and capacity (Letourneux et al., 2014).
•
•
5. An integrated reconfigurable control and self-organizing communication framework In order to enhance resilient and economic operation of CRMs, operation/control strategies and the overlaid communication infrastructure must be effectively coordinated in a flexible manner. Thus, we plan to investigate an integrated reconfigurable control and self-organizing communication framework for exploring their interactions, so that one can be adaptively adjusted in response to real-time changes of the other.
5.2. Reconfiguring self-organizing communication based on real-time CRM operation status The self-organizing communication network is used to monitor CRM operations and gather data at multiple time-scales, and to actuate controllers when needed. A common assumption is that the communication network is static and the central communication network manager has global information on the network topology and link qualities. However, such assumption may not be realistic in the CRM environment because of the complex heterogeneity of data sources and dynamic plug-and-play characters of DERs. Indeed, the communication network must be adaptively self-organized to synchronize with the topology and operation status of CRM. In order to design an effective way for adaptively reconfiguring the communication network according to the real-time topology and operation status of CRM, the following questions need be answered:
5.1. Reconfiguring centralized/distributed control based on real-time communication status Centralized and distributed operation/control strategies present different communication complexity and varying performances in terms of global optimality. Indeed, the extensive utilization of centralized control may deteriorate the communication network performance, make the time scales of control and communication unsynchronized, and in turn jeopardize the resilient and economic operation of CRMs. Thus, an adaptive scheme is needed to allow flexible switching between centralized and distributed strategies according to the real-time communication performance. Specifically, the centralized control strategy is used for pursuing global optimality to the maximum extend as long as the real-time communication performance is not compromised, and the distributed control strategy is adopted to critical applications for ensuring resiliency when the communication performance is degraded. Adaptive schemes that combine optimality of centralized control and resiliency of distributed control could provide robust operation and control against communication failures or degraded performance of communication links. Indeed, in order to design an effective way for adaptively reconfiguring the centralized and distributed control strategies according to real-time congestion and latency status of the communication network, the following questions need to be answered:
• How could the radio resource allocation be dynamically adjusted?
•
• How can the computation and communication complexities of dis-
•
capability for effectively supporting centralized computation and distributed control strategies. How can adaptive control strategies mitigate the impact of non-ideal communication links such as message loss and delay? Although a communication network design can incorporate mechanisms to achieve high reliability, one cannot assume that all the information will always be delivered without delays or losses all the time. Thus, the relationship between the convergence rate of the distributed algorithm, the restorability of network topologies, and the message loss rate needs to be quantified. It is worth noting that restorability of network topologies needs to be formulated as a function of redundancy in these networks (Gurel and Keskin, 2017). How could major disasters impact the communication network and result in partial disconnection of certain networks? To evaluate such situations and avoid their consequences, a survivable and selfhealing design for the critical loads and suppliers needs to be in place. In addition, CRM operation needs to be robust under performance degradation of the communication network due to failures.
tributed and centralized information and control be optimally leveraged? This would guarantee a robust performance in response to volatility and uncertainty in CRM and the communication network, as well as enhance resiliency and economic efficiency in both gridconnected and islanded modes. What are the optimal thresholds to adaptively switch between centralized and distributed controls? The thresholds could be quantified in terms of availability and reliability of the communication network under multi-failure conditions, including its topology and key parameters, which would reflect the communication
6
Although the cooperation among multiple partners in the CRM is enabled by the communication network, individual partners may present different levels of autonomy and authority, asymmetric data generation rates, and various objectives such as minimum loss, minimum latency, or minimum power consumption. Indeed, communication requirements including reliability, response time, latency, and sampling rates may be different, and a real-time dynamically configurable communication platform is needed to provide rapid flexibility. Furthermore, load patterns of the CRM, new equipment association, contingencies, weather conditions, and extreme events all necessitate the reconfiguration of the communication network. In this perspective, the communication platform needs to be self-organized for providing optimal reliability, maximum availability, and the shortest possible latency. What are the optimal frequency and aggressiveness levels of the data aggregation strategy? Data gathered from heterogeneous DERs, loads, and control equipment of multiple partners in the CRM need to be aggregated at several levels and prioritized based on the dynamic operation status of the CRM. This can be correlated with DERs’ outputs and load profiles. The adaptive data aggregation algorithm can operate in peak and off-peak regimes, as well as normal and extraordinary operating conditions. Observing the real-time CRM operation, data generated by equipment, and control commands released by operators, the adaptive data aggregation from multiple sensing points can be fine-tuned. Aggregation window size and variation of the data stream will play the key role in
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
•
•
communication infrastructure of CRMs. The article hopes to bring more attention from energy and utility industry, academia, and power grid agencies to investigate effective operational and control tools for the effective development and widespread deployment of CRMs to improve disaster response and recovery for our society.
determining optimal frequency and aggressiveness levels of the data aggregation strategy (Kolozali et al., 2016). It is worth mentioning that real-time distributed data aggregation does not only experience computational challenges but also increases communications overhead. As reported by a related work (Riker et al., 2016), adopting a multi-tier approach in data aggregation may improve computational and communication efficiency. That being said, data redundancy can be handled at one tier whereas the communications overhead can be handled by the second tier. How can low-complexity algorithms be designed to effectively perform the dynamic self-organization of a communications network in real time? The communications network reconfiguration needs to be done fast enough for meeting the stringent latency requirements; for instance, primary control cannot tolerate delays beyond milliseconds. Reconfiguration of small cell networks can dynamically release or acquire radio resources to or through macro cells in real time. In turn, self-organization and data aggregation addressed above need to be implemented with low-complexity algorithms. Indeed, as mentioned above, introducing an initial tier that would eliminate redundant data can help reduce excessive computational loads on the aggregation processes. How can small cell networks in the self-organization communications architecture be configured in a resilient manner? Data gathering and control signal delivery should not be affected in the presence of any failure. However, major disasters may cause communication outage in multiple cells. This calls for recovery using macro cells or adjacent cells which both require the fast handover of the devices as well as the context. To prevent abrupt outage of communications links for critical CRM components, resource allocation schemes can dynamically vacate mobile user traffic. This has to be done diligently as wireless networks and small cells are planned to be used for public safety in a close future. Therefore, how to allocate radio resources to critical parties under failures of single or multiple cells will need to be addressed. In addition, the small cell network needs to be self-healing and recover from failures with minimum manual configuration.
