- Email: [email protected]
Design and control of energy systems in Denmark – Challenges and opportunities
Author’s Accepted Manuscript Design and Control of Energy Systems in Denmark - Challenges and Opportunities Tommy Moelbak
www.elsevier.com/locate/ejcon
PII: DOI: Reference:
S0947-3580(16)30010-3 http://dx.doi.org/10.1016/j.ejcon.2016.04.005 EJCON166
To appear in: European Journal of Control Received date: 1 February 2016 Accepted date: 20 April 2016 Cite this article as: Tommy Moelbak, Design and Control of Energy Systems in Denmark - Challenges and Opportunities, European Journal of Control, http://dx.doi.org/10.1016/j.ejcon.2016.04.005 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Design and Control of Energy Systems in Denmark - Challenges and Opportunities Tommy Moelbak
Abstract— The Danish energy system has been a front-runner within integration of renewables for the last decades. Historically this has led to needs and new developments within design and control of production and supply systems, on individual plant level as well as on portfolio level. These developments include applications within adaptive control, multivariable control, portfolio balancing control and load planning systems. In future the needs for more detailed utilization of plant and system knowledge will grow further due to increased market and technology complexity. Improved utilization of hybrid knowledge, i.e. modelling on different timescales and modelling with diverse fidelity, contains large potential improvements in the design phase as well as in the operational phase. These challenges should be taken up in research and eventually be applied commercially.
Keywords Energy systems, control, design, performance, flexibility I. INTRODUCTION Denmark has been and still is a front-runner in the development of energy systems based on renewable resources like wind, solar, biomass, waste, etc. This has led to design and operational challenges due to increased complexity in terms of efficiency and stability. These challenges have also led to the need for development of new methods within integrated design, optimization and control. This paper resumes the past achievements and identifies the future needs within these fields, in which control and system R&D has unique challenges and opportunities. The challenges within development of the Danish energy system and the corresponding control challenges are shown on a time line in figure 1. In system terms, development comes from a very controllable centralized situation with primary focus on production efficiency to a predicted future based on a very distributed uncontrollable system focusing on multi-product (power, heat, biofuels, etc.) flexibility. This system development has catalysed the needs within optimization and control, which initially focused on control of individual components (pumps, superheaters, turbines, etc.) and on a future which demands integrated optimization across complex plants. In the 90s Danish energy production and supply were based on centralized fossil CHP (Combined Heat and Power) production. Most production capacity was large controllable units, and operational optimization was focusing on efficiency, availability and lifetime consumption. In terms of on-line optimization and control applications, R&D was focused on components or sub-processes within plants, most often SISO systems. An example is given in [1], which describes different approaches to superheater control, including fuzzy based Control and IMC (Internal Model Control) of a superheater system.
Tommy Moelbak is with Added Values, Lysholt Allé 10, 7100 Vejle, Denmark. Phone: +45 24479544. E-mail: [email protected].
Figure 1 Historical headlines in Danish energy supply and control
In the late 90s and beginning of the 0s, wind power production became a significant part of the power supply – see figure 2. In the same period decentralized production increased. All together this led to reduced controllability of the power system – wind power disturbances increased and control capacity decreased. Optimization of load following capability of controllable capacity was the answer to these challenges. An example is given in [2], in which multivariable LQG control was used to optimize the load speed capability of boilers. These multivariable principles are still used, usually during concept design, and then implementation in C&I systems is approximated using SISO functionalities. During the 0s the demand for product flexibility increased further. Wind power production increased and controllable share of power decreased, see figure 2. Solutions were found in further exploitation of existing hardware, e.g. low load operation, district heat exchange on storage tanks and condensate modulation on individual plants, and balancing across several plants. Initiatives were still based on existing hardware and solutions were primary found in new control and optimization methods. [3] includes results on how new master control methods could increase power load changes by utilizing the individual plant in the above mentioned ways. The need for optimized load planning, power and district heat production increased in order to optimize power market business while still ensuring district delivery as a public obligation – [3] also includes examples on some of the first developments of such load scheduling tools, which are now widely used on plants.
Figure 2 Development in Danish power production
Figure 3 A future energy system in Denmark [8]
During the 0s, portfolio owners also exploited the opportunity to optimize total power balancing by using different types of dynamic handles across plants. This implies tight coordination of different dynamic features: different time response, different persistence and different costs. Model based control is needed for this purpose and [4] describes how MPC can be used to control the power balance of a diverse production portfolio with the objective of optimizing economic benefits on balancing markets. In the mid 0s and beginning of 10s smart grid initiatives became very intensive as a solution for increasing the overall ability on power balancing. Especially power consumption was seen as a cheap solution for balancing the power grid. In [5] and [6] different solutions for mobilizing consumers were explored, including a predictive and an agile strategy. Research
within the smart grid area has been very intensive, and still the real application has been quite limited in terms of power balancing volume and primarily it has been applied for short term services. Since then the smart grid terminology has been included as part for the more general work within smart energy – perspectives on this has been discussed in [7], in which it is argued that the smart energy approach is much more valuable. These challenges have put plant-wide control in focus in order to handle complex operation of several energy producing and consuming plants in an optimal manner. These challenges are still not fully solved and challenges will further increase in future. This paper will take up these challenges and identify needs for development of new methods.
