Demand response scheduling method based on user preference and mixed integer linear programming

Abstract

The present invention is applicable to the technical field of demand response scheduling, and provides a demand response scheduling method based on user preference and mixed integer linear programming, including: performing fuzzy C-means clustering on the historical data of NSAs load and CAs load respectively; obtaining the energy cost tolerance and equipment information of residential users, calculating the user willingness to pay of each residential user to quantify the user preference, and calculating the allocated energy; based on the user preference, using mixed integer linear programming to perform demand response scheduling of residential communities to minimize the total energy cost of the residential communities. The demand response scheduling method of the present invention takes into account the energy cost tolerance and equipment information, and takes into account the user preference to minimize the total energy cost of the residential community, and has the characteristics of reducing the total energy cost and communication traffic of the community.

Description

Technical Field

[0001] The present invention belongs to the technical field of demand response scheduling, and in particular relates to a demand response scheduling method based on user preference and mixed integer linear programming. Background Art

[0002] Demand response (DR) is the abbreviation of electricity demand response. It means that when the wholesale electricity market price rises or the system reliability is threatened, the electricity user receives the direct compensation notice of induced load reduction or the signal of rising electricity price from the power supplier, and changes its inherent habitual electricity consumption mode to reduce or shift the electricity load in a certain period of time to respond to the power supply, thereby ensuring the stability of the power grid and suppressing the short-term behavior of rising electricity prices, so as to manage demand in the smart grid to match the available energy. Through DR, the smart grid can deliver electricity from the power generation end to active consumers in a controllable and intelligent way. Residential energy consumption accounts for about 30-40% of global energy consumption, so exploring DR of demand-side entities in residential communities has important theoretical and practical significance.

[0003] Due to different user preferences, the load of residential communities in smart grids is diverse and flexible. Community users can decide whether to reduce electricity consumption based on real-time electricity prices, which helps shift loads and reduce energy costs. However, different user preferences within the same residential community make it difficult to find a universal DR schedule without affecting the needs of different residential users.

[0004] There has been a lot of research progress on DR scheduling based on user preferences, and the main methods can be summarized into three categories. The first method is based on the user's energy consumption behavior; however, the user's energy consumption behavior involves a large amount of energy consumption data, which makes it more difficult to extract user preferences. The second method is based on the affordability of energy costs, which reflects the subjective factors of user preferences. The third method is based on home appliance information, which reflects the objective factors of user preferences. However, the current methods cannot adapt to different user preference requirements and cannot minimize the total energy cost of residential communities while considering user preferences. Summary of the invention

[0005] In view of the above problems, the purpose of the present invention is to provide a demand response scheduling method based on user preferences and mixed integer linear programming, aiming to solve the technical problem that existing methods cannot minimize the total energy cost of residential communities while considering user preferences.

[0006] The present invention adopts the following technical solution:

[0007] The demand response scheduling method based on user preference and mixed integer linear programming is used for demand response scheduling of residential communities, wherein a load aggregator is provided on the residential community side and a home energy management system is provided on the residential user side. The method of the present invention comprises the following steps:

[0008] Step S1, the load aggregator obtains the historical data of the load of non-schedulable devices NSAs and the load of controllable devices CAs with labels, and performs fuzzy C-means clustering on the two historical data respectively;

[0009] Step S2: The load aggregator obtains the energy cost tolerance and equipment information of residential users from each home energy management system, calculates the user willingness to pay of each residential user, and calculates the energy allocated to each residential user based on the user willingness to pay and the real-time electricity price, i.e., allocated energy;

[0010] Step S3: For each residential user, the home energy management system calculates the minimum demand of the residential user. If the allocated energy is less than the minimum demand of the current residential user, the energy cost tolerance of the residential user is increased and the allocated energy is calculated again until the allocated energy is greater than or equal to the minimum demand.

[0011] Step S4: The load aggregator uses mixed integer linear programming to perform demand response scheduling for residential communities.