References Ahmed, F., Tirkkonen, O., 2016. Simulated annealing variants for self-organized resource allocation in small cell networks. Appl. Soft Comput. 38, 762–770. Alorainy, A., Hossain, M.J., Alouini, M.S., 2016. Multi-flow carrier aggregation in heterogeneous networks: cross-layer performance analysis. In: Washington, DC, USA. IEEE Globecom Workshops (GC Wkshps) 2016. pp. 1–7. Bennis, M., Perlaza, S.M., Blasco, P., Han, Z., Poor, H.V., 2013. Self-Organization in small cell networks: a reinforcement learning approach IEEE. Trans. Wireless Commun. 2 (July (7)), 3202–3212. Canadian Disaster Database, https://www.publicsafety.gc.ca/cnt/rsrcs/cndn-dsstr-dtbs/ index-en.aspx. Chang, B., Liou, S., 2017. Adaptive cooperative communication for maximizing reliability and reward in ultra-dense small cells LTE-A toward 5G cellular networking. Comput. Networks 115 (March), 16–28 Early Access: January 25. Design of a Resilient Underground Microgrid in Potsdam, NY. NYSERDA Project 41309. Clarkson University prime contractor. Er-Rahmadi, B., Bagaa, M., Ksentini, A., Meddour, D.E., 2016. Cost-efficient data aggregation schemes for small cell networks. In: 2016 International Wireless Communications and Mobile Computing Conference (IWCMC). Paphos. pp. 77–83. Gurel, G., Keskin, S., 2015. Emergency recovery for small cell management system. 23rd Signal Processing and Communications Applications Conference. Kolozali, S., Puschmann, D., Bermudez-Edo, M., Barnaghi, P., 2016. On the effect of adaptive and nonadaptive analysis of time-series sensory data. IEEE IoT J. 3 (December (6)), 1084–1098. Letourneux, F., Guivarch, S., Lostanlen, Y., 2014. Impact of modeling tools on outdoor small-cell deployment cost in a realistic urban scenario. In: Vancouver, BC. 2014 IEEE 80th Vehicular Technology Conference (VTC2014-Fall) 2. pp. 1–5. Li, J., Liu, Y., Wu, L., 2016. Optimal operation for community based multi-party microgrid in grid-connected and islanded modes. IEEE Trans. Smart Grid. http://dx.doi. org/10.1109/TSG.2016.2564645. Liu, Y., Li, J., Wu, L., 2017. Distribution system restructuring: distribution LMP via unbalanced ACOPF. IEEE Trans. Smart Grid. http://dx.doi.org/10.1109/TSG.2016. 2647692. Madueno, G., Nielsen, J., Kim, D., Pratas, N., Stefanovic, C., Popovski, P., 2016. Assessment of LTE wireless access for monitoring of energy distribution in the smart grid. IEEE J. Sel. Areas Commun. 34 (3), 675–688. Morozs Grace, N.D., Clarke, T., 2013. Case-based reinforcement learning for cognitive spectrum assignment in cellular networks with dynamic topologies. In: 2013 Military Communications and Information Systems Conference. St.-Malo. pp. 1–6. NOAA, 2013. Hurricane/post-tropical Cyclone Sandy, October 22–29, 2012. National Oceanic and Atmospheric Administration. June 2013. [Online]. Available: http:// www.nws.noaa.gov/os/assessments/pdfs/Sandy13.pdf. NOAA, 2014. Billion-dollar Weather/climate Disasters. National Oceanic and Atmospheric Administration. January 2014. [Online]. Available: http://www.ncdc. noaa.gov/billions/events. Governor Cuomo announces NY Prize resiliency competition to launch this Fall, Press Release, New York State, Aug. 26, 2014. Peng, M., Liang, D., Wei, Y., Li, J., Chen, H., 2013. Self-configuration and self-optimization in LTE-advanced heterogeneous networks. IEEE Commun. Mag. 51 (May (5)), 36–45. REV, 2015. New York State Department of Public Service–Reforming the Energy Vision. http://www3.dps.ny.gov/W/PSCWeb.nsf/All/ 26BE8A93967E604785257CC40066B91A?OpenDocument. Rahmati, A., Shah-Mansouri, V., Safari, M., 2017. Price-based resource allocation for selfbackhauled small cell networks. Comput. Commun. 97 (January), 72–80. Riker, A., Cerqueira, E., Curado, M., Monteiro, E., 2016. A two-tier adaptive data aggregation approach for M2 M group-Communication. IEEE Sens. J. 16 (3), 823–835 Feb. 1. Ross, T., Lott, N., 2016. A climatology of recent extreme weather and climatic events, National Climatic Data Center. Techn. Rep. 2000–2002. Samarakoon, S., Bennis, M., Saad, W., Debbah, M., Latva-aho, M., 2016. Ultra dense small cell networks: turning density into energy efficiency. IEEE J. Sel. Areas Commun. 34 (5), 1267–1280. Xu, J., Tang, L., Chen, Q., Yi, L., 2014. Study on based reinforcement Q-learning for mobile load balancing techniques in LTE-A HetNets. IEEE 17th International Conference on Computational Science and Engineering, Che. Zhang, G., Zhang, G., Gao, Y., Lu, J., 2011. Competitive strategic bidding optimization in electricity markets using bilevel programming and swarm technique. IEEE Trans. Ind. Electron. 58 (June (6)), 2138–2146. Zhang, H., Jiang, C., Hu, R., Qian, Y., 2016. Self-organization in disaster-resilient heterogeneous small cell networks. IEEE, Network(March/April).
6. Final words The successful development and widespread deployment of effective CRM technology have the potential to provide significant improvements for disaster response and recovery for our society. They will also have a profound impact on communities during normal operating conditions, including additional revenues for DER owners derived from operational knowledge and efficiencies. However, in order to enable the full functionality of CRMs for providing seamlessly resilient electric service to the recovery effort following catastrophes while ensuring electrical, economic, and societal goals, the communications network is of critical value in CRM operation and control. Specifically, CRM operators must have access to real-time CRM operation information gleaned from high-fidelity sensors so as to guarantee satisfactory realtime performance on enhancing the resilient and economic operation under both normal situations and during major disruption events. However, the communications network could be inadequate to the challenges of moving large amounts of data in real time to the places they are needed, and in turn control and communication are not synchronized. To address such critical challenges, this article discussed unique operational characteristics of CRMs across multiple timescales in different operation modes, and presented coordinated control and communication solutions for the resilient and economic operation of CRMs. A practical CRM project that is underway for serving Potsdam, N.Y., was also presented. The article further outlined several important issues related to the design of an integrated reconfigurable control and selforganizing communications framework, in order to effectively and adaptively coordinate operation/control strategies and the overlaid
Lei Wu is an Associate Professor of Electrical and Computer Engineering at Clarkson University, Potsdam, New York, with research interests that include power systems operation and planning, energy economics, and community resilience microgrid. He received a Ph.D. in Electrical Engineering from Illinois Institute of Technology (IIT),
7
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
IEEE’s Technical Committee on Green Communications and Computing and the research group leader for IEEE Smart Grid and Big Data Standardization. Her main research interests are wireless networks, the smart grid, cyber-physical systems, electric vehicles and wireless sensor networks.
