administração pública

1

Demand Side Load Control with Smart Meters Frank D. Silva, Student Member, IEEE, and Osama Mohammed, Fellow, IEEE, Energy Systems Research Laboratory, Department of Electrical and Computer Engineering, Florida International University, Miami, FL 33174

Abstract— Utility companies will seek additional electricity generation to satisfy the needs projected by increasing demands. Additional generation means investing in new power plants that will increase costs for the companies and customers. The increase in electric demand also includes increases in peak demands. Peak demands are mostly satisfied with costly and inefficient generators. In system planning for the future, given these circumstances, demand side management (DSM) needs to be given closer consideration, since it could be used to reduce peak demands. This paper proposes a DSM system using smart meters (SM) for load control. The proposed load control would reduce peak demands while not affecting the total power consumption by customers. The load control system would target appliances that consume larger amounts of power. Depending on a priority order, established by the customer, these appliances would be switched off for a period of time during the peak in demand. Additionally, customers would be able to use the Smart Meters to communicate back with the utility companies, reporting appliance use available for load control when the previous control periods are not active. This feature, our proposal, would allow a customer to switch on an appliance and decide to signal the utility company the availability for load control. The utility company could use this information and take action when system operation may require doing so. Index Terms— Appliance manager, demand side management, home area network, smart meter, load control.

I. INTRODUCTION

S

MART meters are the foundation of the electric Smart Grid. Smart meters (SM) are used to record electric energy consumption by customers and remotely transmit the information to the utility company. Many utility companies are using SM technology to determine the electric bill of customers. About one third of homes in the United States have SM. Utility companies are currently able to use the up-to-date information provided by SM to determine customer consumption at different rate periods. SM technology is capable of two-way communication with the utility companies and is also capable of communicating within a home area network (HAN). The HAN can be used by customers to view and control their electrical consumption. According to the U.S. Energy Information Administration [1], residential and commercial electric demands are expected to slightly decrease in the next couple of years, followed by constant increase well into 2030. In order to supply all electric

978-1-4799-1303-9/13/$31.00 ©2013 IEEE

needs, generation must equal demand at all times. The demand increase, with the associated peaks, will require additional operation of less efficient generators, increasing costs for consumers. This increase in demand is also expected to require construction of new generators and transmission lines. The costs associated would be undetermined but expected to continue increasing. Labor and material costs are known to increase, but there is no control over fuel costs. In addition, government regulations and environmental controls directly affect the construction of new generators and transmission lines. Demand side management (DSM) can be used to avert increases in electric demand, need for generation, and the associated costs. DSM can shave demand peaks, requiring less use of lower efficiency generators. With proper management, DSM can be used to allocate loads from shaved demand peaks into lower generation levels. This load management would result in a more leveled demand and in turn cheaper economic dispatch of electric generation. II. RELATED WORK Previous related work contained in technical paper [2] identified individual appliances within a particular household in terms of their energy demand. Identification of individual appliances was achieved using a set of algorithms, smart meters, and smartphones. The system achieved 87% recognition rates when 8 devices were tested simultaneously. Another technical paper [3] concentrated on future residential load forecasting. The method used in this case required larger amounts of real-time data in order to improve accuracy. Technical paper [4] studied smart networking for a smart house system. The smart house system consists of appliances with smart cards installed. These smart appliances communicate with the SM to provide load consumption information. Smart houses are connected with a town server, which is able to control power, generated by renewable sources, stored for future use, or provided by the utility company. Technical paper [5] also studied Smart Homes and introduced a nonintrusive appliance load monitoring system. Load shedding using SM was investigated in technical paper [6]. The method proposed uses a SM controlled scheme to shed loads to support primary frequency. In the proposed

