A cooperative net zero energy community to improve load matching

Renewable Energy 93 (2016) 1e13

Contents lists available at ScienceDirect

Renewable Energy journal homepage: www.elsevier.com/locate/renene

A cooperative net zero energy community to improve load matching ~o Martins a, b, Daniel Aelenei a, b, Celson Pantoja Lima c, d Rui Amaral Lopes a, b, *, Joa a

Centre of Technology and Systems/UNINOVA, Almada, Portugal Faculty of Science and Technology of Universidade Nova de Lisboa, Portugal c , Santar Federal University of Western Para em, Brazil d Industrial Performance Center, MIT, Cambridge, USA b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 23 October 2015 Received in revised form 13 January 2016 Accepted 15 February 2016 Available online xxx

The work reported here addresses load matching improvement in Net Zero Energy Buildings (Net-ZEBs). The related relevant literature shows that currently research work is mainly focused on improving the load matching of individual buildings. In this paper the concept of a Cooperative Net Zero Energy Community (CNet-ZEC) is introduced, extending discussion to the enhancement of load matching at a wider community level. Both building and community levels are compared in order to assess the work proposed here, through the analysis of three distinct scenarios where five Net-ZEBs work individually or in community. The results presented here were obtained through a detailed simulation based on 1-min resolution stochastic load profiles and recorded weather data. The results indicate that over the period of a year the CNet-ZEC has the potential to increase the electrical demand covered by onsite electricity generation up to 21% and the on-site generation that is used by the building up to 15%. The following elements are considered by the CNet-ZEC in order to produce those results: (i) demand heterogeneity of the buildings integrating the community; (ii) the higher number of controllable devices; and (iii) the potential higher amount of energy available to satisfy the community demand. © 2016 Elsevier Ltd. All rights reserved.

Keywords: Net zero energy buildings Renewable energy Load matching Grid interaction Demand side management

1. Introduction 1.1. Background and motivation Buildings are a main topic in European Union's (EU) energy efficiency policy, as nearly 40% of final energy consumption and 36% of greenhouse gas emissions are in houses, offices, shops and other buildings [1]. Therefore it is mandatory to improve the energy performance of European building stock to: (i) reach European Strategy for Energy and Climate Change objectives [2], and (ii) meet the longer term objectives of EU climate strategy to restrict global warming due to human-related CO2 emissions to less than 2
C [3]. In Europe, a legal framework is already in place aimed at improving the energy efficiency of buildings, which includes the directive 2010/31/EU on the energy performance of buildings [4]. Regarding Net Zero Energy Buildings (Net-ZEB), that directive targets 31st December 2020 as the horizon when all new buildings

* Corresponding author. Centre of Technology and Systems/UNINOVA, Almada, Portugal. E-mail address: [email protected] (R.A. Lopes). http://dx.doi.org/10.1016/j.renene.2016.02.044 0960-1481/© 2016 Elsevier Ltd. All rights reserved.

shall be nearly Net-ZEBs. Similar policies were also adopted in other developed countries, e.g. USA and Canada [5,6]. Although several different definitions of Net Zero Energy Buildings are found in the literature, there is a common understanding that the Net-ZEB concept regards a building that, over a certain period of time (typically one year), has a neutral energy balance (i.e. it produces as much energy as it consumes from the supply grid) when certain energy efficiency measures are implemented. Once the renewable energy produced on-site is governed by the availability of the respective primary energy source, there is often a mismatch between the local generation and the Net-ZEB demand [7]. This mismatch might impose a negative impact on the performance of the electrical grid at high penetration levels of the Net-ZEB concept, such as voltage fluctuations [8]. The importance of taking into account the match between the Net-ZEB's on-site energy production and its demand is highlighted by new policies encouraging self-consumption of on-site generation. For example, Portugal approved new legislation for distributed generation allowing self-consumption and the exporting of surplus energy [9]. In Germany, between 2009 and 2012, there was a special rate for self-consumed electricity, besides the traditional feed-

2

R.A. Lopes et al. / Renewable Energy 93 (2016) 1e13

Fig. 1. Cooperative Net Zero Energy Community: general overview.

Table 1 Summary description of scenarios used. Scenario Level 1 2

3

Description

Building Building

This is the baseline scenario. Buildings work individually and their electricity demand profile is not modified by the DSM method. In this scenario buildings act independently, as in scenario 1. However, the electricity demand profile of a specific building is modified by the application of the DSM method without taking into consideration the demand profile of other buildings. The operating times of the controllable devices are shifted by the DSM method to periods that maximize load matching of the particular building. Community All buildings work as a cooperative net zero energy community. The DSM method is applied to shift the operating times of all controllable devices to periods that maximize CAE consumption.