Chicago, in 2008. From 2008 to 2010, he was a Senior Research Associate with the Robert W. Galvin Center for Electricity Innovation at IIT. He also worked as summer Visiting Faculty at the New York Independent System Operator (NYISO) in 2012. Jie Li is an Assistant Professor in the Electrical and Computer Engineering Department at Clarkson University in Potsdam, New York. She received a Ph.D. in Electrical Engineering from the Illinois Institute of Technology (IIT), Chicago, in 2012. From 2006 to 2008, she was a Research Engineer with IBM China Research Lab. From 2012 to 2013, she was a Power System Application Engineer with GE Energy Consulting. Her research interests include power systems restructuring and bidding strategy.
Burak Kantarci is an Assistant Professor at the School of Electrical Engineering and Computer Science at the University of Ottawa. Prior to joining the University of Ottawa, he was an Assistant Professor at Clarkson University, Potsdam, New York. He received his Ph.D. in Computer Engineering from Istanbul Technical University in 2009. He is the founding director of the Next Generation Communications and Computing Networks (NEXTCON) research group. He is currently an editor for IEEE Transactions on Green Communications and Networking, editor for IEEE Communications Surveys and Tutorials, and Secretary of IEEE’s Communication Systems Integration and Modelling Technical Committee. His main research interests are mobile computing and communications, the internet of things, wireless sensor networks, digital health, and cloud computing.
Melike Erol-Kantarci is an Assistant Professor at the School of Electrical Engineering and Computer Science at the University of Ottawa. Prior to joining the University of Ottawa, she was an Assistant Professor at Clarkson University, Potsdam, New York. She received a Ph.D. in Computer Engineering from Istanbul Technical University in 2009. She is currently the vice-chair of the Green Smart Grid Communications special interest group of
8
Contents lists available at ScienceDirect
The Electricity Journal journal homepage: www.elsevier.com/locate/tej
Special Issue: Contemporary Strategies for Microgrid Operation & Control
An integrated reconfigurable control and self-organizing communication framework for community resilience microgrids ⁎
Lei Wua, , Jie Lia, Melike Erol-Kantarcib, Burak Kantarcib a b
Electrical and Computer Engineering Department, Clarkson University, Potsdam, New York, United States School of Electrical Engineering and Computer Science, University of Ottawa, Ottawa, Canada
A R T I C L E I N F O
A B S T R A C T
Keywords: Community resilience micrgrid distributed energy resource resilience integrated control and communication
CRMs can effectively share distributed energy resources of multiple owners and enhance resilient electricity supply in communities during disruptions. However, in contrast to single-entity microgrids, CRMs present a unique structure and bring new challenges for operations and control, which can often be interrupted due to cascaded failures in interconnected electrical/communication components. An installation in Potsdam, New York, incorporates some control/communication solutions for resilient and economic operation of CRMs.
1. Introduction of community resilience microgrids There is an increasing frequency of catastrophic weather events in the United States and globally, which inflict serious social and economic impacts. The critical issue associated with such catastrophes is the availability of electricity for recovery efforts. In fact, according to a report by the National Oceanic and Atmospheric Administration (NOAA), the U.S. has sustained 188 weather and climate disasters since 1980, more than 80% of them bringing damages to the nation’s electricity infrastructure (NOAA, 2014). The most notable example is Hurricane Sandy in 2012, which cost some 8.5 million customers their power during the storm and its aftermath (NOAA, 2013). Similar incidents, mostly due to ice storms, were also observed in Canada (Canadian Disaster Database, 2017) and other places around the globe. Fortunately, as an emerging technology, community resilience microgrids (CRM) can enhance resilient electricity supply to critical loads in the community during such disruption events. A CRM includes multiple distributed energy resources (DER) and critical loads that are owned and controlled by individual entities within a clearly defined electrical boundary, which are connected via primary distribution lines owned by a local regulated power company. Indeed, as shown in Fig. 1, CRMs extend benefits of traditional single-entity, behind-the-meter microgrid systems by effectively sharing DERs among multiple partners within a community, which could enhance energy resiliency by reducing the overall impact of and improving response and recovery to critical events for the entire community. However, a CRM presents a unique structure and brings new challenges for its operation and control. Specifically, DERs owned and controlled by individual partners can either represent individual ⁎
stakeholders’ financial interests when the CRM is operated in gridconnected mode, or collaborate as a single controllable entity for enhancing resilient electricity supply to critical loads when CRM is islanded. That is, in grid-connected mode, each CRM partner acts as a self-interested entity and operates DERs/flexible loads according to its specific objectives; during disruption events, on the other hand, CRM partners can coordinate with each other by sharing onsite generation capacities and flexible load curtailment/shifting capabilities, in order to supply community critical loads based on certain agreements. Thus, in different CRM operation modes, partners in a CRM would act as either self-interested competitors or as altruistic collaborators following the guidance of the CRM operator. Furthermore, CRMs are complex networked systems, which not only interconnect DERs and loads of multiple owners through distribution lines but also are overlaid with a communication and control system that gathers inputs and sends out control signals to multiple owners for enabling resilient and economic operation. Indeed, CRM operations can often be interrupted or halted due to the cascaded growth of failures in interconnected electrical/communication components or unwillingness/inability of individual CRM partners to respond. Thus, in order to enable the full functionality of CRMs, an integrated control and communication framework is needed for facilitating CRM operations while complying with system control and communication requirements under different operation modes. This article discusses the unique operational characteristics of CRMs across multiple timescales in different operation modes, and presents coordinated system control and communication solutions for enhancing resilient and economic operation of CRMs. Section 2 discusses a practical CRM project that is underway for serving the city of Potsdam, N.Y.