II. CHALLENGES IN FUTURE ENERGY SYSTEM Danish energy policies and strategies aim for a fossil free energy system based on a number of different green technologies and based on synergies across different energy forms. Figure 3 illustrates one example of a future energy system in Denmark. Figure 3 is discussed in [8], and this example includes a total energy system which is divided into three layers:
A resource layer which consists of the production plants. Basic resources include fluctuating sources like wind and solar, controllable biomass sources and controllable fossil sources. Deliveries from this layer are power, heat and gas.
A system layer which consists of the transportation infrastructure, pipes and storage for the district heat and gas system and grid for the power system. This layer is also connected to the international power and gas grid.
A service layer which consists of the consumers. Consumers need services, but can also in this example deliver supportive services to the system operation, e.g. electrical vehicles can provide ancillary services to the power grid.
Usually the resource layer plants and the system layer plants are owned by local asset owners, who will pursue the total business efficiency. In future, the service layer may even be included in these asset owners’ business model, e.g. investing and operating household heat pumps. This type of energy system includes an increased design and operational complexity in several directions:
It is complex in design. Optimization criteria across a diverse portfolio are also diverse: resource efficiency, product flexibility, delivery redundancy, etc.
It is complex in operation. Optimized utilization of the hardware requires intelligence in on-line systems in order to support the operators efficiently.
It is complex in development. Development from the asset owner’s point of view will be incremental, i.e. existing plants, decommissioning and planning of new plants should be taken into consideration.
It is complex in geography. Asset owners will face a local balancing problem on district heat, a regional balancing problem on gas and a national/ international balancing problem.
Altogether, the energy system is in transition, a shift of paradigm, which means that the complexity in optimization – process design, control design and operational efficiency - is certainly very challenging.
III. THE OPTIMIZATION CHALLENGES The general optimization process is shown in figure 4. This is often regarded as a pure sequence, but as complexity grows it will take an increasingly iterative nature, i.e. feedback information backwards in the chain will grow.
Figure 4 General optimization process Taking the chain into the shift of paradigm in the energy system, the feedback iterations will grow enormously, and the process will most probably not converge to anything near a global optimum. Illustrated by a still quite simple example taking the asset owner’s point of view: An asset owner operates a number of different plants: waste-to-energy CHP, biomass CHP, solar power, gas-fired peakload boilers, a district heat storage tank and a district heat distribution grid. The owner now considers investing in an electrical heat pump or in technology enabling by-pass operation on the biomass CHP plant. The business objective is to increase the product flexibility: Take out power production in low-price hours and increase balancing capabilities on shortterm power markets. This type of business case will traditionally be evaluated through focus on the present situation, which means the expected gross revenue and investment costs for each of the two cases. Detailed considerations on cross synergies with other plants, delivery of balancing and peak-load services to power markets and analysis of risks in operation and markets are typically done in late phases of the optimization process. All together this will result in many iterations, e.g. optimized dynamic interaction between plants will be considered during commissioning and reveal a number of redesign demands which cannot be implemented at this stage. Further, in the operation phase the plants will interact optimal due to reduced design analysis and due to lack of tools for supporting operation on portfolio level. Generalizing on this, still quite simple, example, two types of improvements in the optimization process are needed:
Optimized integrated design. There is a need for taking design and operation details into consideration at an early stage and combining these with portfolio characteristics in a systematic manner.
Optimized integrated operation. There is a need for new on-line tools which can support optimized and robust operation of several plants of different types.
In control language, the objective is to reduce the need for feedback in the optimization process in figure 4 by increasing the efficiency in feedforward. This will require totally new methods in the design phases and in the operational phase.
IV. OPTIMIZED INTEGRATED DESIGN In the design phase, it is necessary to make detailed considerations on a number of different time-scales in order to obtain final optimized operation and robustness:
Steady-state characteristics: Include issues such as thermodynamic efficiency, optimized utilization of fuels, operational capacity and maintenance consequences.
Dynamic characteristics: Include issues such as product balancing, failure robustness, start/stop characteristics, control concepts and load-depending dynamics.
Portfolio characteristics: Include issues such as operational envelopes, security of supply, market robustness and road map robustness.
These different optimization criteria should be brought together in an overall optimization as illustrated in figure 5.
Figure 5 Iterative design optimization
An alternative to an iterative approach based on simulations on several time-scales will be one global model based optimization, but this is considered to be unrealistic. The different types of model based simulations will be elaborated below.