in tariffs. This special rate was abandoned because feed-in tariffs had become so low by 2012 that they provided a natural support for self-consumption [10]. In all electricity markets that lack netmetering or feed-in support, and where retail prices are higher than wholesale prices, a high match between production and consumption is automatically promoted [11]. 1.2. Current options to improve load matching The number of studies on load matching improvement has increased rapidly in recent years. These studies often refer to load matching as self-consumption and are mainly focused on buildings with Photovoltaic (PV) systems. Three different approaches have been studied to improve load matching: 1) energy storage [12e30]; 2) Demand Side Management (DSM) [31e35]; and 3) a combination of energy storage and DSM [36e43]. Improvement rates reported in the literature are promising but defined in distinct conditions (e.g. size of the on-site energy generation system, storage capacity, and time-resolution used on the simulation), which makes the

comparison between different studies difficult and sometimes highly subjective. Energy storage approaches use solid state batteries [19e30], hydrogen storage tanks [14], and thermal storage [12] [13] [15e18]. Solid state batteries and hydrogen storage tanks are each used to improve self-consumption by storing surplus generated energy and later allowing the consumption of this stored energy to reduce the imports when the on-site generation from renewable sources is insufficient to meet the building's demand. Regarding the thermal storage, it is normally used to anticipate the energy consumption of a certain electrical device (e.g. air-conditioner, electrical water tank or heat pump), using the thermal properties of the device itself or the respective building to reduce the import of energy later. While there is no single definition for DSM, a common concept is that it relates to a range of measures to improve the energy system from the demand side [44]. Literature reports that methods based on DSM normally improve load matching by rescheduling the demand of electrical devices (e.g. washing machines, clothes dryers, and dishwashers) to periods with a generation surplus

R.A. Lopes et al. / Renewable Energy 93 (2016) 1e13

3

Fig. 2. Simulation schema.

Table 2 PV model parameters values used for the reference system with 1.5 kWp. Parameter

Value

Unit

A

1 0.9 0.15 0.0045 25 47 20 800 0.9

m2 e e
1 C
C
C
C Wm2 e

hINV hSTC m TC,STC TC,NOCT Ta,NOCT GNOCT

ta

[31e35]. DSM can also be combined with battery storage to further improve self-consumption [36e43].

1.3. Major original contribution Relevant literature reports extensively on improving load matching of individual buildings. In this work the concept of CNetZEC is proposed aimed at the load matching of multiple buildings at the community-level, taking account of the following features to improve the load matching across all Net-ZEBs within a given community: (i) heterogeneity among Net-ZEBs energy demand; (ii) the greater number of controllable energy consuming devices; (iii) greater on-site energy generation capacity. Fig. 1 presents a general overview of CNet-ZEC. A CNet-ZEC has N buildings (with N  2) connected through the local community electrical grid. Each building generates energy on-site, described

here as Building Produced Energy (BPE), and is responsible for compensating the energy consumption of the respective devices (any device within the CNet-ZEC is associated to a specific building that compensates its energy consumption). The BPE of a specific Net-ZEB is therefore available to the entire community, contributing to the Community Available Energy (CAE). The CAE is then distributed by the community electrical grid with a building giving or receiving energy from the CAE to meet cumulative demand of all buildings. The cooperative nature of the CNet-ZEC relates to the management of each building's energy demand with the objective of improving overall load matching of all Net-ZEBs in the community. Once a CNet-ZEC is composed only by Net Zero Energy Buildings, its annual energy balance is zero. As in the Net-ZEB concept, when the CAE is not sufficient to meet the community energy demand, the remaining needed energy is imported from the electrical grid. On the other way round, excess of CAE is exported to the electrical grid. In this study the management of each Net-ZEB energy demand is performed by a DSM method, based on a genetic algorithm that controls the operating times of associated electrical devices. By shifting the operating times of these devices, the proposed DSM method aims to maximize the load matching of all buildings in the community. This paper is structured as follows. Section 2 details the methodology adopted in this study and the investigated scenarios. Section 3 presents and discusses the results of this research. Finally, section 4 presents the conclusions and introduces the future works to be explored.

4

R.A. Lopes et al. / Renewable Energy 93 (2016) 1e13

Table 3 Electrical devices considered in this study. Device

Mean cycle length [m]

Mean cycle power [W]

Standby power [W]

Controllable

Refrigerator Iron Vacuum Personal computer Printer TV Oven Microwave Dish washer Clothes Drier Washing Machine Lighting

18 30 20 300 4 73 27 30 60 60 138 Usage Dependent

110 1000 2000 141 335 124 2125 1250 1131 2500 406 Usage Dependent

0 0 0 5 4 3 3 2 0 1 1 0

NO NO NO NO NO NO NO NO YES YES YES NO

Table 4 Occupancy description. Designation

Number of occupants

Size (m2)

Building#1 Building#2 Building#3 Building#4 Building#5

1 2 3 4 5

60 80 100 120 140

houses that generate on-site energy through integrated Photovoltaic (PV) systems. Table 1 summarizes the scenarios considered. Through detailed simulation, using 1 min-resolution data over 1 year period, the load matching of five individual buildings is analyzed in the first two scenarios. At the community-level, these five Net-ZEBs form a CNet-ZEC. Each building has the following controllable domestic appliances: dishwasher, washing machine, and clothes-drier. In the second and third scenarios, the DSM method shifts the operating times of these appliances over a maximum period of 24 h Fig. 2 shows the simulation schema and sections 2.1e2.4 detail each one of the simulation phases. 2.1. Input data The simulation supporting this study uses meteorological data as input, gathered during 2013. The data covers the incident solar radiation (GT) and the ambient temperature (Ta), obtained at 1 min resolution from the meteorological station located at Faculty of Science and Technology of Universidade Nova de Lisboa (38
390 3600 N/9
120 1100 W). 2.2. Models

Fig. 3. Distribution of energy consumption for house heating by type of energy source in Portugal ([49], page 44).