Corresponding author. E-mail address: [email protected] (L. Wu).
http://dx.doi.org/10.1016/j.tej.2017.03.011
1040-6190/ © 2017 Elsevier Inc. All rights reserved.
Please cite this article as: Wu, L., The Electricity Journal (2017), http://dx.doi.org/10.1016/j.tej.2017.03.011
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
Fig. 1. Single-entity microgrid versus CRM.
Catastrophic weather events in Upstate New York, such as ice storms, major snow events, and micro-burst wind events, have historically caused widespread damages. The risks posed to residents and the coordination of emergency services are extremely challenging. One of the most devastating events was the ice storm of 1998, which affected most of northern New York State, southern Canada, and northern New England, and caused over $1.4 billion in damages/costs and 16 deaths in the U.S. (Ross and Lott, 2016). In fact, the availability of a functioning CRM after the disruption of the main grid during these events can enable operations of critical services and mobilization of first responders. Such a CRM can also furnish a functional staging area for teams to perform disaster recovery tasks in the region and address the immediate needs of the most affected population. The Potsdam CRM project is led by Clarkson University, partners with GE Energy Consulting, Nova Energy Specialists, and National Grid. Subsequently, during 2014 to 2017, four synergistic projects have been funded by U.S. DOE, NSF, and New York State Department of Public Service (NYDPS). The DOE project led by GE Global Research funds the development and testing of a microgrid controller, in which Clarkson is subcontractor. One NSF project led by Clarkson University focuses on the human-machine operational impacts of the CRM during normal operations and disaster response, while the other NSF project led by
Section 3 evaluates different operation strategies of a CRM in order to optimally manage dispatches of DERs in grid-connected mode and enhance resiliency in islanded mode. Section 4 presents a self-organizing, small cell-based infrastructure for ensuring flexible, fast, and reliable communication requirements of CRM operations. Section 5 describes an integrated reconfigurable control and self-organizing communication framework to study the interdependency and interaction between control strategies and communication requirements. The paper is concluded in Section 6.
2. Current efforts of the CRM Project in Potsdam The development of CRM technology to address disaster response is currently underway through multiple projects across the U.S. One of such projects is the “Design of a Resilient Underground Microgrid in Potsdam, NY,” (Clarkson University prime contractor, 2014) jointly funded by the New York State Energy Research and Development Authority (NYSERDA) and National Grid. This project is unique in that it involves four electrical generation owners and as many as six additional entities served by the CRM, with interconnections owned by National Grid, the regulated utility serving the area. A preliminary plan for the Potsdam CRM is shown in Fig. 2.
Fig. 2. Conceptual diagram of the Potsdam CRM.
2
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
Fig. 3. CRM market-based operation in grid-connected mode.
objective, a market-based operation and hierarchical control architecture can be deployed, with the CRM operator acting as the mini-ISO in the community market in charge of the CRM level electricity trading while maintaining system reliability. Furthermore, such a market operation and hierarchical control architecture enables flexible plug-andplay participation of other community partners interested in joining the CRM. The market-based operation of the CRM is motivated by the facts that CRM partners prefer extensive autonomy in scheduling their local resources during normal grid-connected operation status, while requiring an effectively coordinated energy trading mechanism. A market-based solution could ensure the fairness and transparency among CRM partners through a well-defined bidding structure, which grants participants options on whether and how much to bid into the CRM market under unified pricing criteria. In addition, such a marketbased operation needs to present a multiple-time-scale architecture to be seamlessly coordinated with the current wholesale market practice with both forward and real-time markets. In turn, CRM partners with quick-start generation or fast demand response flexibilities may adjust their bidding strategies in forward and real-time markets to pursue greater profits. Fig. 3 shows the proposed CRM market structure, which starts by collecting electricity sales/purchase bids from all CRM partners. The CRM operator will execute a market-clearing tool to clear the CRM market and determine optimal electricity sales/purchases quantities and prices for individual partners during each scheduling period (i.e., hourly or by the minute) for maximizing the community social benefits. Such a market-clearing tool should consider the facts of an unbalanced distribution network inside the CRM with unbalanced three-phase loads, untransposed line segments, and single- or two-phase laterals, as well as distinct operational characteristics associated with different distributed generation (DG) facilities owned by CRM partners, such as gas turbines, fuel cells, energy storage devices, and renewable resources (Liu et al., 2017). Important physical constraints such as minimum on/ off time limits of dispatchable DG units, exclusive charging/discharging/idle statuses of energy storage devices, and reconfigurable distribution feeders need to be modeled by introducing binary variables, which lead to a three-phase unbalanced AC-constrained unit commitment (ACUC) problem for clearing the CRM energy market. After the CRM market is cleared, per-phase locational market prices (PLMPs) are calculated and sent back to individual CRM partners. Optimal operation of the CRM under grid-connected mode is conducted through a multiple time-scale market architecture, which sets the most economical operation points for individual generation and demand response facilities of different CRM partners. Such information usually is exchanged on an hourly or several-minute basis in order to regulate power flows among multiple CRM partners as well as between the CRM and the upstream main grid. However, in recognizing the fact that CRM exhibits a variety of dynamical behaviors ranging from minutes and hours to milliseconds, such as real-time uncertainties of community customer loads and renewable energy resource outputs,
Clarkson University investigates an integrated control and communication framework for enhancing operations of CRM in both gridconnected and islanded modes. The NYDPS project led by GE Energy Consulting and partnered with Clarkson University intends to conduct a detailed engineering design and a business plan assessment of the Potsdam CRM. On the completion of those projects, funds will be sought to build the Potsdam CRM from a variety of sources, including the $40 million dollar New York Prize program (NYS, 2014). 3. CRM operation and hierarchical control strategies in gridconnected and islanded modes In recognizing the distinct structure and unique operating goals of CRMs with multiple owners of generators and critical loads, a marketbased operation and hierarchical control strategy is proposed for operating CRMs in grid-connected mode, in which each CRM partner acts as a self-interested entity and operates DERs/flexible loads according to its specific objectives. While in islanded mode during disruption events, a centralized energy management strategy is sought to effectively coordinate CRM partners and guarantee energy resiliency with limited generation resources and competing needs from multiple owners’ critical loads (Li et al., 2016). 3.1. Market-based operation and hierarchical control strategies in gridconnected mode When a CRM is operated in grid-connected mode, i.e., during bluesky days, CRM partners act as self-interested entities to operate their own generation and load resources, and interact with the CRM and/or the main grid to pursue distinct operating objectives, such as minimum electricity cost, maximum generation sales profits, or maximum power supply reliability. For instance, it is common practice for small-capacity partners purchase or sell electricity from or to the main grid through time-of-use (TOU) prices and power purchase agreements (PPA). Indeed, as emphasized in the New York REV Reforming the Energy Vision (REV) Initiative (REV, 2015), distribution-level participants can interact directly with New York Intendent System Operator (NYISO) programs, each CRM partner can make proper market participation decisions by interacting with the CRM operator, the local regulated utility company, and the bulk power system operator by optimally controlling their generation resources and flexible loads in multiple timescales. Specifically, from the CRM operator’s point of view, multiple CRM partners’ participation in the community-level market for energy sales/purchases, voltage regulation, and frequency stabilization can be economically coordinated, while preserving secure operation of the system and maximum community social benefits. From the utility system and bulk power system’s point of view, the CRM operator as a single entity represents all community resources devoted into the community markets, and such resources could be optimally used by the CRM operator to participate in the utility or bulk system’s energy/ancillary service markets. In order to fulfill such a CRM system operation 3
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
Fig. 4. Key functional blocks of the three-level hierarchical control.