[0012] The beneficial effects of the present invention are as follows: the present invention first clusters the devices according to power based on the historical data of the non-dispatchable devices and the controllable devices, and introduces the user willingness to pay to quantify the user preference by obtaining only a small amount of user information, including energy cost tolerance and device information, and combining the power and operation time of the dispatchable devices, and finally minimizes the total energy cost of the residential community based on the user preference and mixed integer linear programming. The method of the present invention is a complete residential community demand response scheduling method, which has the characteristics of reducing the total energy cost and communication traffic of the community. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic diagram of the layout of residential communities and residential users;

[0014] Figure 2 is a flow chart of a demand response scheduling method based on user preference and mixed integer linear programming provided by an embodiment of the present invention;

[0015] Figure 3 It is a comparison chart of the load assigned by UWP and the original load in a residential community during a day;

[0016] Figure 4 is the hourly load comparison diagram of the residential community before dispatch, after MILP and DR-UWP-MILP dispatch;

[0017] Figure 5 It is the UWP distribution load of residential user 1 and the hourly load comparison diagram before and after scheduling. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] In order to illustrate the technical solution of the present invention, a specific embodiment is provided below for illustration.

[0020] like Figure 1 As shown, this embodiment provides a demand response scheduling method based on user preferences and mixed integer linear programming, which is applied to demand response scheduling in residential communities. A load aggregator is provided on the residential community side, which is connected to the smart grid. A home energy management system HEMS is provided on the residential user side, and the types of electrical appliances on the residential user side include non-shiftable appliances (NSAs), controllable appliances (CAs) and shiftable appliances (SAs). Figure 2 As shown, the method of the present invention includes the following steps:.

[0021] Step S1, the load aggregator obtains the historical data of the labeled unschedulable equipment load and the controllable equipment load, and performs fuzzy C-means clustering on the two types of historical data respectively.

[0022] As mentioned above, residential users' electrical equipment is generally divided into three categories: NSAs, CAs, and SAs. NSAs are turned on 24 hours a day, cannot be turned off by DR scheduling, and work at a fixed power; for example, security systems run 24 hours a day. CAs can operate at different rated powers and can be adjusted between maximum power and minimum power; for example, air conditioners can operate at different power levels. SAs have fixed rated power and operating hours. These appliances can be switched to other times to run. Once turned on, they cannot be turned off before the end of the operation, including electric vehicles, water heaters, etc. The following Table 1 is an example of a residential community. The residential community has a load aggregator and 35 residential users. The number of electrical equipment in the residential community is shown in Table 1:

[0023] Table 1

[0024]

[0025] The specific process of this step is as follows:

[0026] S11. Classify the historical data according to the labels, where the load of non-dispatchable equipment is recorded as NSA load and the load of controllable equipment is recorded as CAs load.

[0027] Since both NSAs and CAs payloads are labeled data, the payload data is called a data object, denoted as First, load the NSAs according to the label Load with CAs separate, is the device category. Here =1 means the device is NSAs, =2 means the device is CAs, =3 indicates the device is SAs.

[0028] S12. Fuzzy C-means clustering is performed on NSAs load and CAs load respectively. The number of clusters for both sets of historical data is 3. The cluster centers obtained after clustering are , is the device category, are the different power levels of clustering, .

[0029] The following describes the clustering process using NSA load as an example.

[0030] set up is the degree of membership of the data object to the cluster center, which is:

[0031] (1)

[0032] are the different power levels of clustering, , , is an index of NSAs in residential communities.

[0033] The objective function of fuzzy C-means is defined as:

[0034] (2)

[0035] in, is the Euclidean distance from the data object to the cluster center, and m is the fuzzy factor.

[0036] Membership and cluster centers The expression is as follows:

[0037] (3)

[0038] (4)

[0039] In the formula, the initial cluster center is randomly selected, and then each data object is calculated To cluster center The initial membership degree, membership degree and cluster centers Iterative update. Where r represents the rth category.

[0040] Similarly, for CAs load, equations (1)-(4) also hold true. , let E represent the index of CAs in the residential community.

[0041] Since the number of clusters for both sets of data is 3, the clustering results include the cluster centers of the unschedulable devices. , , , the cluster centers of controllable devices are , , .

[0042] Step S2: The load aggregator obtains the energy cost tolerance and equipment information of residential users from each home energy management system, calculates the user willingness to pay of each residential user, and calculates the energy allocated to each residential user based on the user willingness to pay and the real-time electricity price, i.e., allocated energy.