2. Methodology The assessment of the CNet-ZEC concept proposed here (considering the load matching improvement) uses and compares three different scenarios. The first corresponds to the building-level without any DSM method applied, which is used as the baseline for this work. The second scenario also addresses the building-level but with the influence of a DSM method that shifts the operating time of some electrical devices to periods that maximize the load matching of the individual Net-ZEB. The third scenario covers the community-level where Net-ZEBs cooperate to improve the load matching of the whole community. In this last scenario, the energy generated by each building is made available to the entire community, contributing to the aggregated CAE, and the operating times of the controllable devices of each building are shifted to improve the consumption profile of this community available energy. The Net-ZEBs considered in this study are five detached

The electricity demand and generation profiles of each Net-ZEB are required to assess the load matching in each of the adopted scenarios. The analytical model chosen to generate the PV system output of each building is described in section 2.2.1. To generate the electricity demand profiles, the stochastic model developed by Richardson et al. [45] is used as described in section 2.2.2. 2.2.1. Photovoltaic model The PV system model uses the meteorological data described in Section 2.1 (i.e. the incident solar radiation and the ambient temperature) to produce a 1 min resolution data series for the power produced by a typical residential PV system, composed by a PV array and a power inverter. The model considers the incident solar radiation (GT) and electrical power (P) as described in Eq. (1):

P ¼ NAGT hM hINV

(1)

where N denotes the number of PV modules, A the surface area of a single module, hM and hINV the module and inverter electrical efficiency, respectively. The electrical efficiency of the module is given by Eq. (2) [46]:


 TC;NOCT  Ta;NOCT  h  1  STC hM ¼ hSTC 1 þ m Ta  TC;STC þ GT GNOCT ta (2)

R.A. Lopes et al. / Renewable Energy 93 (2016) 1e13

5

2.3. Demand side management

Table 5 GA4S parameters values. Parameter

Value

Population size Number of Generations Number of elite chromosomes Crossover Rate Mutation Rate Mutation Probability

100 150 2 0.8 0.2 0.01

PSTC where hSTC ¼ AG is the electrical efficiency of the reference STC module at Standard Test Conditions (STC), defined as the ratio between the module output power at STC and the total STC irradiance incident over the module area. In Eq. (2) m is the temperature coefficient, TC,STC and TC,NOCT are the reference cell temperature at STC and nominal operating cell temperature (NOCT), respectively, GNOCT is the solar radiation at NOCT, t is the transmittance of the cell cover and a is the solar absorptance of the PV layer. The PV system simulated has a rated peak power of 1.5 kW for N ¼ 10. In order to reach the Net-ZEB status, the PV systems from each building were linearly scaled to match its electricity demand over one year. The values of the parameters of Eqs. (1) and (2) are presented in Table 2.

2.2.2. Electricity load profiles The domestic electricity demand model presented by Richardson et al. in Ref. [45] follows a “bottom-up” approach where the individual domestic electricity loads are the basic building blocks. This model uses stochastic occupancy profiles [47] and information related to the respective activities performed by a building's occupants when at home and awake to define, with 1 min resolution, the state of each load in the building (i.e. if each load is on or off). The meteorological data received by Richardson's model is incident solar radiation (GT) used to define the lighting demand of the building. In this study, Richardson's model is used to generate the demand profiles of the five Net-ZEBs considering that the electricity demand of each building is defined by the operation of the loads presented in Table 3. Due to the occupancy dependence of this model, the probability that some devices have to operate depends on the number of occupants of the respective building. As a result, the annual energy of a specific building is also dependent on its number of occupants. The occupancy is also used to define each Net-ZEB size, which is set according to the Portuguese general regulation of urban construction. Table 4 shows the number of occupants considered in each building, associated sizes and the respective designations. From the loads shown in Table 3, it is assumed that the operating times of the “Dish Washer”, “Washing Machine”, and “Clothes Drier” are controllable by the DSM method over a 24 h period. The controllable devices chosen are normally used only once a day [48] and thus the time-shifting of 0e24 h has limited impacts on the users' comfort. However in a real world application, users' preferences regarding the stop times of controlled appliances would have to be considered. Under the Portuguese energy context 68% of the dwellings heating energy comes from biomass (see Fig. 3). Thus the present study considers that buildings use biomass based systems for space heating. Summer thermal comfort is achieved through natural ventilation and gas boilers are used for water heating, also reflecting Portuguese reality [49].

Both the second and third scenarios require a DSM method to set the starting time of the appropriate operating cycles for each controllable device. The objective is to find the starting times' vector t¼[t1,t2,…,tN] that minimizes the difference between production P and demand L(t), where N is the number of controllable devices, P¼[P1,P2,…,PG], and L(t)¼[L(t)1,L(t)2,…,L(t)G] are discrete data series representing the respective average power (in Watt) in each time interval, and G is the number of time slots regarding one day analysis (in this study G¼60x24¼1440). At the building-level P is equal to BPE while at the community-level P is equal to CAE. This objective can be described by Eq. (3). min tεΤ jP

 LðtÞj

(3)

When the DSM method is applied, the electricity demand is equal to the base load, which is not controllable, plus the sum of the shifted loads as it is described by Eq. (4). The shifted load of device i is denoted as L*i ðtÞ and it is equal to the original load of the controllable device i shifted by Refs. ti; this is L*i ðtÞ ¼ Li ðt  ti Þ. Regarding the load considered by the DSM method, the main difference between building and community levels is the number of controllable devices. Additionally, it should be noted that at the community-level the base load is the sum of the base loads of all Net-ZEBs integrating the CNet-ZEC.