A. Steady-state characteristics In general, steady-state characteristics of individual plants have traditionally drawn a lot of attention in the design phases of energy production and supply plants. An example of focus areas is shown in figure 6, a combined heat and power plant (CHP). This type of production plant has in recent years experienced an increased demand for product flexibility, including stretching the operational envelope as indicated in figure 6. Taking this example into general considerations in the design phases, it is traditionally done important to optimize
P [MW]
Q [MJ/s]
Figure 6 Stretching the operational envelope of a CHP plant
steady-state characteristics of the new plant to be build. In addition, the increasing complexity dimensions described in section II will demand focus on total portfolio characteristics, e.g. considerations on other existing plant steady state characteristics. Increased attention on portfolio characteristics will potentially induce improved portfolio characteristics followed by reduced investment costs. Existing simulation and design tools within this area, commercially available and in-house developed, have limitations with respect to these needs:
Model fidelity is typically high, which enhances ability to optimized design of the individual plant, but it is not supportive on bringing the new plant into portfolio context. Model scope is limited to individual plants or sub processes of these, which implies that there are no well-defined methods for modelling steady-state characteristics across several plants.
B. Dynamic characteristics Like steady state characteristics, dynamic characteristics are traditionally optimized on individual plants as well. In addition, the optimization process is most often carried out as a sequence, i.e. firstly the process is designed, secondly the C&I is conceptually designed, thirdly the controls are tuned and adjusted during commissioning. An example of dynamic issues is shown in figure 7. Boiler dynamics is often the bottleneck for improving the overall plant stability during volatile operation. This is typically one of many issues when operational flexibility of CHPs is optimized, e.g. as done using multivariable control in [2]. As an isolated case this will very often imply a good business case due to improved load change capability, but it will also often result in increased costs on O&M. A more integrated approach also considering other possibilities at the same plant or at other plants will most likely relax the demands on the boiler and still be able to perform well in portfolio terms.
Firing rate Brændsel
FeedFødevand water flow
Tryk Pressure
Enthalpy Entalpi
Figure 7 Optimizing dynamic response of a boiler
Tools for dynamic simulations and analysis have the following features:
Model fidelity is chosen as appropriate for the specific case and problem. For plant design it will typically include physical properties directly or indirectly. In portfolio respect plant models are often aggregated transfer function models which include process, instrumentation and controls. Model scope is typically not necessarily limited, but often aggregated non-physical models are used for large systems. Many system models suffer from stiffness and often compromises are taken.
C. Portfolio characteristics On portfolio level of energy production and supply, the focus is often on simulating operation on longer time-horizons, e.g. a one-year-period with 1 hour resolution. In some cases design is based on even rougher models, e.g. monthly average or yearly duration curves. An example is shown in figure 8 with a schematic model of a system including district heat piping, heat producing units, storage tanks and CHP units. For such examples analysis often emphasizes on identifying potential benefits in new plants, e.g. a storage tank, or in new operational strategies, e.g. by-pass operation in CHPs. In the design phase focus is purely on very superior design issues like sizing and placement of a storage tank Tools for simulation and analysis of energy supply systems are characterized by the following:
Model fidelity is always quite low because the purpose is on overview level and because very high fidelity will result in too high complexity in such a model.
HILLERØD
OMR6 etape A
VAK H
OMR4 NOVO SLK 2*7,5 MW
VAK N
SLK ny
Egedal
VAK W
OMR1 Max 60 MW
Consumers Upstream bottleneck
Vest
CTR/GLC
Nord
Max 110 MW
Hedegården SLK 2*18 MW VAK S
Vestforbrænding SLK 3*20 MW VF5+VF6
Power
Waste Summercoolers 30 MW
Min 10 MW Max 75 MW
VEKS
Figure 8 Model of energy production and supply system
In principle, model scope can vary from one site to a larger local area, to a region and even to a country. Scope will depend on the purpose and relevance from asset owner’s point of view.
V. OPTIMIZED INTEGRATED OPERATION In the operation phase it is in principle the same knowledge base as in the design phase. It is “just” an on-line real-time situation which means that all functionalities should be automatically executable and able to take in and deliver information automatically while still being robust on different types of errors and abnormalities. Taking the future needs and complexities described in section II in account this will increase the need for new functionalities significantly. Figure 9 illustrates an example of how many plants are already operated. In this case only very superior functionalities are shown and they are split between a portfolio level and an individual level, in which the shown functionalities are of course repeated on every plant. The functionalities shown in the example can briefly be described as:
Load planning should optimize the utilization of plants with respect to the markets, presently power and district heat markets, in future with increased complexity and also more market places will be relevant.
Balancing controller should make short time corrections in case of internal imbalances or in case of external requests.
Port folio level
Load planning (24 h)
Powerprices, PBA, contracts
Production plan
Correction
Balancing of production
Marketorders
Setpoint schedule for individual plants
Power production
Individual plants Performance indicators for operator
Marginal costs
Productionplan
Performance monitoring Correction of setpoints
Measure ments
Master Control Measure ments
Setpoints
C&I
Power plants
Figure 9 Structure for on-line optimized operation of plants
Master Control functionality is crucial to convert delivery demands into coordinated control and stability of the individual plant.
Performance monitoring is crucial on each plant to ensure persistent efficiency, quick handling of failures and updated calculation of plant economics.