critical load interruption using local resources while maintaining system reliability. Considering the fact that, during an emergency, not all community loads could be sufficiently supplied by onsite generation capacities, a critical load prioritization should be systematically identified and coordinated among all CRM partners. In islanded operation mode, the three-level hierarchical control architecture in Fig. 4 also applies for efficiently coordinating the power exchange among multiple DER owners and minimizing critical load interruption of the community. Specifically, (1) in the tertiary control layer, sufficient DERs and adjustable loads will be brought online by the strong cooperation among multiple parties, in order to enable a seamless transfer to islanded mode and enhance the resilient operation under disruption events; (2) in the secondary control layer, the synchronization control will be performed to seamlessly island and resynchronize the CRM system with the main grid; and (3) in the primary control layer, DERs will exhibit fast response to instantaneously match critical demands and stabilize CRM voltage/frequency in real time, when the CRM distribution network loses its voltage and/or frequency stability due to power mismatches subsequent to islanding events.
maintaining the CRM’s distribution system reliability in terms of voltage and frequency stability requires greater real-time response from dispatchable generation resources, flexible demand response assets, as well as the upstream utility grid support. Thus, a three-level hierarchical control architecture, as shown in Fig. 4, is deployed to realize secure and cost-effective operation and coordinate multiple partners in grid-connected mode. 3.1.1. Tertiary control Tertiary control ensures the economic operation of CRM by determining optimal set points of DERs and loads, and regulates power flows among multiple partners as well as between the community microgrid and the main grid at a temporal granularity level of hours and minutes. Specifically, with information from loads and DERs of multiple partners as well as the topological status of the distribution network, the tertiary control would execute unbalanced ACUC and unbalanced AC-constrained economical dispatch (ACED) for procuring cost-efficient scheduling of DERs, adjustable loads, and the main grid power transfer.Secondary control Secondary control is implemented to restore frequency and voltage levels of the CRM distribution network, and to compensate for deviations induced by the primary control at a temporal granularity level of seconds. Specifically, with a sudden change in demand, dispatches of DERs will be adjusted to balance demands. Once balanced, if frequency/voltage deviates from the rated value, the secondary control would generate a frequency/voltage compensation signal to restore the rated frequency/voltage with respect to secondary control parameters.
4. Self-organizing small cell based communication network for CRM operation and control The success of CRM heavily relies on the availability and efficiency of the underlying communication infrastructure. In CRMs, the combination of fast reaction time requirement in response to extreme events and the heterogeneity of data sources, together with their various operation/control constraints, call for a flexible, fast, and reliable communication network. Heterogeneous wireless networks (HetNet) provide broad coverage and ubiquity, which make them an ideal candidate for connecting CRM partners and physical assets. Yet, the co-existence of macro and small cells as well as the organization of small cells including cell identification, interference, and mobility management are challenging issues that need to be addressed even for the mobile user traffic. Indeed, the CRM traffic has tighter quality-of-service (QoS) expectations, which calls for urgently addressing these challenges. In particular, during an emergency, when the CRM operates in islanded mode, control decisions rely more on real-time information relaying capability of the communication network. The dynamic nature of the CRM calls for dynamic management of small cells. Indeed, by recognizing that manually configuring small cells is not practical, self-organization of small cells emerges as a primary requirement for providing a reliable and fast communications medium for the CRM. Self-organization, and more generally network management, has been studied in the context of long term evolution (LTE) and LTE-A networks. Network densification and difficulties related to manually configuring overlapping cells have been discussed in (Madueno et al., 2016; Zhang et al., 2011). The number of small cells is expected to grow significantly in the near future. Indeed, Nokia estimates that by 2020, wireless networks will be dominated by small cells immensely (Peng et al., 2013). In turn, as the number of cells increases, manually configuring, optimizing, and healing of the network become a tedious task and self-organization becomes essential. However, traditional self-organization techniques suffer from instantaneous network management decisions that do not consider predictability of certain
3.1.2. Primary control Primary control realizes active and reactive load sharing among parallel-connected DERs with plug-and-play capabilities, and stabilizes CRM distribution network voltage and frequency in real time. In recognizing that DERs are parallel-connected and most of them have limited capacities while being interfaced via power electronic converters, the most widely used primary control strategy is droop control. In the droop control scheme, dispatch levels of individual DERs are calculated based on droop characteristics as well as local frequency and voltage measurements. 3.2. Centralized operation and hierarchical control strategies in islanded mode When a CRM is operated in islanded mode during major disruption events, multiple CRM partners need to collaborate following CRM operator’s guidance, by sharing onsite generation capacities and flexible load curtailment/shifting capabilities to guarantee resilient energy supply to critical loads of all community partners, according to predefined agreements. Under such an emergency operation mode, a centralized operation and control architecture becomes necessary to effectively utilize multiple CRM partners’ resources for system resiliency and recovery, as shown in Fig. 5. All subscribed CRM partners need to grant the control of their generation and demand response resources to the CRM operator, and identify their respective critical and non-critical loads. From the CRM operator’s point of view, the sole operation objective in islanded mode is to minimize the community 4
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
Fig. 5. CRM centralized operation in islanded mode.
