Você está na página 1de 10

Energy 42 (2012) 321e330

Contents lists available at SciVerse ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Optimal control of a residential microgrid


Phillip Oliver Kriett*, Matteo Salani
Dalle Molle Institute for Articial Intelligence, Galleria 2, 6928 Manno-Lugano, Switzerland

a r t i c l e i n f o
Article history: Received 16 June 2011 Received in revised form 26 February 2012 Accepted 21 March 2012 Available online 21 April 2012 Keywords: Residential microgrid Electrical Thermal Supply side management Demand side management Mixed integer linear programming

a b s t r a c t
We propose a generic mixed integer linear programming model to minimize the operating cost of a residential microgrid. We model supply and demand of both electrical and thermal energy as decision variables. The modeled microgrid is operated in grid-connected mode. It covers solar energy, distributed generators, energy storages, and loads, among them controllable load jobs released by home appliances and electric vehicles. We propose a model predictive control scheme to iteratively produce a control sequence for the studied microgrid. Our case study reveals the performance of minimum cost control by comparison with benchmark control policies. We consider three price scenarios in our analyses which include two market-based scenarios. Numerical results from our study indicate savings in annual operating cost between 3.1 and 7.6 percent. 2012 Elsevier Ltd. All rights reserved.

1. Introduction Residential energy needs typically include electricity and heat. Traditionally, these needs are met by large-scale centralized power generation and on-site heaters that burn some type of fossil fuel. Nowadays, homeowners are also interested in the use of renewable energy resources, distributed power generation, energy storage technology, and advanced automated systems for measurement and control. Some of the reasons for this include tightened regulation of carbon emissions, governmental support schemes to improve energy efciency, and rising energy prices. A microgrid is formed by integrating distributed generators, energy storage devices, and electric loads. In addition to the control of local voltage and frequency, a microgrid controller facilitates supply side management and demand side management [15]. Clearly the concept of a microgrid can be applied to residential energy systems. When a microgrid includes combined heat and power (CHP) generation, electrical and thermal energy balances are interconnected. In this paper, we extend the concept of a residential

* Corresponding author. Present address: TUM School of Management, Technical University Munich, Arcisstrasse 21, 80333 Munich, Germany. Tel.: 49 89 289 24874; fax: 49 89 289 24872. E-mail addresses: phillip.kriett@tum.de (P.O. Kriett), matteo.salani@idsia.ch (M. Salani). URLs: http://www.scm.wi.tum.de/index.php?idkriett, http://www.idsia.ch/ salani/ 0360-5442/$ e see front matter 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.energy.2012.03.049

microgrid to supply side management and demand side management of both electrical and thermal energy. Several factors, such as the availability of renewable energy resources, energy consumption, and price of energy, can change stochastically over time, affecting optimal control of a residential microgrid. We expect that the use of electronic systems in measurement and control, abundant computing power, and high speed transfer of information will make automated control of residential microgrids possible in real-time. We propose a model that facilitates minimum cost control of a residential microgrid and assess its economic value from a homeowners point of view. In this context, scientic literature often refers to the unit commitment problem (UCP) and the economic dispatch problem where the latter is commonly solved as a subproblem of the former. The question is which dispatchable unit should run at what time (i.e. the unit commitment problem) and at what level (i.e. the economic dispatch problem) such that energy requirements and technical constraints are satised while costs are minimized [33]. Optimal control of microgrids is an active eld of research. The most common objective is to minimize operating cost. Published studies mainly differ in their solution techniques and scope of the modeled microgrid. Proposed solution techniques include mathematical programming [3,6e9,16,21], heuristics [5,17,18,22,32], and priority rules [27]. The microgrid concept originates from electrical sub-networks. In these networks, minimum cost control affects the operation of distributed power generators, electrical storages, and

322

P.O. Kriett, M. Salani / Energy 42 (2012) 321e330

electric power requirements [17,22]. Residential microgrids, however, include both electrical and thermal requirements. This suggests the use of combined heat and power generation [5] in connection with thermal energy storage [8,21]. Refs. [6,7,9,18] propose management of electrical and thermal supply considering renewable energy resources, distributed generators, and energy storages. Fewer studies look at the simultaneous management of supply and demand considering both electrical and thermal energy in a residential microgrid [3,16]. These studies fail, however, to investigate the opportunities due to supply side management of components such as thermal solar power, CHP generation, and thermal energy storage. Our paper lls this gap with a modeling and optimization approach that considers the supply and demand of both electricity and heat in a well-equipped residential microgrid. Reported savings in operating cost of microgrids under optimal control vary remarkably depending on microgrid conguration, optimization method, and benchmark model. Cost reductions range from single digit percentages up to 35 percent [3,16,21,22,27]. Ristic et al. report a 17.24 percent cost reduction compared to heatled control of a CHP unit embedded in a grid-connected microgrid with boiler, thermal storage, and bidirectional electricity exchange with the main grid [21]. For a residential microgrid that includes a wind turbine, batteries, a controllable air conditioner, and a connection to the main grid, Livengood and Larson report a 40 percent cost reduction as a result of optimal control of storage and air conditioning [16]. Previous studies consider cost reduction statements based on analyses that observe microgrid operation during a few hours up to four days. Our results are obtained using one-year-long time series data based on historical records and technical specications of commercially available state-of-the-art technology. Demand side management in residential microgrids is about controlling electrical and thermal consumer loads. Technical specications of consumer loads as well as user preferences can be very diverse. The literature in this area proposes models of reducible loads that do not require ex post feed [3] and reducible loads that require ex ante feed or ex post feed [16]. We propose further models of shiftable and non-interruptible as well as shiftable, interruptible, and reversible load jobs that have an earliest start date and a deadline. In summary, the scientic contributions of this study include:  a generic mixed integer linear programming model to nd the minimum cost operating schedule of a residential microgrid;