2 scheme, the residential loads are grouped based on their criticality and shed depending on the drop of frequency. Technical paper [7] investigated the use of SM to control residential loads during system emergencies. Direct load control through the SM in response to frequency changes was discarded based on communication delays. The document studied an alternative load control scheme that used a local frequency measurement from the SM. DSM mechanisms were studied in technical paper [8]. The document proposes using smart grid infrastructures to reduce power demand at peak hours by means of dynamic pricing strategies. The proposed system would use a wireless power meter sensor network to monitor home appliances consumption. Data provided by sensors are processed every 24 hours and used to forecast the following day consumption. Technical paper [9] also discusses DSM as a strategy to reduce peak consumption and avoid construction of new power generators. III. PROBLEM STATEMENT This paper proposes use of SM technology to conduct DSM load control. Utility companies (UC) could accomplish this residential load control using total load consumption readings from SMs. The UC would not need access to information of the customer’s internal consumption in order to achieve DSM. However, it is recognized that technical paper [2] established the technical feasibility of using algorithms to identify individual loads. The proposed DSM load control is not intended to reduce electrical consumption, but rather delay this consumption to lower cost periods. While the proposed load control process does not give the UC access to individual appliance consumption within a residence, it would still achieve DSM on individual appliances during demand peak periods. The individual appliances that will be controlled are: water heater, dishwasher, pool pump, clothes washer, and clothes dryer. The SM would continue to provide the UC with up-to-date customer consumption. In many cases the customer also has access to this up-to-date information, even though they have no access to changing SM features. The up-to-date information would be used to manage this proposed DSM load control process. During load demand peak periods, a UC aggregator would use the SM two-way external communication system to set up a desired maximum consumption level for a group of customers during certain period of time. After receiving a desired maximum consumption level signal, the SM would become a load controller for the customer. The appliances would be connected to the outlet through a disconnect device, or appliance manager (AM). The SM would use the HAN communication system to send signals to the AM of the individual appliances. The individual appliances would be enabled or disabled using the AM. In cases when the SM cannot be used as the load controller, a separate controller could serve this purpose. This separate

controller would require input signals from the SM for maximum desired consumption and control period. These signals would come from the UC aggregator through the SM. The SM would use the desired maximum consumption level supplied by the aggregator of the UC to monitor and control internal consumption. If the total load exceeds the control level, the SM would use the HAN to send a signal to the AM of one of the appliances. The SM would continue to send disconnect signals until the total load is under the desired maximum consumption level. The appliance order of priority would be established by the customer. If all participating appliances are disconnected and the load continues to exceed the control level, the SM will not disconnect any additional loads. Incentivized customers could use the HAN to set up their participation in the DSM. Through the HAN customers would establish the order of priority in which individual appliances would be disconnected. Customers could be notified in advance of DSM controls to be applied and of the current control status. The HAN system could notify the customer of any appliance disconnections and provide record of their participation in the DSM program. An additional feature the AM could have would be a switch that customers could use to indicate availability for load control when they turn on an appliance. The AM would communicate with the SM, which in turn would send this information to the UC. A UC aggregator could use this information to order disconnects of appliances when the operational needs arise. This feature would be in addition to the process previously explained. Motivated customers could take advantage of incentives offered in a program with this second feature to include other large appliances, such as air conditioning units. IV. SIMULATION AND TESTING SM control and household loads were simulated using MATLAB and Excel files. The system simulated different random residential loads measured by the SM. The SM readings were added up and controlled by a dispatch aggregator. The dispatch aggregator used cumulative SM information to establish DSM controls. In order to run this simulation, it was necessary to create the load data to be tested. An Excel spreadsheet was created with different loads containing random combinations of periods of usage for the five appliances to be considered for control. The table contains the consumption per appliance, starting time for the appliance usage, amount of time the appliance is used, and the switching hierarchy for the appliances. All of these measures were random values from within a set of parameters. In the spreadsheet, consumption is measured in current (Amperes) and time is measured in minutes. The spreadsheet information, combined with the demand curve, allowed a minute by minute simulation of the household load. The typical household load demand curve utilized was obtained from estimated values. The load curve for each of the