LðtÞ ¼ LB þ

N X

Li ðtÞ

(4)

i¼1

The main challenge of our DSM method concerns solving Eq. (3). As in any scheduling problem involving finding an optimal time allocation for a large number of controllable units, the number of possible combinations of L*i ðtÞ grows exponentially with the number of considered controllable devices. Therefore, to find the optimal vector of operating starting times t, testing all the possible combinations, might not be a feasible option. For instance, to find the operating starting times of N controllable devices, having a time scheduling window of 24 h with a 1 min resolution, would require testing 1440N combinations. Taking this into account, a Genetic Algorithm for Scheduling (GA4S) was developed to find the vector of operating starting times t that satisfies Eq. (3). Due to its capabilities to deal with the combinatorial nature of the problem and the size of the search space, GA4S is not a “greedy animal” in terms of computational effort when solving the combinatorial optimization problem. GA4S is based on a selection process that mimics biological evolution, representing a vector of operating start times t as a chromosome composed by N genes, where each gene refers to an operation starting time ti. The algorithm starts by creating a random initial population of chromosomes (i.e. vectors of operation starting times t). Then over a specified number of generations, GA4S creates a sequence of new populations with the objective of finding the chromosome that satisfies Eq. (3). At each step, GA4S uses the individuals in the current generation to create the next population according to the following sequence: 1. Evaluating the quality of each chromosome in the current population by computing the fitness function, which measures how similar the resulting demand L(t) is to the production P (note that P ¼ BPE at building level; P ¼ CAE at community level). 2. Selecting the chromosomes of the current population that will produce the next generation (so-called Parents). The GA4S uses a “roulette wheel” based selection where the area of the section

6

R.A. Lopes et al. / Renewable Energy 93 (2016) 1e13

Fig. 4. Load and BPE profiles at building-level. a) before applying the DSM method. b) after applying the DSM method.

Fig. 5. Load and CAE profiles at community-level. a) before applying the DSM method. b) after applying the DSM method.

Table 6 Daily load and supply cover factors. Designation

Building#1 Building#2 Building#3 Building#4 Building#5

Building-level without DSM

Building-level with DSM

Community-level

gload,day

gsupply,day

gload,day

gsupply,day

gload,day

gsupply,day

8 11 8 11 21

13 23 15 46 62

15 32 26 14 30

26 64 52 60 88

49

75

of the wheel corresponding to a parent is proportional to its result of the fitness function. Then GA4S generates a random number to select one of the sections, resulting on a probability to select a specific chromosome equal to its section area.

3. Based on the results of step 1, selecting the chromosomes in the current population with the best quality to directly integrate the next generation. This process is called Elitism and ensures that the results of the next generation are at least as good as the ones of the current population. 4. Generating the remaining individuals of the new generation through two distinct processes: crossover and mutation. In the first one, a chromosome of the new generation is formed by combining a pair of parents randomly chosen, resulting on a chromosome composed by operating start times of two distinct vectors t. In the second process, GA4S creates new chromosomes by randomly changing some operating start times of certain individual parents. Table 5 shows the specific parameters used by the GA4S presented in this study'

R.A. Lopes et al. / Renewable Energy 93 (2016) 1e13

7

Fig. 6. Daily number of active controllable devices. a) building-level. b) community-level.

2.4. Results aggregation and analysis

3. Results and analysis

The last step of the simulation process handles the aggregation of results for the three scenarios and the respective analysis. Results are based on the metrics described here and the outputs addressed in Section 3. At any instant, the absolute load instantly matched by the local production, denoted by M, is delimited by the load itself or by the available on-site generation. It can be described as:

In this section, unless otherwise stated, the results regarding the building-level refer to Building#5. Considering that at the buildinglevel the results do not vary significantly from building to building, Building#5 was chosen because it presents the best load matching related results among all individual buildings. The results presented start by showing the affects introduced by our DSM method on the electricity demand. Then the results regarding the load matching are presented and analyzed, comparing building and community levels. Once the load matching performance of the building is highly related with its grid interaction, this section ends by showing some grid interaction related results for both levels.

MðtÞ ¼ minðLðtÞ; PðtÞÞ

(5)

Eq. (5) gives the load instantly matched, measured in Watt. However, similarly to [11] and [35], the load matching in this study is characterized by two distinct factors, namely: the load cover factor, defined as the percentage of the electrical demand that is covered by the on-site electricity generation; and the supply cover factor, defined as the percentage of the on-site generation that is directly used by the building. These two factors can be denoted by Eqs. (6) and (7), where t1 and t2 define the period of analysis.