 an integrated model of electrical as well as thermal supply and demand including load jobs released by home appliances and electric vehicles;  a comparative evaluation of minimum cost operation of a wellequipped residential microgrid over one year with the help of a realistically modeled case. The rest of the paper is structured as follows. In Section 2, we describe the problem to be studied. In Section 3, we formulate a comprehensive mixed integer linear programming (MILP) model of the stated problem. It includes models of microgrid components, energy balance constraints, and the objective function. Next, we evaluate the proposed model with the help of a realistic case. In Section 4, we describe the setting of model parameters. The problem is solved by a MILP solver, embedded in a model predictive control (MPC) scheme [2]. In Section 5, we present numerical results about operating cost and utilization levels of selected units considering three electric rate scenarios. Financial savings are revealed by comparison with benchmark models. We examine the sensitivity of the model to variations in the estimated parameters and discuss the solution quality of the MPC scheme. We conclude by summarizing the most interesting ndings in Section 6. 2. Problem description We consider a residential microgrid equipped with photovoltaic panels, thermal solar panels, combined heat and power units, boilers, and electrical as well as thermal energy storages. Consumer loads include shiftable loads such as dishwashers, washing machines, and dryers, preemptive loads such as as electric vehicles, refrigerators, and heat requirements, and non-controllable loads such as cooking, lighting and entertainment. The residential microgrid is operated in grid-connected mode and electricity can be imported and exported through the point of common coupling (PCC). The price of energy is assumed to be known in advance. We ignore subsidies and other governmental interventions affecting fair market prices. This situation is approximated by assuming the same rate for electricity imports and for electricity exports at every point in time. The description of the microgrid system is presented in Fig. 1. The thin lines in Fig. 1 represent electrical energy ows while the thick lines represent thermal energy ows; dashed lines represent the ow of fossil fuels, such as natural gas; double arrows indicate that energy can be both imported and exported by the connected

Fig. 1. Microgrid model.

P.O. Kriett, M. Salani / Energy 42 (2012) 321e330

323

device. The reader will notice that the thermal system is connected indirectly with the electrical system via the combined heat and power unit. The optimal control problem of the given residential microgrid receives as input the energy costs, the technical specications of involved units, and energy demands expressed in terms of energy requirement and desired time of fulllment. The control decisions include the dispatch level of generators, the state of charge of storages, the operating schedules of controllable consumer loads, and the energy exchange at the PCC. Note that optimal control of microgrid units should not counteract user preferences and habits. The problem is to identify the minimum cost operating schedule for such a well-equipped residential microgrid. This problem is of interest because recent technological developments, e.g. low-cost information technology, have opened up new possibilities for advanced planning and control of supply and demand even at the residential scale. 3. Problem modeling The unit commitment problem in a residential microgrid is modeled as a mixed integer linear program. The problem is modeled over a nite prediction horizon of length n with discrete time steps t T (T : 1; 2; .n). Dt is a constant and models the time interval t ; t 1 for any t T . For simplicity, the model is composed of a limited number of parameter types:

Table 1 Superscripts. Modeled units (rst position) Superscript ael boi chp dsw ees esp evh fri hhw mgd ngs tdu tes tsp Meaning Aggregated electric load Boiler Combined heat and power Dishwasher Electrical energy storage Electrical solar panel Electric vehicle Refrigerator Heat and hot water Main grid Natural gas supply Thermal dump Thermal energy storage Thermal solar panel Features and ows (following positions) Superscript buf csc eex eim due gex gim htp lb oim pop prp rel sdc tex tim ub Meaning Buffer Cold start cost Electrical export Electrical import Due Natural gas export Natural gas import Heat-to-power Lower bound Overall import Postponed Preponed Release Self discharging rate Thermal export Thermal import Upper bound

C E P K L U V

Cost Energy Power Release date, due date Length On/off status Start Efciency

R!0 R!0 R!0 f0; 1g N!0 f0; 1g f0; 1g 1; 100

CHF/kWh, CHF/(kWh Dt ), CHF/Start kWh kW No unit Number of time steps No unit No unit %

arbitrary unit into the cables and pipes of the microgrid while exports refer to the opposite. In the following we present generic MILP models of the modeled units, balance constraints, and the objective function. In general, the model is based on energy balance equations. We approximate laws of both thermodynamics and electronics by a conservative estimation of parameters. In other words, efciency parameters have been set rather low in order not to overestimate the performance of exchanges and transfers of energy (see also Ref. [12]). The boilers thermal power output equals the natural gas input multiplied by conversion efciency hboi (1). Technical constraints on output modulation are modeled in (2). Cold starts of the boiler in (3). are tracked by vboi t
;gex boi ;tim h pboi pboi t t boi;tim uboi t Pmin ;tim pboi t boi;tim uboi t Pmax

(1) (2) (3)

Likewise, types of decision variables are:

vboi ! uboi uboi t t t 1


uboi t 1
boi Uini

e p u v

Energy Power On/off status Start

R!0 R!0 f0; 1g f0; 1g

kWh kW No unit No unit

with when t 1. The CHP unit burns natural gas and generates both heat and electricity. The overall conversion efciency hchp;oim as well as the heat-to-power ratio mchp;htp can be approximated by linear functions of the CHPs electrical power output. In (4) and (5), we denote
;eim ;eim these linear functions as hchp;oim pchp and mchp;htp pchp t t respectively. The right-hand side of Eqs. (4) and (5) are consequently nonlinear functions of the CHPs electrical power output and we approximate these terms by four-piece linearizations according to Ref. [14] (see also [12, pp. 12e13, 56e59]). Eqs. (6) and (7) model the maximum ramp rate of the CHPs electrical output. Technical constraints on output modulation are modeled in Eq. (8) and CHP cold starts are tracked in Eq. (9).

Subscripts and superscripts, respectively, specify parameters and decision variables. Parameters and decision variables with subscript t describe a state at the end of the time interval t ; t 1 for all t T . Other subscripts used in this study are:

ini max min

Initial value, i.e. the value at the beginning of t 1 Maximum value Minimum value

pt
Superscripts are separated by a comma. Superscripts are structured such that the rst position species the modeled unit. The residential microgrid is composed of fourteen distinguishable submodels as listed in column Modeled Units in Table 1. Superscripts at the second and the following positions provide information about features and energy ows, in particular energy import and energy export (see column Features & Flows in Table 1). As indicated in Fig. 1, imports refer to a ow from an

chp;eim chp;tim

pt

chp;tim

pt

chp;gex chp;oim

pt

chp;eim

(4) (5)

pt

pt

chp;eim chp;htp

pt

chp;eim

chp;eim chp;eim chp;eim DPmax ! pt pt 1 chp;eim chp;eim chp;eim DPmax ! pt 1 pt

(6) (7)

324

P.O. Kriett, M. Salani / Energy 42 (2012) 321e330

ut

chp chp;eim Pmin

pt

chp;eim

ut

chp chp;eim Pmax

(8) (9)

vchp ! uchp uchp t t t 1

;eim chp;eim chp with pchp Pini and uchp t 1 t 1 Uini when t 1. The electrical energy storage model is based on ywheel energy storage technology. The charging rate (10), the discharging rate (10), and the state of charge (11) are bounded according to technical specications. The energy conservation constraint (12) considers conversion efciency and the self discharging rate. Since standstill is not a problem for ywheels, (13) enforces self-discharging only when the state of charge is non-zero.