3

Appliance

Time start between

Duration On

Yearly Demand

Washer

10:40 am & 10:00 pm 30 to 60 Mins

125 kWh

Dryer

12:20 pm & 11:40 pm 40 to 90 Mins

1000 kWh

Heater

6:00 am & 5:00 pm

6 to 24 Hours

2300 kWh

Pool Pump

6:00 am & 6:00 pm

3 to 6 Hours

800 kWh

Dishwasher

5:00 pm & 10:00 pm

40 to 60 Mins

130 kWh

Fig. 2 Appliance specifications to be used in load simulation

The residential loads were combined, for a total of 10,000 residential customer loads simulated. Of these, 5,000 were not affected by the load control this paper proposes. The other 5,000 residential customer loads were divided into 25 control groups with different amounts of customers each. The 23 control periods lasted 30 minutes and started at 3:20 pm. The control over the last group ended at 6:10 pm. The desired maximum consumption was set for this simulation at 30.5 A, or 3.66 kW. The MATLAB simulation program first considered the individual household load composition during a full day. It determined whether any of the appliances was operating and summed up the consumption values for each minute during the 24 hour period. The resulting load consumption followed the typical demand curve. Figure 3 shows the demand for the total 10,000 residential customers without DSM load control. The total Demand for the 24-hour period was 616 MWh. The Peak Demand was 37.94 MW. Once the controlled operation was established, the program considered the desired maximum consumption, as calculated by the system aggregator. The program simulated the reaction of the individual residential SM to the beginning of the load

40

35

Consumption in MW

100 individual loads represented slightly varied (± 10%) from the average. In order to follow the demand curve, the sum of active appliances and other loads were also considered. These additional loads were not considered by individual appliance. The typical appliance consumption was obtained from the U.S. Department of Energy website [10]. The yearly values were divided by 365 to obtain each appliance’s daily demand. The power demand was converted to current to simplify calculations. The power demand of each appliance was divided by the daily usage in order to determine the average current consumption of each appliance. The final value assigned was randomly selected between 90% and 100% of this value. Figure 2 is a table that shows the values utilized to simulate individual appliance loads. It is important to note that all appliance usage, which involves start time, duration, and current drawn, were assigned randomly using established maximum and minimum parameters. In addition, the appliance order of priority, to be established by the customer, was also randomly simulated. Individual residential loads were obtained using random combinations of the appliance loads and adding a residual load to complete the typical residential load demands. The table shows the possible maximum and minimum values that could be randomly assumed.

30

25

20

15

10

0

500

1000

1500

Time in hours

Fig. 3 Total load Demand without DSM load control for a 24-hour period

control. If the desired maximum consumption was higher than the SM reading, the SM would send a signal to the first priority appliance AM. After this, the SM would verify the resulting load level and again compare it to the desired level. If the load level was lower than the desired maximum consumption, the SM would take no further action, except continue to monitor the load during the control period. If after sending a signal to disconnect the first appliance, the SM load reading continues to be higher than the desired maximum consumption, then the SM would send a disconnect signal to the second appliance AM. The process would continue sending disconnect signals to the rest of the appliances until the SM reading is lower than the desired maximum consumption. If after all appliances have been disconnected, and the SM load reading continues to be higher, the SM will take no further action until the control period has expired. After the control period expires the SM must reestablish service to the appliances that were shut down during the control period. The simulation program dealt with three possible conditions regarding the five appliances involved in this test. First condition would be the appliance started and finished normally and was not affected by the SM load control. In this case the simulation would take no action. The other two conditions would require the program to simulate the real life situation. The real life situation would be that the appliance be used for the necessary period of time before or after the control period. The second condition would involve the appliance being in operation when the control period begins. In this case, the program simulated the appliance completed the duration of its operation period after the control period finished. It is important to note that this action has the effect of delaying consumption, which would only shift the demand curve for the period the control action is in effect. For this reason different groups of loads must be controlled at different control periods in order to flatten the demand curve and move the load consumption into non-peak periods.

4

Start

40

35

Consumption in MW

The third condition would be when the appliance begins operation during the control period and is disconnected by the process. In this case the program simulates the appliance starting after the control period and operating during the remaining normal operation duration. In the real life scenario, it is possible that residential customers could voluntarily move the time period of operation of some of their appliances if early notice of upcoming control periods is provided.