Pt2 t¼t MðtÞ  100 gload ¼ Pt2 1 t¼t1 LðtÞ

(6)

Pt2 t¼t MðtÞ gsupply ¼ Pt2 1  100 t¼t1 PðtÞ

(7)

It is also important to characterize the power exporting profile of the building and the respective accumulated energy exported, which are given by Eqs. (8) and (9), respectively. In Eq. (9) the accumulated energy exported is expressed in Wh, and Dt refers to the time interval used in the simulation, expressed in hours (in this study Dt¼1/60).

netðtÞ ¼ PðtÞ  LðtÞ 2 NET ¼ Dt4

t2 X t¼t1

PðtÞ 

(8) t2 X t¼t1

3 LðtÞ5

(9)

3.1. Original and shifted load profiles Figs. 4 and 5 show the electricity demand profiles for a certain day of November at building and community levels, respectively, before and after the application of the DSM method. Additionally, those figures show the respective on-site power generation profiles (represented by the dotted lines). At the building-level the dotted lines represent the BPE while at the community-level these lines represent the CAE. For the specific day represented in Fig. 4, the DSM method shifted the operating start times of the three controllable devices of the Building#5 to periods where the use of on-site produced energy is increased, as shown in Fig. 4a and b. By shifting the operating start times of these devices, the daily load cover factor was improved by 9% (from 21% to 30%), while the daily supply cover factor was increased by 26% (from 62% to 88%). Regarding the community-level, presented in Fig. 5, the DSM method shifted the operating start times of eight controllable devices. Fig. 5a and b shows that the operating start times of these devices were shifted to periods that increase the consumption of the CAE. As a result, considering the initial individual performance of Building#5 (Scenario 1), the daily load and supply cover factors improved by 28% (from 21% to 49%) and 13% (from 62% to 75%), respectively. Table 6 presents the daily load and supply cover factors for all buildings operating individually and as CNet-ZEC. Due to the availability of more controllable devices at the

8

R.A. Lopes et al. / Renewable Energy 93 (2016) 1e13

community-level, the capacity to shift energy demand to periods with greater energy surplus is increased. While at the buildinglevel the possible number of active controllable devices during one day is three, at the community-level this number raises to fifteen. However as it can be seen in Fig. 6b (showing the daily number of active controllable devices during a year period), at the community-level the maximum registered number of effective controllable devices in one day was only 11. This is due to the stochastic behavior of the model used to generate the load profiles, where each device has a certain probability to operate during a day. As a consequence there was not a single day where all controllable devices of the five buildings were active. Fig. 7 presents the yearly mean demand and generation profiles for all the considered scenarios. By analyzing Fig. 7a, it can be seen that the demand profile at the building-level has a reduced value during the night, followed by a morning peak at breakfast time. During the day, the energy demand is relatively even, until midafternoon, when it rises towards the evening peak. After this second peak, the demand falls again until reaching the night-time lower values. Still on the building-level, Fig. 7b shows the effects of the DSM method on the yearly mean demand profile. It is clear the demand shifting to periods with higher generation of BPE. At the community-level it is evident the demand peaks around noon due to the operation shifting of the. controllable devices to periods with higher availability of CAE (Fig. 7c). The high demand in the evening results from the base load of all buildings in the community. To further reduce the difference between the yearly mean demand and generation profiles, additional controllable devices would have to be considered. 3.2. Load matching

Fig. 7. Yearly mean demand and generation profiles. a) scenario 1. b) scenario 2. c) scenario 3.

Despite the net zero balance over a year period, the instantaneous load matching of a Net-ZEB is normally far from being perfect and it can be verified by the load cover and supply factors. Fig. 8 shows these two factors for all the scenarios considered, calculated on a monthly basis. To better understand the seasonal variation of gload and gsupply, Figs. 9 and 10 present the mean of the referred factors calculated on an hourly basis over four months (January, April, July, and October) representing the different seasons of the year. Table 7 summarizes all the load matching related results for the analyzed year. The significant seasonal variation of gload and gsupply registered in the three scenarios (Figs. 8e10) is the consequence of the yearly solar azimuth and altitude variation registered in Portugal throughout the year. The load cover factor has its higher value during the summer months due to the greater availability of PV generation and the lower values during the winter, when the PV generated energy is reduced. Regarding the supply load cover, it is higher during the winter when there is less on-site produced energy to be locally consumed, and lower during the summer months when energy surpluses are common. As shown in Fig. 8, the CNetZEC presents the best results regarding both gload and gsupply during all months analyzed, while the individual building without DSM presents the worst results. Over the entire year, gload and gsupply were improved by 18% and 14% respectively, comparing scenarios 1 and 3 for Building#5 (Table 7). Fig. 9 shows that gload is higher around noon and zero at night, following the availability of the on-site produced energy. Scenario 3 has the best results during all seasons. This is due to the greater availability of on-site produced energy and to the heterogeneity among the energy demands of the five buildings comprising the community. The CNet-ZEC reaches values of gload close to 100% at noon in the winter, spring, and autumn and complete selfsufficiency during the summer from 11am to 4pm. Comparing the

R.A. Lopes et al. / Renewable Energy 93 (2016) 1e13

9

Fig. 8. Load matching factors. a) load cover factor 1. b) supply cover factor.