8 evh Et evh;rel if t t evh;rel > > h i > < evh Emin if t t evh;rel 1; t evh;rel 2; .; t evh;due 2 evh;lb Et > evh evh;due 1 > > : Emax if t t 0 otherwise

(19)

The state of charge of the electric vehicles battery is bounded by evh;ub evh;lb and Et (20). Charging power and discharging power are Et evh;ub eevh bounded in (21). Ktevh Et t 1 in energy conservation on earliest start dates and constraint (22) adjusts the level of eevh t after deadlines.
evh;lb Et

ees;r Pmin ees Emin

;r pees t

ees;r Pmax ; ees Emax

eevh t
;r pevh t

evh;ub Et evh;r Pmax ;

(20) with r feex; eimg


;eim pevh t

with r feex; eimg

(10)
evh;r Pmin

(21)

eees t

(11)
;eim pees t ees h ;eim

!
;sdc ees;sdc Dt uees P t

eees eees t t 1

;eex ees;eex h pees t

eevh eevh t t 1 Ktevh 

evh;eex pt *hevh;eex evh;ub Et

(12)
;sdc uees ! t

eevh t 1

hevh;eim

(22)

Dt

eees t ees Emax

(13)

ees with eees t 1 Eini when t 1. The thermal energy storage is a strongly simplied model of a water tank with thermal stratication. Charging rate, discharging rate, and the state of charge are bounded in (14) and (15). The energy conservation constraint (16) is different from (12) as a result of the assumption that the minimum temperature of the storage medium is sufciently high such that self-discharging never stops.

evsh with eevh t 1 Eini when t 1. Dishwashing cycles are modeled as non-preemptive load jobs between release date and due date. The electrical energy requirement and the power requirement of a dishwashing cycle are given dsw and P dsw;eex respectively. K dsw models earliest by parameters Emax t start dates according to case (23).

 Ktdsw

1 0

if t t dsw;due otherwise

(23)

tes;r Pmin tes Emin

;r ptes t

tes;r Pmax ; tes Emax

with r ftex; timg

(14) (15)

etes t etes t 1

dsw;ub Et

h i dsw if t t dsw;rel ; t dsw;rel 1; .; t dsw;due 1 Emax 0 otherwise

(24)

etes t

;tex tes;tex h ptes t

;tim ptes t

!
tes;sdc

P tes;tim

Dt

(16)

dsw;lb Et

dsw if t t dsw;due 1 Emax 0 otherwise

(25)

tes with etes t 1 Eini when t 1. Electric vehicle charging is modeled as a preemptive and reversible load job between release date and due date. Spontaneous discharging is feasible during connection to the microgrid. evh;ub evh;lb , and Et . Ktevh indiCharging jobs are modeled by Ktevh, Et cates earliest start dates and deadlines according to case (17). Cases evh;ub and (18) and (19) explain the assignment of parameters Et evh;lb evh;ub evh;lb . Et and Et are set to: Et

The feasible time window and the electrical energy requirement of


dsw;ub dsw;lb and Et according to a dishwashing job are modeled by Et cases (24) and (25) and enforced by constraint (26). Eq. (27) ensures energy conservation between time steps and resets the state of to zero after the deadline. Ldsw models the charge variable edsw t dsw Ldsw P dsw;eex Dt . Note that (28) allows cycle length such that Emax at most one dishwashing job within the prediction horizon T P > 1 would violate (26). because sT vdsw s

 the initial state of charge at the earliest start date,  the technical capacity constraints between earliest start and deadline,  the required state of charge by the deadline, and  zero at all other times.

dsw;lb Et

edsw t

dsw;ub Et

(26) (27) (28)

dsw;eex Dt Ktdsw edsw edsw edsw t t 1 pt t 1

Ktevh

8 <1 1 : 0

X if t t evh;rel if t t evh;due otherwise (17) o n sT : s ! max 1; t Ldsw 1 o n ^s min t ; n Ldsw 1

;eex vdsw P dsw;eex c t T pdsw s t

8 evh evh;rel > < Et evh;relif t ht i evh;ub evh if t t evh;rel 1; t evh;rel 2; .; t evh;due 1 Et Emax > : 0 otherwise

(18)

dsw with edsw t 1 Eini when t 1. We assume that cooled goods inside a refrigerator have sufciently high thermal inertia. Therefore, postponing or preponing the cooling process with respect to the original cooling schedule

P.O. Kriett, M. Salani / Energy 42 (2012) 321e330

325

has a delayed impact on the temperature of cooled goods. By postponing and preponing cooling, electrical energy consumption can be postponed and preponed. In other words, thermal inertia can be used to buffer electricity consumption of the refrigerator. Capacity bounds of the electrical energy buffer are modeled in (29). The left hand side of (30) models the refrigerators original operating schedule in terms of energy consumption. On the right hand denes the actual operating side of (30), decision variable ufri t schedule. The actual energy consumption is balanced with buffer ;buf ;prp fri;buf ;pop ; et . variables efri t

comfort due to temperature uctuations are modeled by buffer usage costs C fri;buf and C hhw;buf respectively. The boiler and the CHP unit suffer from thermal stress due to repeated power on and off. Depreciation due to thermal stress is modeled by cold start costs C boi;csc and C chp;csc . We do not consider acquisition cost, maintenance cost, or subsidized feed-in tariffs.

min

Xh
t T

  mgd ;eim ;eex ngs ngs;gim Ct pmgd Ct pmgd pt t t (37)

;buf ;r efri t

fri;buf Emax ;

with r fpop; prpg

(29) (30) (31)

fri fri;eex fri;eex Dt at1 at ufri Dt Ut P t P ;eex fri;eex ufri pfri t P t


fri;buf ;prp fri;buf ;pop fri;buf ;pop fri;buf ;pop

   ;buf ;pop ;buf ;prp ;eim Dt C fri;buf efri C evh;eim pevh efri t t t   hhw;buf ;pop hhw;buf ;prp C boi;csc vboi C hhw;buf et et t i chp;csc chp C vt