30

25

20

15

T=T+1 T = 0 to 1440 Minutes

10 Wdur over?

No

Yes

No

Ddur over? Yes

Wdur

Hdur over?

No

Yes Hdur

Ddur

Con

No

Pdur over?

Kdur over?

Yes Pdur

No Control on

Yes Kdur

No Control off

Yes

Coff

Coff

Coff Smart Meter SM > Max

AC

AC

Wdur

Ddur

Hdur

No

Yes

WLC Washer

Timer for Wdur

WLC

Dryer

Timer for Ddur

DLC

Heater

Timer for Hdur

HLC

Pool Pump

Timer for Pdur

PLC

Dishwasher

Timer for Kdur

Yes

Yes

W Priority

Yes

AC

Pdur

Kdur

PLC

Yes

P Priority No

KLC KLC

Yes

K Priority No

AC

Other Appliances

38.5

H Priority No

AC

39

D Priority No

HLC

1500

period when the DSM load control was established. The graph of the DSM controlled test has lower amplitude and a chopped-off appearance. This effect is due to the control action. The difference between both graphs would be the amount of peak demand shaved.

No

DLC

1000

Fig. 5 Total load Demand with DSM load control for a 24-hour period

Con

AC

500

Yes

Con

Operator input Max Consumption

Null

0

Time in hours

End

Fig. 4 Flowchart for the Simulation Program

Figure 4 is a flowchart of the MATLAB simulation program. Switching on and off of the appliances is accomplished with relay contacts, which are represented in the chart. The combined 24-hour period energy consumption for the 10,000 residential customers was 616 MWh. The combined peak demand for the 10,000 residential loads without DSM load control was 37.94 MW, while it was 37.51MW when the proposed DSM load control was applied to half of the residential loads. The difference in peak demand was 430 kW. Figure 5 shows the demand for the total 10,000 residential customers. In this case 5,000 residential loads were not subject to DSM load control, while the other 5,000 was subjected. From the graph it can be noted that the demand shaved from the peak period was allocated in the period that follows. The period that follows the peak had a decaying demand, which would allow access to lower cost energy. Figure 6 illustrates a detail of the results of the test with and without DSM load control. The graph concentrates on the peak

Consumption in MW

No

38 37.5 37 36.5 36 920

940

960

980

1000 1020 Time in hours

1040

1060

1080

Fig. 6 Closer look comparing Peak Demand difference before and after DSM Load Control

Narrowing down the results to an individual residential load, each one reacted differently to the various control periods. Half of the 10,000 residential loads were not under DSM control and remained unchanged. The other half of the total residential loads was placed under DSM Load Control. The result of the control was shutting down appliances during 30 minutes. The amount of appliances disconnected depended on the individual residential load configuration and the priority order set by the customer. Based on the simulation and the models used for the test, the DSM load control would be an effective means to lower peak residential demand. With a sample of 10,000 residential customers, the test resulted in a reduction of 430 kW from the

5 peak demand. Multiples of these results could represent significant MW savings, lower costs for the UC and cheaper electricity for the customers. For example, for a residential customer base of 1 million, peak demand savings could be nearly 43 MW. Lowering the peak demand could not only avoid spending resources on new generators and transmission lines, but could represent avoiding use of less efficient peak generators. Generators used during peak periods are usually less efficient and more harmful to the environment. In that sense, DSM load control is friendly to the environment. The MATLAB program simulated 25 customer groups under DSM load control. A real system with many customers subscribed to the program would be separated into more groups. This would allow the aggregator to better control and flatten the peak demand. Observing Figs. 5 and 6 it is clear that better efficiency could be accomplished to lower even more the peak demand. This effect can be obtained by better spreading the effects of the load control, which can be done with more control groups. Another detail that must be pointed out is the fact that all customers had similar demand curves with different amplitudes (± 10%). In the case of our test, in order to target half of the total residential load (10,000) it was necessary to place half of the residential customer loads (5,000) in DSM load control. In reality, in order to target the top half of the residential load, less than half of the residential customers would need to be subscribed in the DSM load control program. In this case, only the largest residential loads would be included. Adding the feature that would allow customers to designate current use of appliances as loads available for DSM would be highly beneficial for the UC. Instead of recurring to load shedding, UC aggregators could automatically disconnect appliances designated by customers. Customers could let the UC know of load control availability for appliances such as air conditioning units and the other mentioned in this proposal using their smart phones. This feature would require the AM being able to communicate with the UC through the SM. One concern for implementing controls using SMs could come from customer privacy concerns. The DSM load control program allows the SM to control the appliances without letting the UC know which appliances are being switched off. The switching hierarchy for the appliances is established by the customer. An important aspect of implementing a DSM load control using SMs system is to prepare the infrastructure for future changes in residential loads. UCs and Public Service Boards have already noted that, in the near future, electric vehicles will place a high demand on their generation capabilities. The proposed system could also be used to control battery charging periods of electric vehicles.