All of the above described functionalities are often called “the economical control room”. There is a need for new developments on portfolio level as well as on individual plant level. A. Portfolio control Historically advanced portfolio control schemes have mostly been applied by large asset owners with the challenge of planning and balancing a combination of controllable plants and non-controllable plants. A specific example of this is shown in figure 10. In this case an in-house developed load planning application and MPC based balancing controller were used [9]. In general this type of control scheme will have to be developed further in future for a number of reasons:
Plant portfolios will be even more complex in composition. Plants will individually be smaller in capacity which increases the number of plants to handle but also probably reduces the total capacity. Markets will increase in complexity, e.g. power and district heat markets, and more market places will occur, e.g. on fuels and residues.
This calls for, at least, extrapolation of the present methods if up-scaling is possible, otherwise new concepts and methods are needed.
Figure 10 Portfolio control scheme [9]
B. Plant control Control in this context should be interpreted in its widest definition, i.e. as classical control, but also as tracking and supervising the state of the plant in general. Classical control, feedback loops and feedforward control, have drawn a lot of attention in research and also in practical utilization. Pragmatic utilization of advanced knowledge has evolved in industry rather than utilization of advanced control methods. This has been driven by the demand for simplicity and transparency, and will certainly be a key driver in the future as well. Supportive performance functions in different versions have also been applied widely in large plants and will probably propagate into smaller plants in new versions as the service demand from small plants increases. These functions are typically not critical on a short-term time-scale, because they focus on long-term issues or on issues that will not influence short-term plant availability. Until now these functions have primarily been used for supportive functions, i.e. always with a plant operator as receiver. In future there is a need for increasing intelligence in control while still keeping robustness and simplicity in real-time automation on a high level. Figure 11 illustrates a step in this direction. The idea is to have a bank of intelligence, different types of plant models combined with intelligence that can transform basic physical knowledge into relevant operational information. The intelligence bank is separated from real-time automation which can always operate independently of the bank. This type of architecture which separates complex knowledge and real-time critical functionality could be a possible solution to future demands.
FlexIQ Subscribers
FlexIQ Flex capacity Flex costs Fuel efficiency Maintenance costs Failure state …
Corrective actions Assets Figure 11 Intelligence bank for advanced control
VI. CONCLUSIONS The Danish energy system has been a front-runner with respect to integration of renewables, and historically the development has challenged the design and control of production plants and supply systems. This trend will continue in the coming decades and result in new challenges and opportunities within design and control. New challenges will primarily be driven by demand for optimized design and operation of distributed hybrid systems:
Demand for portfolio flexibility on energy resources and on products will increase due to the political ambitions on renewables. Demand for product diversity and plant integration will increase due to the need for improved stakeholder value in utilities. Demand for integrated utilization of new small-scale distributed technologies due to the need for reduced investment risks and need for production flexibility.
These challenges will change the research and application agenda in direction of developments crossing traditional research areas. Basically the fundamental knowledge is present, but initiatives for developments combining the basic knowledge are needed:
In the design phase there is a need for combining modelling on different time-scales and with different fidelity into overall optimized design.
In the operation phase there is a need for taking more detailed and hybrid intelligence into on-line robust utilization.
There are no ready answers to these needs, but the challenges are certainly becoming clearer, and research within systems, modelling and control are capable of taking up the challenges.
REFERENCES [1] Moelbak, T., Advanced control of superheater steam temperatures - An evaluation based on practical applications, IFAC Conference on Control of Power Systems and Power Plants 1997 (cpspp'97), pp. 161-166
[2] J.H. Mortensen, T. Moelbak, P. Andersen and T.S. Pedersen, Optimization of boiler control to improve the load-following capability of power-plant units, Control Engineering Practice 6/12, p. 1531-1539, 1998
[3] Joergensen, C., Mortensen, J.H., Moelbak, T. and Nielsen, E.O.: “Model Based Fleet Optimisation and Master Control of a Power Production System”, Proceedings of IFAC Symposium on Power Plants and Power Systems Control, Canada, June, 2006
[4] K. Edlund, J. Bendtsen, T. Moelbak, Simple Models for Model-based Portfolio Load Balancing Controller Synthesis, Proc. of the 2008 IFAC Symposium on Power Plants and Power Systems Control, 2009
[5] M. K. Petersen, L. H. Hansen and T. Mølbak, Exploring the Value of Flexibility: A Smart Grid Discussion, Proceedings of IFAC Symposium on Power Plants and Power System Control, vol. 8, no. 1, pp. 43-48, 2012
[6] Juelsgaard, M., Totu, L. C., Shafiei, S. E., Wisniewski, R., Stoustrup, J., Control Structures for Smart Grid Balancing, Proceedings of the 2013 4th IEEE PES Innovative Smart Grid Technologies Europe (ISGT Europe). IEEE Press, 2013
[7] Henrik Lund, Anders N. Andersen, Poul Alberg,Østergaard, Brian Vad Mathiesen, David Connolly, From electricity smart grids to smart energy systems – A market operation based approach and understanding, Energy,Volume 42, Issue 1, June 2012, Pages 96–102
[8] P. Meibom, K. B.Hilger, H. Madsen, D. Vinther, Energy Comes together in Denmark, ieee power & energy magazine, September/October 2013, p. 4655
[9] Edlund, K. S., Dynamic Load Balancing of a Power System Portfolio. Ph.D thesis, Aalborg University, 2010.