4. Case description We apply the proposed model to a hypothetical single family house located in the south of Switzerland. We briey present the corresponding data and data sources. A detailed description of the data and its preparation can be found in Ref. [12]. Constant parameter values are listed in the appendix. The modeled system is a residential microgrid. The state of the system either changes continuously (e.g. solar panel output) or in discrete points in time (e.g. operating mode of the refrigerators compressor) as a result of exogenous (e.g. solar irradiation) and endogenous (e.g. user behavior) events. In this study, events are modeled as parameter time series. Data points of parameter time series model average values over 15 min time intervals (Dt 15 min). Table 2 shows that parameter time series are either based on historical records or generated based on assumptions. We emphasize that all technical assumptions are based on commercially available state-of-the-art units preferably designed for domestic applications. Most time series data were recorded in the south of Switzerland over the course of a year between May 1, 2006 and April 30, 2007. For data collected from a different time period or location, we make appropriate adjustments. We consider three electric rate scenarios. The Switzerland electric rate scenario (CH electric rate) is a two-tiered day/night rate as it is offered at the time of writing by the Swiss utility company Sankt Galler Stadtwerke:  0.1628 CHF/kWh on Monday through Friday, 7 am until 8 pm, and on Saturday from 7 am until 1 pm;  0.0748 CHF/kWh at all other times. Market-based scenarios are based on hourly spot prices from West Denmark (WDK electric rate) and Sweden (SE electric rate). We scale these spot prices such that the annual average electric rate is equal to the annual average electric rate of the Switzerland scenario.
Table 2 Parameter time series.

with at : et et , et 1 Eini and ;buf ;prp fri;buf ;prp Eini when t 1. efri t 1 The thermal inertia of the living space makes it a thermal energy buffer. Heat and consequently hot water requirements can be preponed or postponed. Capacity bounds are modeled in (32). While the the refrigerators compressor is operated in binary hhw;tex Dt can operating mode, the heat and hot water requirement Pt be met partially as modeled in (33)

;buf ;r ehhw t

hhw;buf Emax ;

with r fpop; prpg

(32) (33)

hhw;buf ;pop Eini

hhw;tex hhw;tex Dt bt1 bt pt Dt Pt


hhw;buf ;prp hhw;buf ;pop hhw;buf ;pop with bt : et et , et 1 ;buf ;prp hhw;buf ;prp E when t 1. ehhv t 1 ini

and

The electrical solar panel output and the thermal solar panel esp;eim tsp;tim and Pt respectively. output are modeled by parameters Pt Natural gas imports from the utility company are modeled by ;gim . The bidirectional exchange of electricity decision variable pngs t mgd;eim and with the main grid is modeled by decision variables pt ;eex . The aggregated electric load is modeled by parameter pmgd t ael;eex ;tex . The thermal dump decision variable ptdu is necessary Pt t because excess heat cannot be sold to some type of upper grid. In every time interval, the generation of electrical energy, the generation of thermal energy, as well as the provision of natural gas must meet the corresponding requirements. Since Dt is a constant, it is sufcient to impose balance equations on electrical power (34), thermal power (35), and the natural gas ow rate (36).
;eim ;eex esp;eim ;eim ;eim ;eex 0 pmgd pmgd Pt pchp pees pees t t t t t ;eim evh;eex ;eex ;eex ael;eex pevh pt pdsw pfri Pt t t t

34

0 Pt

tsp;tim

boi;tim pt pt

chp;tim

tes;tim ;tex ;tex ;tex pt ptes phhw ptdu t t t

(35) 0
ngs;gim pt

boi;gex pt

chp;gex pt

(36)

Type Cost Cost Generation Generation Requirement Requirement Requirement Requirement Requirement

Name Electric rate Natural gas rate Electrical solar panel Thermal solar panel Aggregated electric load Dishwasher Electric vehicle Heat and hot water Refrigerator

Basis Historical data and assumptions Historical data Historical data Historical data & assumptions Historical data Assumptions Historical data and assumptions Historical data Assumptions

The objective function (37) sums the microgrids operating costs over the prediction horizon. Operating costs include electricity mgd ngs and Ct . costs and natural gas costs according to prices Ct Selling electricity back to the main grid creates negative costs, i.e. revenues. The depreciation rate C evh;eim of the electric vehicles battery can be calculated based on acquisition cost and the expected number of the battery charging cycles [12, pp. 62e65]. Degradation of cooled goods inside the refrigerator as well as loss of

326

P.O. Kriett, M. Salani / Energy 42 (2012) 321e330 Table 3 Total generation and total requirement. Generation Electrical solar panel Thermal solar panel kWh/yr 137.9 2462.0 Requirement Aggregated electric load Dishwasher Electric vehicle Heat and hot water Refrigerator kWh/yr 1106.4 382.2 1616.1 22259.4 258.6

The natural gas rate is set to 0.064 CHF/kWh. This is based on the rate offered by Sankt Galler Stadtwerke to its residential customers. Electric rates as well as the natural gas rate include fees and taxes. The output of the electrical solar panel is modeled based on a 0.9 square meter 125 Watt photovoltaic module [13] located in Lugano. Between May 1, 2006 and April 30, 2007, a maximum power point tracker has recorded the average output in 15 minute time intervals. We scale the records with a factor of 0.9 assuming that the modules output can be converted and transformed at a loss of 10 percent to meet the voltage requirements of the microgrid. According to manufacturers, this is a conservative assumption [24]. The output of the thermal solar panel is computed as a function of collector temperature, ambient air temperature, and global solar irradiance [28]. Technical assumptions are based on a 4.65 square meter at-plate collector [31]. Meteorological data were recorded at every quarter of an hour between May 1, 2006 and April 30, 2007 in Lugano. The collectors return and supply temperatures are assumed to be 15 and 45 degrees Celsius respectively. The heat transfer medium is water and through ow is assumed to be innitely variable. In case weather conditions are insufcient to allow for the heating of at least one liter of water per minute up by 30 degrees Celsius, the collectors thermal output is set to zero. The aggregated electric load is modeled on the total load of a single family household in Lugano, Switzerland. Provided data reveal the average aggregated power requirement between 2009 and 2010 in every quarter of an hour. We adjust the data for the weather conditions between May 1, 2006 and April 30, 2007. The adjustment considers daily average ambient air temperature, daily average global irradiance, as well as the weekday. The length, the energy requirement, and power requirement of a dishwashing cycle are determined based on a standard cycle of a state-of-the-art dishwasher [11]. We assume one dishwashing job per day. In 80 percent of the days, the earliest start of a dishwashing job is uniformly distributed between 7 pm and 11 pm. Otherwise, it is uniformly distributed between 1:30 pm and 2:30 pm. The deadline is either set to 6 am on the following day for evening jobs or to 6 pm for the afternoon jobs. The energy requirement of electric vehicle charging jobs is modeled based on records from a eld test near Lugano [23]. The monitored device, a Peugeot 106 Electric, was used by a family in their everyday life. The vehicle is powered by twenty Saft NiCd STM5-100 MRE monoblocks [4]. Data records reveal the charged energy on a daily basis but not the time of charging. We assume that the earliest start of charging jobs is uniformly distributed between 5 pm and 10 pm. The deadline is set to 7 am on the following day. The heat and hot water requirement is modeled based on the total heat requirement between 2008 and 2009 of a Minenergiecertied low-energy house with a heated area of 474 square meters in the north of Switzerland [25]. The data reveal the average fuel consumption per hour of both heat and hot water supply. We adjust the data for weather conditions in Lugano. The adjustments consider ambient air temperature in winter time and the factor weekday in summer time. The refrigerators original cooling schedule turns on the compressor for 15 minutes at the beginning of every hour. Further, we assume a 10 percent chance that cooling is scheduled at any other quarter of the hour. The energy requirement is given by the technical specications of a state-of-the-art refrigerator [10]. Table 3 summarizes the total energy supply from solar panels and the total energy requirements over a year for the studied case. The parameter settings of the boiler and the CHP unit are based on a conventional burner [30] and the ecopower MicroCHP unit from PowerPlus Technologies GmbH [26] respectively. Model parameters of the electrical energy storage and the thermal energy