[0043] User willingness to pay (UWP) is used to quantify user preferences. Different residential users have different willingness to pay and affordability for electricity. This step calculates the user willingness to pay for each residential user, and finally calculates the energy allocated to each residential user. Taking residential user i as an example, the specific process of this step is as follows:

[0044] S21. For each residential user i, obtain the energy cost tolerance of residential user i in time slot t , and the number of residential users’ two types of equipment clustered into three power levels, recorded as .

[0045] The 24 hours of a day are divided into time slots, generally one hour is a time slot, and a time slot is a time period. The energy cost tolerance is set by the home energy management system according to the characteristics of residential users. According to the clustering results of fuzzy C-means, the NSAs and CAs of residential users can be divided into three power levels. The load aggregator sends the clustering results to the home energy management system, and the home energy management system sends its own user information to the load aggregator according to the three power levels of NSAs and CAs.

[0046] User information includes the number of electrical appliances of corresponding power level provided by each residential user , and the operating power of dispatchable equipment SAs and run Time, and energy cost affordability .

[0047] It is not necessary to provide information on all electrical equipment.

[0048] S22. Calculate the user willingness to pay for residential user i :

[0049] (5)

[0050] in , , is a normalization parameter, for example, the possible values ​​are as follows: 1 =0.59, ω 2 =0.16, ω 3 =0.25, the operating power of the dispatchable equipment SAs is , the running time is .

[0051] In the above steps, clustering can be used to obtain the cluster center of each group of NSAs and the cluster center of each group of CAs, and the cluster center is power. According to the expression of formula (5), the user willingness to pay of residential user i can be calculated to evaluate the preference of residential users.

[0052] S23. Calculate the energy allocated to each residential user every hour, i.e., the allocated energy ,in It is the real-time electricity price.

[0053] The energy allocated to each residential user is calculated based on UWP and real-time electricity prices, which is recorded as allocated energy. For residential user i, its allocated energy is (6).

[0054] Step S3: For each residential user, the home energy management system calculates the minimum demand of the residential user. If the allocated energy is less than the minimum demand of the current residential user, the residential user's energy cost tolerance is increased and the allocated energy is calculated again until the allocated energy is greater than or equal to the minimum demand.

[0055] The minimum demand of residential users is set by the users and has a starting amount. If the allocated energy calculated above is less than the minimum demand of the current residential user, it means that the allocated energy of the residential user is too small, so the allocated energy needs to be increased. Specifically, the energy cost tolerance of the residential user is increased by a fixed amount each time, and the allocated energy is calculated again until the allocated energy is met. Greater than or equal to the minimum demand .

[0056] Step S4: The load aggregator uses mixed integer linear programming to perform residential community demand response scheduling.

[0057] Linear programming has constraints and optimal solutions. Under the premise of satisfying the constraints, the optimal result is obtained, generally the maximum value or the minimum value. In this embodiment, the goal is to minimize the total energy consumption of the community.

[0058] The constraints for mixed integer linear programming are:

[0059] Constraints for the nth NSAs of residential user i in time slot t:

[0060] (7)

[0061] in is the fixed load of user i’s nth NSA.

[0062] Constraints for the e-th CAs of residential user i in time slot t:

[0063] (8)

[0064] in is the minimum load of the e-th CAs of residential user i, is the maximum load of the e-th CAs of residential user i.

[0065] Assuming that a 24-hour day is divided into 24 time slots, the fixed consumption pattern vector of the s-th SAs of residential user i is :

[0066] (9)

[0067] Scheduling of CAs and a circular shift in fixed consumption mode same, is a column vector, is the initial open state, the elements of the column vector The corresponding cyclic shift The power of each time slot. Since the equipment has a specific power in a certain operating period, such as a washing machine with a rinsing power of 0.5kW, a dehydration power of 1.0kW, and a drying power of 1.2kW, for the convenience of representation, it is assumed that rinsing, dehydration, and drying are all in one time slot, expressed as:

[0068]

[0069] The purpose of scheduling equipment is to change the start and end time of equipment operation, while the power of the operating time period remains unchanged. It can represent the power of all possible time periods after all scheduling, and then use the vector (a binary integer vector with only one non-zero value) to obtain the final schedule .Right now:

[0070] vector And the scheduling plan for all possible displacements is as follows:

[0071] (10)

[0072] (11)

[0073] The scheduling plan for the sth SAs is , then define a binary integer vector As the switch control of SAs, there is only one non-zero element in the vector equal to 1.