V. CONCLUSION The proposed DSM load control system could represent peak demand savings of nearly 43 MW for a residential load base of 1 million customers. In addition, including the feature that would allow customers to designate current use of appliances as loads available for DSM would be highly beneficial for the UC. Instead of recurring to load shedding, UC aggregators could automatically disconnect appliances designated by customers. Implementing the DSM load control system would require enabling SMs with a programmable HAN. This HAN would be the channel used by the SM to execute the UC aggregator’s signals for desired maximum consumption, start load control, and end load control. The system would also require implementing AM devices that would be connected to appliances and would perform switching operations based on wireless communication with the SM through the HAN. In the future, the AM devices could be factory installed, making appliances DSM load control friendly. REFERENCES [1]

U.S. Energy Information Administration, http://www.eia.gov/oiaf/aeo/tablebrowser/#release=AEO2012&subject= 0-AEO2012&table=2-AEO2012®ion=1-0&cases=ref2012d020112c, November 8, 2012. [2] Weiss, Markus ; Helfenstein, Adrian ; Mattern, Friedemann ; Staake, Thorsten, “Leveraging smart meter data to recognize home appliances”, 2012 IEEE International Conference on Pervasive Computing and Communications, March 2012, pp.190-197. [3] Ghofrani, M. ; Hassanzadeh, M. ; Etezadi-Amoli, M. ; Fadali, M. S., “Smart meter based short-term load forecasting for residential customers”, 2011 North American Power Symposium, August 2011, pp.1-5. [4] Bilal shahid, Engr. ; Ahmed, Zubair ; Faroqi, Adnan ; Navid-urRehman, Rao M., “Implementation of smart system based on smart grid Smart Meter and smart appliances”, Iranian Conference on Smart Grids, May 2012, pp.1-4. [5] Benyoucef, Dirk ; Klein, Philipp ; Bier, Thomas, “Smart Meter with non-intrusive load monitoring for use in Smart Homes”, 2010 IEEE International Energy Conference, Dec. 2010, pp.96-101. [6] Samarakoon, K. ; Ekanayake, J., “Demand side primary frequency response support through smart meter control”, Universities Power Engineering Conference, Sept. 2009, pp.1-5. [7] Kamalanath Samarakoon, Member, IEEE, Janaka Ekanayake, Senior Member, IEEE, and Nick Jenkins, Fellow, IEEE, “Investigation of Domestic Load Control to Provide Primary Frequency Response Using Smart Meters”, IEEE transactions on Smart Grid, Vol. 3, No. 1, March 2012, pp. 282-292. [8] Barbato, A. ; Capone, A. ; Rodolfi, M. ; Tagliaferri, D., “Forecasting the usage of household appliances through power meter sensors for demand management in the smart grid”, 2011 IEEE International Conference on Smart Grid Communications, Oct. 2011, pp.404-409. [9] Majid, M.S., Fac. of Electr. Eng., Univ. Teknol. Malaysia, Skudai Rahman, H.A. ; Hassan, M.Y. ; Ooi, C.A., “Demand Side Management Using Direct Load Control for Residential”, 4th Student Conference on Research and Development, 2006, pp. 241-245. [10] U.S. Department of Energy, http://energy.gov/energysaver/articles/estimating-appliance-and-homeelectronic-energy-use, November 8, 2012.