cell deployments for a CRM due to multiple ownerships. Therefore, small cells are expected to self-organize. Meanwhile, for meters, devices, and control equipment, cell association translates into reconfiguration of the topology. Self-organizing techniques, learning-based small cell pilot adjustment, and spectrum assignment in particular will allow reliable collection of information from multiple partners of CRM and reliable delivery of control commands. We aim to take advantage of the predictability of events and network traffic in CRM and effectively integrate them into the decision-making process. In particular, a predictive Q-learning-based pilot adjustment and spectrum assignment technique will be used. Q-learning is a reinforcement learning technique which is used to identify optimal action-selection policies for Markov Reward Processes (Morozs Grace and Clarke, 2013). In Qlearning, the algorithm for most of the time selects an action that maximizes Q, while random action is taken in certain steps. The Q function is updated accordingly and each successful step is rewarded. Conventional Q-learning is concerned with current states and actions, while neglecting the prediction of the future. In turn, we will take learning one step further and integrate prediction capability into learning for addressing two network planning problems: (1) optimal pilot adjustment and (2) optimal spectrum assignment for a small-cellenabled CRM. Besides organizing small cells based on prediction-integrated learning, data aggregation plays an important role in the manageability and the recurring costs of the network (Alorainy et al., 2016; ErRahmadi et al., 2016). Most control actions in a CRM, especially in islanded mode, need near-real-time communications. Yet collecting a large amount of data from DERs/loads and delivering control
traffic types, in particular the traffic generated by CRM devices. The increasing density of small cells has been recognized in (Zhang et al., 2016), and a dynamic stochastic game has been formulated to optimally control the transmit power of each base station. Distributed resource management for small cell wireless networks has also been studied in (Rahmati et al., 2017; Samarakoon et al., 2016) using mathematical optimization approaches as well as a matching game strategy, whereas metaheuristics have been studied for self-organized orthogonal resource allocation challenges in small cell networks (Ahmed and Tirkkonen, 2016). Furthermore, inter- and intra-cell interference, multi-cell cooperative communication, cooperating cell selection, and improved resource block utilization have been identified as the crucial challenges in small-cells-based communications (Chang and Liou, 2017). Recently, learning approaches have been considered for load balancing and interference management (Xu et al., 2014; Bennis et al., 2013). The communication infrastructure of CRMs must be flexibly selforganizing in response to real-time dynamic changes in the network topology and operational status of CRMs, such as joining or leaving the status of plug-and-play DERs corresponding to cell association as well as adjustments on data sampling and control signal rates based on the communication network traffic. Thus, a novel prediction-integrated learning approach is explored in order to enhance self-organization of small cells. The conceptual diagram of the small-cell-enabled Potsdam CRM is illustrated in Fig. 6. As shown in this figure, multiple DERs and consumers in the CRM are interconnected through a mix of small cell base stations and macro cell base stations based on LTE, in which small cells can improve spectrum efficiency, capacity, and coverage of traditional wireless networks. In practice, it would be difficult to pre-specify small
Fig. 6. Conceptual diagram of the Potsdam CRM with overlaid communication network.
5
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
commands using the same communication infrastructure could undermine the real-time requirement. Thus, we will leverage cell association in small cells and integrate multi-scale data aggregation to form multiple paths for data streams with varying QoS requirements. Indeed, in a CRM, packets destined to controllers will have the highest priority, while data collected from meters and loads allow aggregation and may also be treated as relatively delay-tolerant data streams. Adaptive algorithms are needed to aim at optimal aggregation of data at the ingress point. In turn, the adaptive algorithms will observe the self-organized small cell and the backhaul to compute the optimal path for each data stream. This can be done periodically at the aggregation intervals, as well as when the underlying CRM topology changes due to scheduled or unexpected events. Alternatively, computing the optimal path for data streams can be adapted through predictive analysis of the network status, which would periodically predict data rates, service coverage, and capacity (Letourneux et al., 2014).
•
•
5. An integrated reconfigurable control and self-organizing communication framework In order to enhance resilient and economic operation of CRMs, operation/control strategies and the overlaid communication infrastructure must be effectively coordinated in a flexible manner. Thus, we plan to investigate an integrated reconfigurable control and self-organizing communication framework for exploring their interactions, so that one can be adaptively adjusted in response to real-time changes of the other.
5.2. Reconfiguring self-organizing communication based on real-time CRM operation status The self-organizing communication network is used to monitor CRM operations and gather data at multiple time-scales, and to actuate controllers when needed. A common assumption is that the communication network is static and the central communication network manager has global information on the network topology and link qualities. However, such assumption may not be realistic in the CRM environment because of the complex heterogeneity of data sources and dynamic plug-and-play characters of DERs. Indeed, the communication network must be adaptively self-organized to synchronize with the topology and operation status of CRM. In order to design an effective way for adaptively reconfiguring the communication network according to the real-time topology and operation status of CRM, the following questions need be answered:
5.1. Reconfiguring centralized/distributed control based on real-time communication status Centralized and distributed operation/control strategies present different communication complexity and varying performances in terms of global optimality. Indeed, the extensive utilization of centralized control may deteriorate the communication network performance, make the time scales of control and communication unsynchronized, and in turn jeopardize the resilient and economic operation of CRMs. Thus, an adaptive scheme is needed to allow flexible switching between centralized and distributed strategies according to the real-time communication performance. Specifically, the centralized control strategy is used for pursuing global optimality to the maximum extend as long as the real-time communication performance is not compromised, and the distributed control strategy is adopted to critical applications for ensuring resiliency when the communication performance is degraded. Adaptive schemes that combine optimality of centralized control and resiliency of distributed control could provide robust operation and control against communication failures or degraded performance of communication links. Indeed, in order to design an effective way for adaptively reconfiguring the centralized and distributed control strategies according to real-time congestion and latency status of the communication network, the following questions need to be answered:
• How could the radio resource allocation be dynamically adjusted?