www.elsevier.com/locate/ejcon
PII: DOI: Reference:
S0947-3580(16)30010-3 http://dx.doi.org/10.1016/j.ejcon.2016.04.005 EJCON166
To appear in: European Journal of Control Received date: 1 February 2016 Accepted date: 20 April 2016 Cite this article as: Tommy Moelbak, Design and Control of Energy Systems in Denmark - Challenges and Opportunities, European Journal of Control, http://dx.doi.org/10.1016/j.ejcon.2016.04.005 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Design and Control of Energy Systems in Denmark - Challenges and Opportunities Tommy Moelbak
Abstract— The Danish energy system has been a front-runner within integration of renewables for the last decades. Historically this has led to needs and new developments within design and control of production and supply systems, on individual plant level as well as on portfolio level. These developments include applications within adaptive control, multivariable control, portfolio balancing control and load planning systems. In future the needs for more detailed utilization of plant and system knowledge will grow further due to increased market and technology complexity. Improved utilization of hybrid knowledge, i.e. modelling on different timescales and modelling with diverse fidelity, contains large potential improvements in the design phase as well as in the operational phase. These challenges should be taken up in research and eventually be applied commercially.
Keywords Energy systems, control, design, performance, flexibility I. INTRODUCTION Denmark has been and still is a front-runner in the development of energy systems based on renewable resources like wind, solar, biomass, waste, etc. This has led to design and operational challenges due to increased complexity in terms of efficiency and stability. These challenges have also led to the need for development of new methods within integrated design, optimization and control. This paper resumes the past achievements and identifies the future needs within these fields, in which control and system R&D has unique challenges and opportunities. The challenges within development of the Danish energy system and the corresponding control challenges are shown on a time line in figure 1. In system terms, development comes from a very controllable centralized situation with primary focus on production efficiency to a predicted future based on a very distributed uncontrollable system focusing on multi-product (power, heat, biofuels, etc.) flexibility. This system development has catalysed the needs within optimization and control, which initially focused on control of individual components (pumps, superheaters, turbines, etc.) and on a future which demands integrated optimization across complex plants. In the 90s Danish energy production and supply were based on centralized fossil CHP (Combined Heat and Power) production. Most production capacity was large controllable units, and operational optimization was focusing on efficiency, availability and lifetime consumption. In terms of on-line optimization and control applications, R&D was focused on components or sub-processes within plants, most often SISO systems. An example is given in [1], which describes different approaches to superheater control, including fuzzy based Control and IMC (Internal Model Control) of a superheater system.
Tommy Moelbak is with Added Values, Lysholt Allé 10, 7100 Vejle, Denmark. Phone: +45 24479544. E-mail: [email protected].
Figure 1 Historical headlines in Danish energy supply and control
In the late 90s and beginning of the 0s, wind power production became a significant part of the power supply – see figure 2. In the same period decentralized production increased. All together this led to reduced controllability of the power system – wind power disturbances increased and control capacity decreased. Optimization of load following capability of controllable capacity was the answer to these challenges. An example is given in [2], in which multivariable LQG control was used to optimize the load speed capability of boilers. These multivariable principles are still used, usually during concept design, and then implementation in C&I systems is approximated using SISO functionalities. During the 0s the demand for product flexibility increased further. Wind power production increased and controllable share of power decreased, see figure 2. Solutions were found in further exploitation of existing hardware, e.g. low load operation, district heat exchange on storage tanks and condensate modulation on individual plants, and balancing across several plants. Initiatives were still based on existing hardware and solutions were primary found in new control and optimization methods. [3] includes results on how new master control methods could increase power load changes by utilizing the individual plant in the above mentioned ways. The need for optimized load planning, power and district heat production increased in order to optimize power market business while still ensuring district delivery as a public obligation – [3] also includes examples on some of the first developments of such load scheduling tools, which are now widely used on plants.
Figure 2 Development in Danish power production
Figure 3 A future energy system in Denmark [8]
During the 0s, portfolio owners also exploited the opportunity to optimize total power balancing by using different types of dynamic handles across plants. This implies tight coordination of different dynamic features: different time response, different persistence and different costs. Model based control is needed for this purpose and [4] describes how MPC can be used to control the power balance of a diverse production portfolio with the objective of optimizing economic benefits on balancing markets. In the mid 0s and beginning of 10s smart grid initiatives became very intensive as a solution for increasing the overall ability on power balancing. Especially power consumption was seen as a cheap solution for balancing the power grid. In [5] and [6] different solutions for mobilizing consumers were explored, including a predictive and an agile strategy. Research
within the smart grid area has been very intensive, and still the real application has been quite limited in terms of power balancing volume and primarily it has been applied for short term services. Since then the smart grid terminology has been included as part for the more general work within smart energy – perspectives on this has been discussed in [7], in which it is argued that the smart energy approach is much more valuable. These challenges have put plant-wide control in focus in order to handle complex operation of several energy producing and consuming plants in an optimal manner. These challenges are still not fully solved and challenges will further increase in future. This paper will take up these challenges and identify needs for development of new methods.