building-level without DSM and the CNet-ZEC, the biggest difference occurs during the winter when the former reaches max(gload)¼64% and the latter max(gload)¼95%. Regarding the supply cover factor, it is minimal around noon due to the greater quantity of on-site generated energy to be locally consumed and maximum during the night when there is no BPE or CAE (Fig. 10). In this case the best results are obtained by the building-level with DSM and by the CNet-ZEC. For these two scenarios, the registered local maximum by noon is a consequence of the load shifting for this period of the day. The second scenario shows better results around noon in spring, summer and autumn than the CNet-ZEC, due the greater availability of on-site generated energy of the latter. However, if considering the entire day, the CNet-ZEC still presents the best results. Although the studied CNet-ZEC (third scenario) has a fixed number of buildings (i.e. 5 Net-ZEBs), it was found that community load matching results are related with the number of buildings integrating the community. Fig. 11 presents the yearly load and supply cover factors as a function of the number of buildings within the CNet-ZEC. Both indicators are monotonically increasing functions of the number of Net-ZEBs integrating the community, delimited by horizontal asymptotes with distinct values. For a large number of buildings, the load cover factor is constrained by the ratio between the amount of energy that can be consumed during the generation period and the energy consumed during early morning and evening. Therefore, to increase the load cover factor for a high number of Net-ZEBs, the level of energy demand satisfied during CAE generation period should be increased. This could be achieved using a higher number of controllable devices in order to increase the ration of controllable demand. Regarding the supply cover factor, its value is limited by the amount of generation that can be instantly consumed within the community. The supply cover factor's horizontal asymptote reflects the balance between the additional amount of CAE introduced by the addition of a new NetZEB and the additional energy demand introduced by the same building during the CAE generation period. To increase the value of the referred asymptote, electricity consumption during the CAE generation period should be increased. This would require a demand shift using a higher number of controllable devices.

3.3. Grid interaction To achieve the net zero balance, a Net-ZEB uses the supply grid as an unlimited virtual storage system. Over the year, the on-site generation that exceeds the demand is exported to the supply grid and the same amount of energy is received back from it when the demand exceeds the on-site generation. On a monthly basis, this can be seen by the net exporting profile of Building#5, as presented in Fig. 12. From April to September, Building#5 generates more energy than it consumes and exports the energy surplus to the grid. The same amount of energy is imported back from the grid from October to March. Fig. 13 shows in more detail, with 1 min resolution, the net exporting profile of the three scenarios over the entire year (525600 data points per scenario). In all cases, it is evident the exporting behavior during the day time and the importing profile when there is no on-site energy generation. The increased number of hours with PV generation during the summer as well as the higher amplitude of the exports is evident. At the building-level, Fig. 13b shows the operating start times of the controllable devices shifted to periods around noon, denoted by the darker dots on the carpet plot (energy importing). These importing periods occur because the building demand is higher than the BPE even during the maximum on-site generation period. At the community-level (Fig. 13c) it is not common to have importing periods during the day time once the CAE is higher and sufficient to satisfy the energy demand of the entire community. The shifting of the operating start times of the controllable devices to periods with higher values of CAE is also evident from Fig. 13c. Comparing building and community levels, it can be seen that the net profiles of the latter are smoother than the former. This is mostly due to the existing heterogeneity among the demand profiles of the Net-ZEBs integrating the community and to the higher availability of on-site generated energy that avoids frequent energy import periods during the day time. 4. Conclusions and future work In this work, three different scenarios concerning the assessment of the load matching in Net Zero Energy Buildings were discussed. The first addresses the common case of an individual Net-

10

R.A. Lopes et al. / Renewable Energy 93 (2016) 1e13

Fig. 9. Mean of hourly load cover factor. a) January. b) April. c) July. d) October. Fig. 10. Mean of hourly supply cover factor. a) January. b) April. c) July. d) October.

R.A. Lopes et al. / Renewable Energy 93 (2016) 1e13 Table 7 Yearly load and supply cover factors. Designation

Building-level without DSM

Building-level with DSM

Community-level

gload,year gsupply,year gload,year gsupply,year gload,year gsupply,year Building#1 Building#2 Building#3 Building#4 Building#5

26 27 28 27 28

24 25 25 24 25

31 34 36 36 38

30 32 33 33 34

46

39

Fig. 11. Load matching as a function of the number of Net-ZEBs integrating the community.

Fig. 12. Accumulated energy exported.

ZEB without any DSM method applied. Still at the building-level, the second scenario concerns an individual Net-ZEB in which the load diagram is modified by a DSM method aiming to improve the local consumption of on-site produced energy. The final scenario refers to the community-level and it is based on the CNet-ZEC concept, in which all buildings cooperate by sharing the overall on-site produced energy and by changing their demand profiles

Fig. 13. Net power balance. a) scenario 1. b) scenario 2. c) scenario 3.

11

12

R.A. Lopes et al. / Renewable Energy 93 (2016) 1e13

with the objective of improving the load matching of all Net-ZEBs integrating the community. To assess the load matching of each scenario, this study used two different metrics: (i) the load cover factor; and (ii) the supply cover factor. Since the load matching of a building is closely related with its grid interaction, this study also addressed the grid interaction regarding each scenario. The CNet-ZEC achieved the best load matching related results with annual load and supply cover factors of 46% and 39%, respectively. Comparing the building-level with no DSM method, the CNet-ZEC improved the annual load cover factor by 18e20% and the supply cover factor by 14e15%. Regarding the second scenario, the CNet-ZEC improved these factors by 8e15% and 5e9%, respectively. The improvements on the load matching introduced by the CNet-ZEC are mainly related to: (i) the community buildings having different demand profiles; (ii) the higher number of devices controlled by the DSM method; and (iii) the potential higher amount of energy available to satisfy the community demand. Considering a real world application of the CNet-ZEC concept, a computational platform that supports the real time cooperation among different Net-ZEBs is to be developed. Additionally, other options to induce the cooperation among the buildings integrating the CNet-ZEC, beyond the developed DSM method, are to be studied.

[18]

[19]

[20]

[21]

[22]

[23]

[24] [25]

[26] [27]

Acknowledgements This study was made possible by doctoral grant SFRH/BD/87733/ 2012 issued by the (Portuguese) Foundation for Science and Technology.