•
• How can the computation and communication complexities of dis-
•
capability for effectively supporting centralized computation and distributed control strategies. How can adaptive control strategies mitigate the impact of non-ideal communication links such as message loss and delay? Although a communication network design can incorporate mechanisms to achieve high reliability, one cannot assume that all the information will always be delivered without delays or losses all the time. Thus, the relationship between the convergence rate of the distributed algorithm, the restorability of network topologies, and the message loss rate needs to be quantified. It is worth noting that restorability of network topologies needs to be formulated as a function of redundancy in these networks (Gurel and Keskin, 2017). How could major disasters impact the communication network and result in partial disconnection of certain networks? To evaluate such situations and avoid their consequences, a survivable and selfhealing design for the critical loads and suppliers needs to be in place. In addition, CRM operation needs to be robust under performance degradation of the communication network due to failures.
tributed and centralized information and control be optimally leveraged? This would guarantee a robust performance in response to volatility and uncertainty in CRM and the communication network, as well as enhance resiliency and economic efficiency in both gridconnected and islanded modes. What are the optimal thresholds to adaptively switch between centralized and distributed controls? The thresholds could be quantified in terms of availability and reliability of the communication network under multi-failure conditions, including its topology and key parameters, which would reflect the communication
6
Although the cooperation among multiple partners in the CRM is enabled by the communication network, individual partners may present different levels of autonomy and authority, asymmetric data generation rates, and various objectives such as minimum loss, minimum latency, or minimum power consumption. Indeed, communication requirements including reliability, response time, latency, and sampling rates may be different, and a real-time dynamically configurable communication platform is needed to provide rapid flexibility. Furthermore, load patterns of the CRM, new equipment association, contingencies, weather conditions, and extreme events all necessitate the reconfiguration of the communication network. In this perspective, the communication platform needs to be self-organized for providing optimal reliability, maximum availability, and the shortest possible latency. What are the optimal frequency and aggressiveness levels of the data aggregation strategy? Data gathered from heterogeneous DERs, loads, and control equipment of multiple partners in the CRM need to be aggregated at several levels and prioritized based on the dynamic operation status of the CRM. This can be correlated with DERs’ outputs and load profiles. The adaptive data aggregation algorithm can operate in peak and off-peak regimes, as well as normal and extraordinary operating conditions. Observing the real-time CRM operation, data generated by equipment, and control commands released by operators, the adaptive data aggregation from multiple sensing points can be fine-tuned. Aggregation window size and variation of the data stream will play the key role in
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
•
•
communication infrastructure of CRMs. The article hopes to bring more attention from energy and utility industry, academia, and power grid agencies to investigate effective operational and control tools for the effective development and widespread deployment of CRMs to improve disaster response and recovery for our society.
determining optimal frequency and aggressiveness levels of the data aggregation strategy (Kolozali et al., 2016). It is worth mentioning that real-time distributed data aggregation does not only experience computational challenges but also increases communications overhead. As reported by a related work (Riker et al., 2016), adopting a multi-tier approach in data aggregation may improve computational and communication efficiency. That being said, data redundancy can be handled at one tier whereas the communications overhead can be handled by the second tier. How can low-complexity algorithms be designed to effectively perform the dynamic self-organization of a communications network in real time? The communications network reconfiguration needs to be done fast enough for meeting the stringent latency requirements; for instance, primary control cannot tolerate delays beyond milliseconds. Reconfiguration of small cell networks can dynamically release or acquire radio resources to or through macro cells in real time. In turn, self-organization and data aggregation addressed above need to be implemented with low-complexity algorithms. Indeed, as mentioned above, introducing an initial tier that would eliminate redundant data can help reduce excessive computational loads on the aggregation processes. How can small cell networks in the self-organization communications architecture be configured in a resilient manner? Data gathering and control signal delivery should not be affected in the presence of any failure. However, major disasters may cause communication outage in multiple cells. This calls for recovery using macro cells or adjacent cells which both require the fast handover of the devices as well as the context. To prevent abrupt outage of communications links for critical CRM components, resource allocation schemes can dynamically vacate mobile user traffic. This has to be done diligently as wireless networks and small cells are planned to be used for public safety in a close future. Therefore, how to allocate radio resources to critical parties under failures of single or multiple cells will need to be addressed. In addition, the small cell network needs to be self-healing and recover from failures with minimum manual configuration.
References Ahmed, F., Tirkkonen, O., 2016. Simulated annealing variants for self-organized resource allocation in small cell networks. Appl. Soft Comput. 38, 762–770. Alorainy, A., Hossain, M.J., Alouini, M.S., 2016. Multi-flow carrier aggregation in heterogeneous networks: cross-layer performance analysis. In: Washington, DC, USA. IEEE Globecom Workshops (GC Wkshps) 2016. pp. 1–7. Bennis, M., Perlaza, S.M., Blasco, P., Han, Z., Poor, H.V., 2013. Self-Organization in small cell networks: a reinforcement learning approach IEEE. Trans. Wireless Commun. 2 (July (7)), 3202–3212. Canadian Disaster Database, https://www.publicsafety.gc.ca/cnt/rsrcs/cndn-dsstr-dtbs/ index-en.aspx. Chang, B., Liou, S., 2017. Adaptive cooperative communication for maximizing reliability and reward in ultra-dense small cells LTE-A toward 5G cellular networking. Comput. Networks 115 (March), 16–28 Early Access: January 25. Design of a Resilient Underground Microgrid in Potsdam, NY. NYSERDA Project 41309. Clarkson University prime contractor. Er-Rahmadi, B., Bagaa, M., Ksentini, A., Meddour, D.E., 2016. Cost-efficient data aggregation schemes for small cell networks. In: 2016 International Wireless Communications and Mobile Computing Conference (IWCMC). Paphos. pp. 77–83. Gurel, G., Keskin, S., 2015. Emergency recovery for small cell management system. 