storage are based on a ywheel energy storage [19, 20] and a 750 liter water cylinder with integral hot water heating [29]. 5. Results The MILP model presented in Section 3 identies the minimum cost operating schedule for a well-equipped residential microgrid. We refer to this model as model M1. In order to reveal the value of minimum cost control in a residential microgrid, we propose two benchmark models M2 and M3. Benchmark model M2 reveals the value of demand side management and optimal storage control. In contrast to model M1, in model M2  electrical and thermal energy storage devices do not exist;  dishwashing jobs are started on the earliest start date;  electric vehicle charging jobs are started on the earliest start date and are not interruptible nor reversible;  heat and hot water requirement and the refrigerators power requirement are met according to original schedules, i.e. no buffering. Benchmark model M3 reveals the effect of operating the CHP unit and the boiler according to electricity price signals. Model M3 is a modication of model M2 such that the CHP unit as well as the boiler are operated in heat-led mode. In heat-led mode,

8 hhw;tex > < 4:25; < boiler if 0:00 pt hhw;tex heat is provided by the CHP unit if 4:25 pt < 12:20; > : ;tex boiler and CHP unit if 12:20 < phhw : t
4.25 kW and 12.20 kW are the minimum and maximum thermal ;tex , the boiler and the output of the CHP unit. In case 12:20 < phhw t CHP unit are operated at minimum operating cost. Table 4 presents a compact comparison of models M1, M2, and M3. To solve the unit commitment and economic dispatch problem in the case described in Section 4, we embed the MILP models M1, M2, and M3 in a model predictive control (MPC) scheme. The idea behind MPC is an iterative optimization over nite overlapping prediction horizons. We set the length of the prediction horizon and the length of the control horizon arbitrarily to 90 time steps
Table 4 Comparison of models M1, M2, and M3. M1 Bidirectional connection to the main grid Natural gas supply Minimum cost operation of the Boiler Minimum cost operation of the CHP Electrical solar panel Thermal solar panel Electrical energy storage Thermal energy storage Aggregated electric load Scheduling dishwashing cycle Scheduling electric vehicle charging Buffering heat and hot water requirement Buffering refrigerator power requirement Y Y Y Y Y Y Y Y Y Y Y Y Y M2 Y Y Y Y Y Y N N Y N N N N M3 Y Y N N Y Y N N Y N N N N

P.O. Kriett, M. Salani / Energy 42 (2012) 321e330

327

(T 90) and 30 time steps respectively. In other words, 7.5-hour control sequences are iteratively computed based on 22.5-hour deterministic forecasts since Dt 15 min. Data handling is translated into a Java program. We use the Gurobi Optimizer 3.0 MILP solver. All computations are carried out on a Pentium D 3.0 GHz workstation with 2 GB RAM. The average solver runtime per iteration is longest for model M1 in the West Denmark electric rate scenario. Model 1 is the most complex model and the West Denmark electric rate is the most volatile price scenario. The average solver runtime per iteration is 28.5 s at an average MIP optimality gap of 0.03 percent. In order to solve the problem over a modeled period of one year, the MPC scheme goes through 1168 iterations. 5.1. Costs Table 5 presents the annual total operating cost produced by the studied residential microgrid. Negative costs model income. Revenues from electricity sales to the main grid and expenses from electricity purchases from the main grid sum up to the electricity cost. Other operating cost include cold start costs, buffer costs, and electric vehicle discharge cost. Please note that the annual average electric rate is the same in all three electric rate scenarios. The lowest annual operating cost are produced by model M1. Depending on the electric rate scenario, the savings with respect to model M2 range between 3.1 and 6.2 percent. These savings are a result of optimal storage control and optimal demand side management. With respect to model M3, model M1 saves between 4.7 and 7.6 percent in annual operating cost. Here the savings are even higher because benchmark model M3 operates the CHP unit as well as the boiler in heat-led mode which causes a jump in cold start costs. The main driver of savings appears to be the efcient generation and sale of electricity. Savings due to minimum cost control of generators, storages, and load jobs are highest in the volatile West Denmark electric rate scenario. However, the total operating cost per year is lowest in the Switzerland day/night rate scenario. Demand side management modeled in M1 affects dishwashing jobs, electric vehicle charging jobs, and the cooling

schedule of the refrigerator. The basic idea of demand side management is to shift loads to time periods of lower electricity prices subject to various constraints. Keep in mind that the total energy requirement of loads remains unchanged. Table 6 presents the savings in annual electricity cost of selected units as a result of demand side management in model M1. In benchmark model M2, load jobs are met on the earliest start date and cannot be interrupted nor reversed. Savings range between 0.2 and 30.4 percent depending on the type of load and the electric rate scenario. Relative savings appear to rise when there is greater freedom to schedule loads and when there are stronger uctuation of the electric rate. 5.2. Utilization Table 7 provides insights into the annual energy turnover of selected units. Positive numbers refer to a ow from a unit into the cables and pipes of the microgrid while negative numbers represent ow in the opposite direction. The total energy consumption in model M1 is a priori higher than the total energy consumption in M2 and M3. This is because self-discharge losses as well as efciency losses of storages do not occur in M2 nor M3. In heat-led mode (M3), heat generation is shifted from the boiler to the CHP unit. This yields much greater electricity sales to the main grid compared to M1. However, this is not reected by higher earnings (see Table 5). Thermal dump in heat-led mode is caused by heat output of the thermal solar panel at times when no heat is required. Table 8 presents the annual utilization of energy storages and energy buffers. Overall electrical energy storage is used the most in the West Denmark electric rate scenario. This indicates that electrical energy storage is best used for short term trading with electrical energy. The usage level of the thermal energy storage coincides with the output level of the CHP unit (see Table 7). This indicates that the CHPs thermal output is stored for later consumption. The electric vehicles battery is not discharged, i.e. is not used as a temporary electrical energy storage by any model in any electric rate scenario. The depreciation rate outweighs potential savings. Buffer utilization is measured in kWh/(Dt yr), which refers to the annual amount of energy that has been buffered over a time interval t ; t 1. The buffer feature of both the refrigerator and the heat requirement are not used much compared to their annual energy requirements (see Table 3).