II. CHALLENGES IN FUTURE ENERGY SYSTEM Danish energy policies and strategies aim for a fossil free energy system based on a number of different green technologies and based on synergies across different energy forms. Figure 3 illustrates one example of a future energy system in Denmark. Figure 3 is discussed in [8], and this example includes a total energy system which is divided into three layers:
A resource layer which consists of the production plants. Basic resources include fluctuating sources like wind and solar, controllable biomass sources and controllable fossil sources. Deliveries from this layer are power, heat and gas.
A system layer which consists of the transportation infrastructure, pipes and storage for the district heat and gas system and grid for the power system. This layer is also connected to the international power and gas grid.
A service layer which consists of the consumers. Consumers need services, but can also in this example deliver supportive services to the system operation, e.g. electrical vehicles can provide ancillary services to the power grid.
Usually the resource layer plants and the system layer plants are owned by local asset owners, who will pursue the total business efficiency. In future, the service layer may even be included in these asset owners’ business model, e.g. investing and operating household heat pumps. This type of energy system includes an increased design and operational complexity in several directions:
It is complex in design. Optimization criteria across a diverse portfolio are also diverse: resource efficiency, product flexibility, delivery redundancy, etc.
It is complex in operation. Optimized utilization of the hardware requires intelligence in on-line systems in order to support the operators efficiently.
It is complex in development. Development from the asset owner’s point of view will be incremental, i.e. existing plants, decommissioning and planning of new plants should be taken into consideration.
It is complex in geography. Asset owners will face a local balancing problem on district heat, a regional balancing problem on gas and a national/ international balancing problem.
Altogether, the energy system is in transition, a shift of paradigm, which means that the complexity in optimization – process design, control design and operational efficiency - is certainly very challenging.
III. THE OPTIMIZATION CHALLENGES The general optimization process is shown in figure 4. This is often regarded as a pure sequence, but as complexity grows it will take an increasingly iterative nature, i.e. feedback information backwards in the chain will grow.
Figure 4 General optimization process Taking the chain into the shift of paradigm in the energy system, the feedback iterations will grow enormously, and the process will most probably not converge to anything near a global optimum. Illustrated by a still quite simple example taking the asset owner’s point of view: An asset owner operates a number of different plants: waste-to-energy CHP, biomass CHP, solar power, gas-fired peakload boilers, a district heat storage tank and a district heat distribution grid. The owner now considers investing in an electrical heat pump or in technology enabling by-pass operation on the biomass CHP plant. The business objective is to increase the product flexibility: Take out power production in low-price hours and increase balancing capabilities on shortterm power markets. This type of business case will traditionally be evaluated through focus on the present situation, which means the expected gross revenue and investment costs for each of the two cases. Detailed considerations on cross synergies with other plants, delivery of balancing and peak-load services to power markets and analysis of risks in operation and markets are typically done in late phases of the optimization process. All together this will result in many iterations, e.g. optimized dynamic interaction between plants will be considered during commissioning and reveal a number of redesign demands which cannot be implemented at this stage. Further, in the operation phase the plants will interact optimal due to reduced design analysis and due to lack of tools for supporting operation on portfolio level. Generalizing on this, still quite simple, example, two types of improvements in the optimization process are needed:
Optimized integrated design. There is a need for taking design and operation details into consideration at an early stage and combining these with portfolio characteristics in a systematic manner.
Optimized integrated operation. There is a need for new on-line tools which can support optimized and robust operation of several plants of different types.
In control language, the objective is to reduce the need for feedback in the optimization process in figure 4 by increasing the efficiency in feedforward. This will require totally new methods in the design phases and in the operational phase.
IV. OPTIMIZED INTEGRATED DESIGN In the design phase, it is necessary to make detailed considerations on a number of different time-scales in order to obtain final optimized operation and robustness:
Steady-state characteristics: Include issues such as thermodynamic efficiency, optimized utilization of fuels, operational capacity and maintenance consequences.
Dynamic characteristics: Include issues such as product balancing, failure robustness, start/stop characteristics, control concepts and load-depending dynamics.
Portfolio characteristics: Include issues such as operational envelopes, security of supply, market robustness and road map robustness.
These different optimization criteria should be brought together in an overall optimization as illustrated in figure 5.
Figure 5 Iterative design optimization
An alternative to an iterative approach based on simulations on several time-scales will be one global model based optimization, but this is considered to be unrealistic. The different types of model based simulations will be elaborated below.
A. Steady-state characteristics In general, steady-state characteristics of individual plants have traditionally drawn a lot of attention in the design phases of energy production and supply plants. An example of focus areas is shown in figure 6, a combined heat and power plant (CHP). This type of production plant has in recent years experienced an increased demand for product flexibility, including stretching the operational envelope as indicated in figure 6. Taking this example into general considerations in the design phases, it is traditionally done important to optimize
P [MW]
Q [MJ/s]
Figure 6 Stretching the operational envelope of a CHP plant
steady-state characteristics of the new plant to be build. In addition, the increasing complexity dimensions described in section II will demand focus on total portfolio characteristics, e.g. considerations on other existing plant steady state characteristics. Increased attention on portfolio characteristics will potentially induce improved portfolio characteristics followed by reduced investment costs. Existing simulation and design tools within this area, commercially available and in-house developed, have limitations with respect to these needs:
Model fidelity is typically high, which enhances ability to optimized design of the individual plant, but it is not supportive on bringing the new plant into portfolio context. Model scope is limited to individual plants or sub processes of these, which implies that there are no well-defined methods for modelling steady-state characteristics across several plants.