23rd Signal Processing and Communications Applications Conference. Kolozali, S., Puschmann, D., Bermudez-Edo, M., Barnaghi, P., 2016. On the effect of adaptive and nonadaptive analysis of time-series sensory data. IEEE IoT J. 3 (December (6)), 1084–1098. Letourneux, F., Guivarch, S., Lostanlen, Y., 2014. Impact of modeling tools on outdoor small-cell deployment cost in a realistic urban scenario. In: Vancouver, BC. 2014 IEEE 80th Vehicular Technology Conference (VTC2014-Fall) 2. pp. 1–5. Li, J., Liu, Y., Wu, L., 2016. Optimal operation for community based multi-party microgrid in grid-connected and islanded modes. IEEE Trans. Smart Grid. http://dx.doi. org/10.1109/TSG.2016.2564645. Liu, Y., Li, J., Wu, L., 2017. Distribution system restructuring: distribution LMP via unbalanced ACOPF. IEEE Trans. Smart Grid. http://dx.doi.org/10.1109/TSG.2016. 2647692. Madueno, G., Nielsen, J., Kim, D., Pratas, N., Stefanovic, C., Popovski, P., 2016. Assessment of LTE wireless access for monitoring of energy distribution in the smart grid. IEEE J. Sel. Areas Commun. 34 (3), 675–688. Morozs Grace, N.D., Clarke, T., 2013. Case-based reinforcement learning for cognitive spectrum assignment in cellular networks with dynamic topologies. In: 2013 Military Communications and Information Systems Conference. St.-Malo. pp. 1–6. NOAA, 2013. Hurricane/post-tropical Cyclone Sandy, October 22–29, 2012. National Oceanic and Atmospheric Administration. June 2013. [Online]. Available: http:// www.nws.noaa.gov/os/assessments/pdfs/Sandy13.pdf. NOAA, 2014. Billion-dollar Weather/climate Disasters. National Oceanic and Atmospheric Administration. January 2014. [Online]. Available: http://www.ncdc. noaa.gov/billions/events. Governor Cuomo announces NY Prize resiliency competition to launch this Fall, Press Release, New York State, Aug. 26, 2014. Peng, M., Liang, D., Wei, Y., Li, J., Chen, H., 2013. Self-configuration and self-optimization in LTE-advanced heterogeneous networks. IEEE Commun. Mag. 51 (May (5)), 36–45. REV, 2015. New York State Department of Public Service–Reforming the Energy Vision. http://www3.dps.ny.gov/W/PSCWeb.nsf/All/ 26BE8A93967E604785257CC40066B91A?OpenDocument. Rahmati, A., Shah-Mansouri, V., Safari, M., 2017. Price-based resource allocation for selfbackhauled small cell networks. Comput. Commun. 97 (January), 72–80. Riker, A., Cerqueira, E., Curado, M., Monteiro, E., 2016. A two-tier adaptive data aggregation approach for M2 M group-Communication. IEEE Sens. J. 16 (3), 823–835 Feb. 1. Ross, T., Lott, N., 2016. A climatology of recent extreme weather and climatic events, National Climatic Data Center. Techn. Rep. 2000–2002. Samarakoon, S., Bennis, M., Saad, W., Debbah, M., Latva-aho, M., 2016. Ultra dense small cell networks: turning density into energy efficiency. IEEE J. Sel. Areas Commun. 34 (5), 1267–1280. Xu, J., Tang, L., Chen, Q., Yi, L., 2014. Study on based reinforcement Q-learning for mobile load balancing techniques in LTE-A HetNets. IEEE 17th International Conference on Computational Science and Engineering, Che. Zhang, G., Zhang, G., Gao, Y., Lu, J., 2011. Competitive strategic bidding optimization in electricity markets using bilevel programming and swarm technique. IEEE Trans. Ind. Electron. 58 (June (6)), 2138–2146. Zhang, H., Jiang, C., Hu, R., Qian, Y., 2016. Self-organization in disaster-resilient heterogeneous small cell networks. IEEE, Network(March/April).
6. Final words The successful development and widespread deployment of effective CRM technology have the potential to provide significant improvements for disaster response and recovery for our society. They will also have a profound impact on communities during normal operating conditions, including additional revenues for DER owners derived from operational knowledge and efficiencies. However, in order to enable the full functionality of CRMs for providing seamlessly resilient electric service to the recovery effort following catastrophes while ensuring electrical, economic, and societal goals, the communications network is of critical value in CRM operation and control. Specifically, CRM operators must have access to real-time CRM operation information gleaned from high-fidelity sensors so as to guarantee satisfactory realtime performance on enhancing the resilient and economic operation under both normal situations and during major disruption events. However, the communications network could be inadequate to the challenges of moving large amounts of data in real time to the places they are needed, and in turn control and communication are not synchronized. To address such critical challenges, this article discussed unique operational characteristics of CRMs across multiple timescales in different operation modes, and presented coordinated control and communication solutions for the resilient and economic operation of CRMs. A practical CRM project that is underway for serving Potsdam, N.Y., was also presented. The article further outlined several important issues related to the design of an integrated reconfigurable control and selforganizing communications framework, in order to effectively and adaptively coordinate operation/control strategies and the overlaid
Lei Wu is an Associate Professor of Electrical and Computer Engineering at Clarkson University, Potsdam, New York, with research interests that include power systems operation and planning, energy economics, and community resilience microgrid. He received a Ph.D. in Electrical Engineering from Illinois Institute of Technology (IIT),
7
The Electricity Journal xxx (xxxx) xxx–xxx
L. Wu et al.
IEEE’s Technical Committee on Green Communications and Computing and the research group leader for IEEE Smart Grid and Big Data Standardization. Her main research interests are wireless networks, the smart grid, cyber-physical systems, electric vehicles and wireless sensor networks.
Chicago, in 2008. From 2008 to 2010, he was a Senior Research Associate with the Robert W. Galvin Center for Electricity Innovation at IIT. He also worked as summer Visiting Faculty at the New York Independent System Operator (NYISO) in 2012. Jie Li is an Assistant Professor in the Electrical and Computer Engineering Department at Clarkson University in Potsdam, New York. She received a Ph.D. in Electrical Engineering from the Illinois Institute of Technology (IIT), Chicago, in 2012. From 2006 to 2008, she was a Research Engineer with IBM China Research Lab. From 2012 to 2013, she was a Power System Application Engineer with GE Energy Consulting. Her research interests include power systems restructuring and bidding strategy.
Burak Kantarci is an Assistant Professor at the School of Electrical Engineering and Computer Science at the University of Ottawa. Prior to joining the University of Ottawa, he was an Assistant Professor at Clarkson University, Potsdam, New York. He received his Ph.D. in Computer Engineering from Istanbul Technical University in 2009. He is the founding director of the Next Generation Communications and Computing Networks (NEXTCON) research group. He is currently an editor for IEEE Transactions on Green Communications and Networking, editor for IEEE Communications Surveys and Tutorials, and Secretary of IEEE’s Communication Systems Integration and Modelling Technical Committee. His main research interests are mobile computing and communications, the internet of things, wireless sensor networks, digital health, and cloud computing.
Melike Erol-Kantarci is an Assistant Professor at the School of Electrical Engineering and Computer Science at the University of Ottawa. Prior to joining the University of Ottawa, she was an Assistant Professor at Clarkson University, Potsdam, New York. She received a Ph.D. in Computer Engineering from Istanbul Technical University in 2009. She is currently the vice-chair of the Green Smart Grid Communications special interest group of
8