Table 5 Annual operating cost in models M1, M2, and M3. M1 CHF/yr CH electric rate Electricity cost Electricity revenues Electricity expenses Natural gas cost Other operating cost Total operating cost SE electric rate Electricity cost Electricity revenues Electricity expenses Natural gas cost Other operating cost Total operating cost WDK electric rate Electricity cost Electricity revenues Electricity expenses Natural gas cost Other operating cost Total operating cost 465.1 712.3 247.1 2086.9 73.7 1695.4 207.3 517.4 310.1 1976.5 61.5 1830.7 404.7 708.9 304.2 2082.2 67.5 1745.0 w.r.t. M2 % 122.8 47.6 9.7 8.9 36.9 3.7 168.9 32.1 1.4 3.3 16.8 3.1 150.1 52.1 0.0 6.0 16.3 6.2 w.r.t. M3 % 20.8 11.8 1.9 1.8 34.6 4.7 11.7 2.8 4.3 3.6 50.4 5.6 40.3 22.3 4.5 1.0 41.4 7.6 M2 CHF/yr 208.8 482.4 273.6 1916.0 53.9 1761.1 77.1 391.6 314.5 1913.6 52.6 1889.1 161.8 466.1 304.3 1965.0 58.1 1861.2 M3 CHF/yr 385.0 636.9 251.9 2050.6 112.8 1778.4 234.8 532.0 297.3 2049.4 123.9 1938.5 288.4 579.6 291.2 2062.0 115.3 1888.9

Table 6 Annual electricity cost of selected loads in models M1 and M2. M1 CHF/yr CH electric rate Dishwasher Electric vehicle Refrigerator Sum SE electric rate Dishwasher Electric vehicle Refrigerator Sum WDK electric rate Dishwasher Electric vehicle Refrigerator Sum 17.3 120.9 28.3 166.5 19.1 163.3 29.0 211.3 16.6 129.1 28.7 174.4 w.r.t. M2 % 2.9 18.2 2.3 14.4 11.3 14.4 0.2 12.4 23.0 30.4 1.0 26.1 M2 CHF/yr 17.9 147.8 29.0 194.6 21.5 190.7 29.0 241.2 21.6 185.5 29.0 236.1

328

P.O. Kriett, M. Salani / Energy 42 (2012) 321e330 Table 9 Sensitivity analysis for parameters Cboi,csc, Cchp,csc, Chhw,buf and Cfri,buf. M3 kWh/yr 2384.4 5609.8 137.9 3363.3 17411.2 4031.9 2462.0 1645.8 22259.4 2370.6 5596.0 137.9 3363.3 17363.9 4077.9 2462.0 1644.4 22259.4 2563.4 5788.9 137.9 3363.3 17905.4 3537.8 2462.0 1645.8 22259.4 Cold start cost boiler Number of cold starts Total operating cost Cold start cost CHP unit Number of cold starts Total operating cost Buffer cost heat requirement Buffered energy Total operating cost Buffer cost refrigerator Buffered energy Total operating cost Variation and effects 80.0% 100.0% 1.0% 80.0% 20.2% 4.1% 80.0% 1,564.4% 0.2% 80.0% 3,100.5% 1.7% Default values 0.05 13.0 105.5 0.05 94.0 105.5 0.05 0.6 105.5 0.05 6.1 105.5 CHF/cold start nb/mo CHF/mo CHF/cold start nb/mo CHF/mo CHF/(kWh Dt) kWh/(Dt mo) CHF/mo CHF/(kWh Dt) kWh/(Dt mo) CHF/mo Variation and effects 100.0% 61.5% 0.5% 100.0% 18.1% 4.0% 100.0% 100.0% 0.0% 100.0% 98.5% 0.2%

Table 7 Annual energy turnover of selected units in models M1, M2, and M3. M1 kWh/yr CH electric rate El. transmission PCC El. CHP output El. solar panel output Sum Th. CHP output Th. boiler output Th. solar panel output Th. dump Sum SE electric rate El. transmission PCC El. CHP output El. solar panel output Sum Th. CHP output Th. boiler output Th. solar panel output Th. dump Sum WDK electric rate El. transmission PCC El. CHP output El. solar panel output Sum Th. CHP output Th. boiler output Th. solar panel output Th. dump Sum 1876.1 5121.1 137.9 3382.9 15631.8 6840.2 2462.0 0.0 24934.0 896.5 4150.1 137.9 3391.5 12843.4 9038.3 2462.0 0.8 24343.0 1649.8 4952.7 137.9 3440.7 15179.3 7372.8 2462.0 6.0 25008.0 w.r.t. M2 % 519.6 45.1 0.0 0.6 41.6 35.5 0.0 100.0 12.0 293.6 20.2 0.0 0.8 18.0 16.1 0.0 100.0 9.4 109.4 23.4 0.0 2.3 20.8 20.3 0.0 99.7 12.3 w.r.t. M3 % 21.3 8.7 0.0 0.6 10.2 69.6 0.0 100.0 12.0 62.2 25.8 0.0 0.8 26.0 121.6 0.0 100.0 9.4 35.6 14.4 0.0 2.3 15.2 108.4 0.0 99.6 12.3 M2 kWh/yr 302.8 3528.2 137.9 3363.3 11039.3 10602.1 2462.0 1844.1 22259.4 227.8 3453.2 137.9 3363.3 10880.4 10775.1 2462.0 1858.2 22259.4 788.0 4013.4 137.9 3363.3 12560.6 9252.5 2462.0 2015.7 22259.4