[28] [29] [30]

References [31] [1] Eurostat, Energy, Transport and Environment Indicators, 2012. [2] M. da Graça Carvalho, EU energy and climate change strategy, Energy 40 (1) (2012) 19e22. [3] International Energy Agency, World Energy Outlook, 2012. [4] European Union, Directive 2010/31/EU of the European Parliament and of the Council of 19 May 2010 on the energy performance of buildings, Off. J. Eur. Union (2010) 13e35. [5] P. Torcellini, S. Pless, M. Deru, Zero Energy Buildings : a Critical Look at the Definition, 2006. [6] A. Athienitis, W. O'Brien, Modeling, Design, and Optimization of Net-zero Energy Buildings, Wiley, 2015.  [7] L. Montuori, M. Alc azar-Ortega, C. Alvarez-Bel, A. Domijan, Integration of renewable energy in microgrids coordinated with demand response resources: Economic evaluation of a biomass gasification plant by Homer Simulator, Appl. Energy 132 (Nov. 2014) 15e22. [8] M.H. Bollen, Integration of Distributed Generation in the Power System, John Wiley & Sons, 2011. rio do Ambiente, Ordenamento do Territo rio e Energia, Decreto-Lei no [9] Ministe rio da República, 2014, pp. 5298e5311. 153/2014 de 20 de outubro, Dia [10] International Energy Agency (IEA), Iea-pvps. Trends in Photovoltaic Applications 2013: Survey Report of Selected IEA Countries between 1992 and 2012, Rep. No. IEA-PVPS T1e232013, 2013. n, J. Candanedo, K.B. Lindberg, Analysis of grid interaction [11] J. Salom, J. Wide indicators in net zero-energy buildings with sub-hourly collected data, Adv. Build. Energy Res. 136 (2014) 1e18. [12] C.J.C. Williams, J.O. Binder, T. Kelm, Demand side management through heat pumps, thermal storage and battery storage to increase local selfconsumption and grid compatibility of PV systems, in: In 2012 3rd IEEE PES Innovative Smart Grid Technologies Europe (ISGT Europe), 2012, pp. 1e6. [13] F. Sossan, A.M. Kosek, S. Martinenas, M. Marinelli, H. Bindner, Scheduling of domestic water heater power demand for maximizing PV self-consumption using model predictive control, in: In IEEE PES ISGT Europe 2013, 2013, pp. 1e5. [14] D. Parra, G.S. Walker, M. Gillott, Modeling of PV generation, battery and hydrogen storage to investigate the benefits of energy storage for single dwelling, Sustain. Cities Soc. 10 (Feb. 2014) 1e10. [15] R. Thygesen, B. Karlsson, Simulation and analysis of a solar assisted heat pump system with two different storage types for high levels of PV electricity selfconsumption, Sol. Energy 103 (May 2014) 19e27. [16] G. Reynders, T. Nuytten, D. Saelens, Potential of structural thermal mass for demand-side management in dwellings, Build. Environ. 64 (Jun. 2013) 187e199. [17] U.I. Dar, I. Sartori, L. Georges, V. Novakovic, Advanced control of heat pumps

[32]

[33]

[34]

[35] [36]

[37]

[38]

[39]

[40]

[41]

[42]

[43]