B. Dynamic characteristics Like steady state characteristics, dynamic characteristics are traditionally optimized on individual plants as well. In addition, the optimization process is most often carried out as a sequence, i.e. firstly the process is designed, secondly the C&I is conceptually designed, thirdly the controls are tuned and adjusted during commissioning. An example of dynamic issues is shown in figure 7. Boiler dynamics is often the bottleneck for improving the overall plant stability during volatile operation. This is typically one of many issues when operational flexibility of CHPs is optimized, e.g. as done using multivariable control in [2]. As an isolated case this will very often imply a good business case due to improved load change capability, but it will also often result in increased costs on O&M. A more integrated approach also considering other possibilities at the same plant or at other plants will most likely relax the demands on the boiler and still be able to perform well in portfolio terms.
Firing rate Brændsel
FeedFødevand water flow
Tryk Pressure
Enthalpy Entalpi
Figure 7 Optimizing dynamic response of a boiler
Tools for dynamic simulations and analysis have the following features:
Model fidelity is chosen as appropriate for the specific case and problem. For plant design it will typically include physical properties directly or indirectly. In portfolio respect plant models are often aggregated transfer function models which include process, instrumentation and controls. Model scope is typically not necessarily limited, but often aggregated non-physical models are used for large systems. Many system models suffer from stiffness and often compromises are taken.
C. Portfolio characteristics On portfolio level of energy production and supply, the focus is often on simulating operation on longer time-horizons, e.g. a one-year-period with 1 hour resolution. In some cases design is based on even rougher models, e.g. monthly average or yearly duration curves. An example is shown in figure 8 with a schematic model of a system including district heat piping, heat producing units, storage tanks and CHP units. For such examples analysis often emphasizes on identifying potential benefits in new plants, e.g. a storage tank, or in new operational strategies, e.g. by-pass operation in CHPs. In the design phase focus is purely on very superior design issues like sizing and placement of a storage tank Tools for simulation and analysis of energy supply systems are characterized by the following:
Model fidelity is always quite low because the purpose is on overview level and because very high fidelity will result in too high complexity in such a model.
HILLERØD
OMR6 etape A
VAK H
OMR4 NOVO SLK 2*7,5 MW
VAK N
SLK ny
Egedal
VAK W
OMR1 Max 60 MW
Consumers Upstream bottleneck
Vest
CTR/GLC
Nord
Max 110 MW
Hedegården SLK 2*18 MW VAK S
Vestforbrænding SLK 3*20 MW VF5+VF6
Power
Waste Summercoolers 30 MW
Min 10 MW Max 75 MW
VEKS
Figure 8 Model of energy production and supply system
In principle, model scope can vary from one site to a larger local area, to a region and even to a country. Scope will depend on the purpose and relevance from asset owner’s point of view.
V. OPTIMIZED INTEGRATED OPERATION In the operation phase it is in principle the same knowledge base as in the design phase. It is “just” an on-line real-time situation which means that all functionalities should be automatically executable and able to take in and deliver information automatically while still being robust on different types of errors and abnormalities. Taking the future needs and complexities described in section II in account this will increase the need for new functionalities significantly. Figure 9 illustrates an example of how many plants are already operated. In this case only very superior functionalities are shown and they are split between a portfolio level and an individual level, in which the shown functionalities are of course repeated on every plant. The functionalities shown in the example can briefly be described as:
Load planning should optimize the utilization of plants with respect to the markets, presently power and district heat markets, in future with increased complexity and also more market places will be relevant.
Balancing controller should make short time corrections in case of internal imbalances or in case of external requests.
Port folio level
Load planning (24 h)
Powerprices, PBA, contracts
Production plan
Correction
Balancing of production
Marketorders
Setpoint schedule for individual plants
Power production
Individual plants Performance indicators for operator
Marginal costs
Productionplan
Performance monitoring Correction of setpoints
Measure ments
Master Control Measure ments
Setpoints
C&I
Power plants
Figure 9 Structure for on-line optimized operation of plants
Master Control functionality is crucial to convert delivery demands into coordinated control and stability of the individual plant.
Performance monitoring is crucial on each plant to ensure persistent efficiency, quick handling of failures and updated calculation of plant economics.