This result is conrmed by sensitivity analyses in other electric rate scenarios [12, pp. 78e79]. 5.4. Solution quality of the MPC scheme We choose the MPC scheme as a solution method for two reasons. First, the UCP over one year is too large to be solved in a single MILP problem, and second, embedded systems in the eld would probably solve the UCP in a similar way [2]. The conceptual drawback of MPC-based optimization is that optimality cannot be guaranteed [1]. In order to assess solution quality, we compare a MPC-based solution with the optimal solution for the same problem. For complexity reasons, this analysis is restricted to the UCP over one month (May 10, 2006 until June 10, 2006). The relative MIP optimality gap of the optimal solution in the three electric rate scenarios turns out to be at most 1.0 percent. A comparison reveals that the MPC-based monthly operating cost deviates from the minimum monthly operating cost by at most 1.1 percent. Therefore, we assume that the implemented MPC scheme produces solutions close to the optimum. 6. Conclusion We consider a generic mixed integer linear programming model to nd the minimum cost operating schedule of both electrical and thermal supply and demand in a residential microgrid. We quantify cost reductions that result from minimum cost control by comparison with benchmark models on the basis of a case study. We discuss the the utilization levels of selected units to provide further insight into the effects of minimum cost control on residential microgrid operation. We also conrm the optimality of the presented results as well as its robustness to changes in estimated parameters. Scheduling home appliances and electric vehicle charging jobs subject to technical constraints and subject to user preferences reduces annual electricity costs of these loads by up to 30.4 percent. Savings increase when the electric rate shows higher volatility and when load scheduling is restricted by fewer constraints. The main driver of savings is the efcient generation and sale of electricity. This explains the effect of the electric rate scenario. Annual total operating cost of the residential microgrid is reduced by 4.7 percent up to 7.6 percent by the minimum cost operating schedule compared to operation in heat-led mode with no storages or demand side management. Optimal storage control and optimal demand side management alone reduces the annual total operating cost by 3.1 percent up to 6.2 percent. The case study conrms that the minimum cost operation of electrical and thermal supply and demand in a residential microgrid can produce remarkable relative savings. Nominal savings, however, seem to be rather small compared to the expected

5.3. Sensitivity analysis At the time of writing, existing literature does not provide recommendations for setting the following parameters:     cold start cost of the boiler C boi;csc ; cold start cost of the CHP unit C chp;csc ; cost of using the refrigerator as an electric energy buffer C fri;buf ; cost of using the heat and hot water requirement as a thermal energy buffer C hhw;buf .

We set these parameters to 0.05. Since this setting is arbitrary, we investigate the robustness of the presented numerical results through sensitivity analysis. The sensitivity analysis is based on model M1 over one month (May 10, 2006 until June 10, 2006) in the Switzerland electric rate scenario. Table 9 shows that a variation of cost parameters can have quite an effect on the operation of the corresponding units. The effect on total operating cost per month, however, is negligible compared to the variation of parameters.
Table 8 Annual energy turnover of storages and buffers in models M1, M2, and M3. CH el. rate kWh/yr El. energy storage charge El. energy storage discharge Th. energy storage charge Th. energy storage discharge El. vehicle charge El. vehicle discharge Refrigerator buffer Heat and hot water buffer 233.4 213.9 3669.8 995.2 1616.1 0.0 kWh/(Dt yr) 7.7 106.4 SE el. rate kWh/yr 336.3 308.2 2729.6 646.0 1616.1 0.0 kWh/(Dt yr) 0.4 86.7 WDK el. rate kWh/yr 832.8 755.4 3781.1 1032.5 1616.1 0.0 kWh/(Dt yr) 4.1 104.3

P.O. Kriett, M. Salani / Energy 42 (2012) 321e330

329

investments that are necessary to allow for the automated control proposed in this study. Acknowledgment The authors would like to thank Nord Pool Spot for providing historical data from power spot markets in Scandinavia. Meteorological services were provided by MeteoSwiss, the Swiss Federal Ofce of Meteorology and Climatology. The authors would also like to thank Selwyn Piramuthu and Vincent Graziano for their help in proofreading this article. Appendix
Boiler