for improved flexibility of Net-ZEB towards the grid, Energy Build. 69 (2014) 74e84. D. Vanhoudt, D. Geysen, B. Claessens, F. Leemans, L. Jespers, J. Van Bael, An actively controlled residential heat pump: Potential on peak shaving and maximization of self-consumption of renewable energy, Renew. Energy 63 (Mar. 2014) 531e543. R. Kanamori, T. Ito, N. Mizutani, Agent-based electrical power management model for houses equipped with storage battery and photovoltaic units, in: In 2011 IEEE International Conference on Service-oriented Computing and Applications (SOCA), 2011, pp. 1e5. S. Park, Y. Wang, Y. Kim, N. Chang, M. Pedram, Battery management for gridconnected PV systems with a battery, in: In Proceedings of the 2012 ACM/IEEE International Symposium on Low Power Electronics and Design - ISLPED ’12, 2012, p. 115. n, Quantifying self-consumption of on-site J. Munkhammar, P. Grahn, J. Wide photovoltaic power generation in households with electric vehicle home charging, Sol. Energy 97 (Nov. 2013) 208e216. G. Mulder, D. Six, B. Claessens, T. Broes, N. Omar, J. Van Mierlo, The dimensioning of PV-battery systems depending on the incentive and selling price conditions, Appl. Energy 111 (Nov. 2013) 1126e1135.  , A. Martínez-Gracia, Sizing criteria of A.A. Bayod-Rújula, M.E. Haro-Larrode hybrid photovoltaic-wind systems with battery storage and self-consumption considering interaction with the grid, Sol. Energy 98 (2013) 582e591. M. Bruch, M. Müller, Calculation of the Cost-effectiveness of a PV Battery System, Energy Procedia 46 (2014) 262e270. M. Braun, K. Büdenbender, D. Magnor, A. Jossen, Photovoltaic Selfconsumption in Germany - Using Lithium-ion Storage to Increase Selfconsumed Photovoltaic Energy, Jan. 2009. G. Mulder, F. De Ridder, D. Six, Electricity storage for grid-connected household dwellings with PV panels, Sol. Energy 84 (7) (Jul. 2010) 1284e1293. M. Osawa, K. Yoshimi, D. Yamashita, R. Yokoyama, T. Masuda, H. Kondou, T. Hirota, Increase the rate of utilization of Residential photovoltaic generation by EV charge-discharge control, in: In IEEE PES Innovative Smart Grid Technologies, 2012, pp. 1e6. J. Li, M.A. Danzer, Optimal charge control strategies for stationary photovoltaic battery systems, J. Power Sour. 258 (Jul. 2014) 365e373. J. Weniger, T. Tjaden, V. Quaschning, Sizing of Residential PV Battery Systems, Energy Procedia 46 (2014) 78e87. E. Waffenschmidt, Dimensioning of Decentralized Photovoltaic Storages with Limited Feed-in Power and their Impact on the Distribution Grid, Energy Procedia 46 (2014) 88e97. M. Chaabene, M. Ben Ammar, A. Elhajjaji, Fuzzy approach for optimal energymanagement of a domestic photovoltaic panel, Appl. Energy 84 (10) (Oct. 2007) 992e1001. n, E. Wa €ckelgård, P.D. Lund, Options for improving the load matching J. Wide capability of distributed photovoltaics: Methodology and application to highlatitude data, Sol. Energy 83 (11) (Nov. 2009) 1953e1966. Y. Zong, L. Mihet-Popa, D. Kullmann, A. Thavlov, O. Gehrke, H.W. Bindner, Model Predictive Controller for Active Demand Side Management with PV self-consumption in an intelligent building, in: In 2012 3rd IEEE PES Innovative Smart Grid Technologies Europe (ISGT Europe), 2012, pp. 1e8. n, A flexible Markov-chain model for simulating deJ. Munkhammar, J. Wide mand side management strategies with applications to distributed photovoltaics, in: Proc.. Of the World Renewable Energy Forum, 2012. n, Improved photovoltaic self-consumption with appliance scheduling J. Wide in 200 single-family buildings, Appl. Energy 126 (2014) 199e212. C. Clastres, T.T. Ha Pham, F. Wurtz, S. Bacha, Ancillary services and optimal household energy management with photovoltaic production, Energy 35 (1) (Jan. 2010) 55e64.  Gutie rrez, F. Monasterio-Huelin, M. Castillo-Cagigal, E. Matallanas, A. ~ o-Martín, D. Masa-Bote, J. Jime nez-Leube, Heterogeneous collaboE. Caaman rative sensor network for electrical management of an automated house with PV energy, Sensors 11 (2011) 11544e11559. rrez, F. Monasterio-Huelin, E. Caaman ~ o-Martín, M. Castillo-Cagigal, A. Gutie nez-Leube, A semi-distributed electric demand-side manageD. Masa, J. Jime ment system with PV generation for self-consumption enhancement, Energy Convers. Manag. 52 (7) (Jul. 2011) 2659e2666. ~ o-Martín, E. Matallanas, D. Masa-Bote, M. Castillo-Cagigal, E. Caaman rrez, F. Monasterio-Huelin, J. Jime nez-Leube, PV self-consumption A. Gutie optimization with storage and Active DSM for the residential sector, Sol. Energy 85 (2011) 2338e2348, 2011. S. Sharma, D.W. Galipeau, Optimization of residential grid-tied PV systems without net-metering using load management, in: In 2012 IEEE Third International Conference on Sustainable Energy Technologies (ICSET), 2012, pp. 6e11. rrez, F. Monasterio-Huelin, E. Matallanas, M. Castillo-Cagigal, A. Gutie ~ o-Martín, D. Masa, J. Jime nez-Leube, Neural network controller for E. Caaman Active Demand-Side Management with PV energy in the residential sector, Appl. Energy 91 (1) (Mar. 2012) 90e97. n, J. Munkhammar, Evaluating the benefits of a solar home energy J. Wide management system: impacts on photovoltaic power production value and grid interaction, in: Proc.. Of the ECEEE 2013 Summer Study, 2013. N. Femia, D. Toledo, W. Zamboni, Storage unit and load management in photovoltaic inverters for residential application, in: In IECON 2013-39th Annual Conference of the IEEE Industrial Electronics Society, 2013, pp.

R.A. Lopes et al. / Renewable Energy 93 (2016) 1e13 6800e6805. [44] P. Palensky, D. Dietrich, Demand Side Management: Demand Response, Intelligent Energy Systems, and Smart Loads, Ind. Info. IEEE Trans. 7 (3) (2011) 381e388. [45] I. Richardson, M. Thomson, D. Infield, C. Clifford, Domestic electricity use: A high-resolution energy demand model, Energy Build. 42 (10) (2010) 1878e1887. [46] J.A. Duffie, W.A. Beckman, Solar Engineering of Thermal Processes, third ed., Wiley, 2006.

13

[47] I. Richardson, M. Thomson, D. Infield, A high-resolution domestic building occupancy model for energy demand simulations, Energy Build. 40 (8) (2008) 1560e1566. [48] D. Livengood, R. Larson, The Energy Box: Locally Automated Optimal Control of Residential Electricity Usage, Serv. Sci. 1 (1) (2009) 1e16. [49] National Statistics Institute, National Directorate of Energy and Geology, ”Survey of Energy Consumption in the Domestic Sector“, Official Statistics, 2010.