All of the above described functionalities are often called “the economical control room”. There is a need for new developments on portfolio level as well as on individual plant level. A. Portfolio control Historically advanced portfolio control schemes have mostly been applied by large asset owners with the challenge of planning and balancing a combination of controllable plants and non-controllable plants. A specific example of this is shown in figure 10. In this case an in-house developed load planning application and MPC based balancing controller were used [9]. In general this type of control scheme will have to be developed further in future for a number of reasons:
Plant portfolios will be even more complex in composition. Plants will individually be smaller in capacity which increases the number of plants to handle but also probably reduces the total capacity. Markets will increase in complexity, e.g. power and district heat markets, and more market places will occur, e.g. on fuels and residues.
This calls for, at least, extrapolation of the present methods if up-scaling is possible, otherwise new concepts and methods are needed.
Figure 10 Portfolio control scheme [9]
B. Plant control Control in this context should be interpreted in its widest definition, i.e. as classical control, but also as tracking and supervising the state of the plant in general. Classical control, feedback loops and feedforward control, have drawn a lot of attention in research and also in practical utilization. Pragmatic utilization of advanced knowledge has evolved in industry rather than utilization of advanced control methods. This has been driven by the demand for simplicity and transparency, and will certainly be a key driver in the future as well. Supportive performance functions in different versions have also been applied widely in large plants and will probably propagate into smaller plants in new versions as the service demand from small plants increases. These functions are typically not critical on a short-term time-scale, because they focus on long-term issues or on issues that will not influence short-term plant availability. Until now these functions have primarily been used for supportive functions, i.e. always with a plant operator as receiver. In future there is a need for increasing intelligence in control while still keeping robustness and simplicity in real-time automation on a high level. Figure 11 illustrates a step in this direction. The idea is to have a bank of intelligence, different types of plant models combined with intelligence that can transform basic physical knowledge into relevant operational information. The intelligence bank is separated from real-time automation which can always operate independently of the bank. This type of architecture which separates complex knowledge and real-time critical functionality could be a possible solution to future demands.
FlexIQ Subscribers
FlexIQ Flex capacity Flex costs Fuel efficiency Maintenance costs Failure state …
Corrective actions Assets Figure 11 Intelligence bank for advanced control
VI. CONCLUSIONS The Danish energy system has been a front-runner with respect to integration of renewables, and historically the development has challenged the design and control of production plants and supply systems. This trend will continue in the coming decades and result in new challenges and opportunities within design and control. New challenges will primarily be driven by demand for optimized design and operation of distributed hybrid systems:
Demand for portfolio flexibility on energy resources and on products will increase due to the political ambitions on renewables. Demand for product diversity and plant integration will increase due to the need for improved stakeholder value in utilities. Demand for integrated utilization of new small-scale distributed technologies due to the need for reduced investment risks and need for production flexibility.
These challenges will change the research and application agenda in direction of developments crossing traditional research areas. Basically the fundamental knowledge is present, but initiatives for developments combining the basic knowledge are needed:
In the design phase there is a need for combining modelling on different time-scales and with different fidelity into overall optimized design.
In the operation phase there is a need for taking more detailed and hybrid intelligence into on-line robust utilization.
There are no ready answers to these needs, but the challenges are certainly becoming clearer, and research within systems, modelling and control are capable of taking up the challenges.
REFERENCES [1] Moelbak, T., Advanced control of superheater steam temperatures - An evaluation based on practical applications, IFAC Conference on Control of Power Systems and Power Plants 1997 (cpspp'97), pp. 161-166
[2] J.H. Mortensen, T. Moelbak, P. Andersen and T.S. Pedersen, Optimization of boiler control to improve the load-following capability of power-plant units, Control Engineering Practice 6/12, p. 1531-1539, 1998
[3] Joergensen, C., Mortensen, J.H., Moelbak, T. and Nielsen, E.O.: “Model Based Fleet Optimisation and Master Control of a Power Production System”, Proceedings of IFAC Symposium on Power Plants and Power Systems Control, Canada, June, 2006
[4] K. Edlund, J. Bendtsen, T. Moelbak, Simple Models for Model-based Portfolio Load Balancing Controller Synthesis, Proc. of the 2008 IFAC Symposium on Power Plants and Power Systems Control, 2009
[5] M. K. Petersen, L. H. Hansen and T. Mølbak, Exploring the Value of Flexibility: A Smart Grid Discussion, Proceedings of IFAC Symposium on Power Plants and Power System Control, vol. 8, no. 1, pp. 43-48, 2012
[6] Juelsgaard, M., Totu, L. C., Shafiei, S. E., Wisniewski, R., Stoustrup, J., Control Structures for Smart Grid Balancing, Proceedings of the 2013 4th IEEE PES Innovative Smart Grid Technologies Europe (ISGT Europe). IEEE Press, 2013
[7] Henrik Lund, Anders N. Andersen, Poul Alberg,Østergaard, Brian Vad Mathiesen, David Connolly, From electricity smart grids to smart energy systems – A market operation based approach and understanding, Energy,Volume 42, Issue 1, June 2012, Pages 96–102
[8] P. Meibom, K. B.Hilger, H. Madsen, D. Vinther, Energy Comes together in Denmark, ieee power & energy magazine, September/October 2013, p. 4655
[9] Edlund, K. S., Dynamic Load Balancing of a Power System Portfolio. Ph.D thesis, Aalborg University, 2010.