References
[1] Baker KR, Peterson DW. An analytic framework for evaluating rolling schedules. Management Science 1979;25(4):341e51. [2] Camacho EF, Bordons Alba C. Model predictive control. In: Advanced textbooks in control and signal processing. 2nd ed. Springer; 2004. [3] Costa L, Kariniotakis G. A stochastic dynamic programming model for optimal use of local energy resources in a market environment. In: Power Tech, 2007 IEEE Lausanne; July 2007. p. 449e54. [4] Domeniconi R, Mona N, Dustmann C-H, Manzoni R, Pulfer M. Mendrisio operating results using nicd and zebra batteries. Tech. rep., infovel Centro di competenze per la mobilit sostenibile, Apr. 2005. [5] El-Sharkh M, Rahman A, Alam M. Evolutionary programming-based methodology for economical output power from pem fuel cell for micro-grid application. Journal of Power Sources 2005;139(1e2):165e9. [6] Handschin E, Neise F, Neumann H, Schultz R. Optimal operation of dispersed generation under uncertainty using mathematical programming. International Journal of Electrical Power and Energy Systems 2006;28(9):618e26. Selection of papers from 15th power systems computation conference, 2005 e PSCC05. [7] Hawkes A, Brett D, Brandon N. Fuel cell micro-chp techno-economics: Part 1model concept and formulation. International Journal of Hydrogen Energy 2009;34(23):9545e57. [8] Hawkes A, Leach M. Cost-effective operating strategy for residential microcombined heat and power. Energy 2007;32(5):711e23. [9] Houwing M, Negenborn R, Heijnen P, De Schutter B, Hellendoorn H. Least-cost model predictive control of residential energy resources when applying mCHP. In: Proceedings of power tech 2007 (Lausanne, Switzerland); July 2007. Paper 291. [10] IKEA AG. FROSTIG BC184/18 Einbaukuehlschrank, http://www.ikea.com/ch/ de/catalog/products/00199500; May 2010. [11] IKEA AG. LAGAN DW60 Geschirrspueler, http://www.ikea.com/ch/de/catalog/ products/80152202; May 2010. [12] Kriett PO. The unit commitment and economic dispatch problem in a residential microgrid. Masters thesis, ENSIMAG, Institut polytechnique de Grenoble, Aug. 2010. [13] Kyocera Corp. KC125GH-2 high efciency polycrystalline photovoltaic module, http://www.kyocerasolar.de/index/products/download/Spanish.cps-83424-les-88756-File.cpsdownload.tmp/DB-E-KC125GH-2.pdf; December 2011. [14] Li H-L, Chang C-T, Tsai J-F. Approximately global optimization for assortment problems using piecewise linearization techniques. European Journal of Operational Research 2002;140(3):584e9. [15] Lidula N, Rajapakse A. Microgrids research: a review of experimental microgrids and test systems. Renewable and Sustainable Energy Reviews 2011; 15(1):186e202. [16] Livengood D, Larson R. The energy box: Locally automated optimal control of residential electricity usage. Service Science 2009;1(1):1e16. [17] Moghaddam AA, Sei A, Niknam T, Pahlavani MRA. Multi-objective operation management of a renewable mg (micro-grid) with back-up micro-turbine/ fuel cell/battery hybrid power source. Energy 2011;36(11):6490e507. [18] Mohamed FA, Koivo HN. System modelling and online optimal management of microgrid using mesh adaptive direct search. International Journal of Electrical Power and Energy Systems 2010;32(5):398e407. [19] Powerthru. PowerThru PowerThru PT190 DC Flywheel Brochure, http:// keyitec.com/PowerTHRU%20Brochure.pdf; December 2011. [20] Powerthru. PowerThru PT190 DC Flywheel FAQs, http://www.power-thru. com/4.html; December 2011. [21] Ristic M, Brujic D, Thoma K. Economic dispatch of distributed combined heat and power systems participating in electricity spot markets. Proceedings of the Institution of Mechanical Engineers, Part A: Journal of Power and Energy 2008;222(7):743e52. [22] Sanseverino ER, Silvestre MLD, Ippolito MG, Paola AD, Re GL. An execution, monitoring and replanning approach for optimal energy management in microgrids. Energy 2011;36(5):3429e36. [23] Schwegler U, Wegmann S. Grossversuch mit leicht-elektromobilen (lem) in mendrisio. Energieschweiz synthesebericht 2001. Swiss Federal Ofce of Energy; 2002. [24] SMA America LLC. Product Guide 2009/2010. 4031 Alvis Court, Rockin, CA 95677, USA, 2009. [25] Stettler S, Toggweiler P. Energie im gebaeude online (egon), hauptphase e jahresbericht 2008. Tech. rep., Swiss Federal Ofce of Energy, Nov. 2008. [26] Thomas B. Kleine Blockheizkraftwerke Grundlagen, Betriebsdaten, Einsatzmglichkeiten. 1st ed. Wuerzburg, Germany: Vogel-Verlag; 2007. [27] Tsikalakis A, Hatziargyriou N. Centralized control for optimizing microgrids operation. IEEE Transactions on Energy Conversion March 2008;23(1):241e8. [28] Viessmann Deutschland GmbH. Planungshandbuch Solarthermie. Viessmannstrasse 1, 35108 Allendorf (Eder), Germany, Nov. 2008. [29] Viessmann Werke GmbH & Co KG. VITOCELL 340-M Datenblatt, http://www. viessmann.de/etc/medialib/internet-global/pdf_documents/pdf-datenblaetter2009.Par.85234.File.File.tmp/DB-5414040_05-2010.pdf; December 2011. [30] Viessmann Werke GmbH & Co KG. VITOPEND 200-W Datenblatt, http://www. viessmann.de/etc/medialib/internet-global/pdf_documents/pdf-datenblaetter2009.Par.85234.File.File.tmp/DB-5414040_05-2010.pdf; December 2011.

hboi

Cboi,csc boi;tim Pmax boi;tim Pmin

85.00 0.05 24.00 10.50

% CHF/cold start kW kW

Energy conversion efciency Cold start cost Maximum thermal output Minimum thermal output Cold start cost Maximum electrical ramp rate Maximum electrical output Minimum electrical output Total electrical requirement Length Electrical requirement Charging efciency Discharging efciency Maximum state of charge Minimum state of charge Maximum charging rate Minimum charging rate Maximum discharging rate Minimum discharging rate Self discharging rate Charging efciency Discharging efciency Cost of discharging Maximum state of charge Minimum state of charge Maximum charging rate Minimum charging rate Maximum discharging rate Minimum discharging rate Buffering cost Maximum buffer capacity Buffering cost Maximum buffer capacity Electrical requirement Natural gas rate Charging efciency Discharging efciency Maximum state of charge Minimum state of charge Maximum charging rate Minimum charging rate Maximum discharging rate Minimum discharging rate Self discharging rate

Combined heat and power unit 0.05 CHF/cold start Cchp,csc chp;eim DPmax 2.22 kW/Dt
chp;eim Pmax

4.44 1.20 1.05 10.00 0.42

kW kW kWh Time steps kW % % kWh kWh kW kW kW kW kW % % CHF/kWh kWh kWh kW kW kW kW CHF/(kWh Dt) kWh CHF/(kWh Dt) kWh kW CHF/kWh % % kWh kWh kW kW kW kW kW

Pmin

chp;eim

Dishwasher dsw Emax Ldsw Pdsw,eex

hees,eex hees,eim
ees Emin

Electrical energy storage 100.00 100.00 ees 0.68 Emax 0.00 30.00 0.34 30.00 0.34 0.25 87.00 87.00 1.20 12.00 0.00 3.20 3.20 3.20 0.00
ees;eex Pmax ees;eim Pmax ees;eim Pmin Pees,sdc ees;eex Pmin

Electric vehicle

hevh,eex hevh,eim
C evh Emax
evh Emin

evh,eim

evh;eex Pmax evh;eex Pmin evh;eim Pmax evh;eim Pmin

Heat and hot water 0.05 Chhw,buf hhw;buf Emax 1.00 Refrigerator Cfri,buf fri;buf Emax Pfri,eex Natural gas Cngs 0.05 0.47 0.12 0.07

htes,tex htes,tim

Thermal energy storage 60.00 60.00 tes 34.87 Emax tes Emin 0.00 tes;tex 14.40 Pmax
tes;tim Pmax tes;tim Pmin Ptes,sdc tes;tex Pmin

0.00 14.40 0.00 0.062

330

P.O. Kriett, M. Salani / Energy 42 (2012) 321e330 Rautenstrauch C, editors. Information technologies in environmental engineering. Berlin Heidelberg: Environmental Science and Engineering, Springer; 2007. p. 119e31. [33] Wood A, Wollenberg B. Power generation, operation, and control, vol. 1. Wiley Interscience Publication, J. Wiley & Sons; 1996.

[31] Viessmann Werke GmbH & Co KG. VITOSOL 200-F Datenblatt, http://www. viessmann.de/etc/medialib/internet-global/pdf_documents/pdf-datenblaetter2009.Par.59740.File.File.tmp/DB-5368786_07-2009.pdf; December 2011. [32] Vogel U, Sonnenschein M. Optimization of adaptive consumers to a timevarying electricity supply. In: Gmez JM, Sonnenschein M, Mller M, Welsch H,

Você também pode gostar