[0074] The power of residential user i in time slot t The power of the residential user at time slot t is obtained by adding the power of the working electrical equipment:

[0075] (12)

[0076] Then calculate the sum of the energy consumption of all residential users, that is, the energy purchased by the residential community from the grid:

[0077] (13)

[0078] The objective function of the mixed integer linear programming in this step is to minimize the total energy consumption of the community, that is:

[0079] (14)

[0080] According to the above objective function, the scheduling is solved according to the constraints of equations (7)-(11).

[0081] The following is an analysis through actual cases of distance to demonstrate the effect of the present invention.

[0082] 1. If Figure 3 The figure shows the load allocated by UWP and the original load in a residential community during the day. The energy allocated by UWP to each residential user is different and higher than the original load. The energy allocated to each residential user is different, reflecting the different preferences of users. The method of the present invention is referred to as the DR-UWP-MILP method. The allocated energy is higher than the original energy because SAs operate during low-price periods.

[0083] Table 2 compares the communication traffic of DR-MILP (i.e., not considering user preferences and only using mixed integer linear programming) and DR-UWP-MILP methods. Without UWP, DR communication needs to transmit information about 259 NSAs, 34 CAs, and 92 SAs. When using UWP, DR communication for NSAs and CAs only needs to transmit the number of different power levels in the residential community. There are 35 residential users in the residential community, so the number of NSAs and CAs that need to be transmitted is 35×6=210.

[0084] Table 2 Comparison of communication traffic between DR-MILP and DR-UWP-MILP

[0085]

[0086] 3. Further verify the effectiveness of the algorithm and introduce the peak average ratio (PAR)

[0087]

[0088] Figure 4 is the hourly load of the community before, after and after MILP and DR-UWP-MILP scheduling. The original load is higher during high-price periods and lower during low-price periods. DR-MILP and DR-UWP-MILP reduce the total energy cost of the residential community by reducing and shifting loads. The DR-MILP load is concentrated during low-price periods, so the PAR is higher than the original load. DR-UWP-MILP takes into account user preferences, improves the load curve, and reduces the total energy cost and PAR of the residential community. The dotted line represents the real-time electricity price as a reference.

[0089] Table 3 compares DR-MILP with DR-UWP-MILP. Compared with DR-MILP, DR-UWP-MILP obviously reduces the communication traffic and PAR. The total energy cost of DR-UWP-MILP is slightly higher than that of DR-MILP. The energy consumption reduction of DR-UWP-MILP and DR-MILP is the same.

[0090] Table 3 DR-MILP vs. DR-UWP-MILP

[0091]

[0092] five, Figure 5is the UWP allocated load of residential user 1 and the hourly load before and after scheduling. Taking the scheduling of residential user 1 as an example, the user has 9 NSAs, 1 CAs and 4 SAs. UWP allocates more load during low-price periods, SAs can work during low-price periods, and high-price periods can maintain the minimum demand of residential users. If SAs cannot be scheduled during low-price periods, they can work in the original periods. CAs operate at minimum power and NSAs operate at fixed power. Compared with the original load, DR-UWP-MILP increases the load during low-price periods and improves the load curve. The dotted line represents the real-time electricity price as a reference.

[0093] VI.,Table 4 shows the performance comparison of residential user 1 before and after,using DR-UWP-MILP.,For residential user 1, after DR-UWP-MILP, the communication,traffic, energy cost and energy consumption are reduced.

[0094] Table 4 Performance comparison of residential user 1 before and after using DR-UWP-MILP

[0095]

[0096] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A demand response scheduling method based on user preference and mixed integer linear programming, characterized in that: The method is used for demand response scheduling of a residential community, wherein a load aggregator is provided on the residential community side and a home energy management system is provided on the residential user side, and the method comprises the following steps: Step S1, the load aggregator obtains the historical data of the load of non-schedulable devices NSAs and the load of controllable devices CAs with labels, and performs fuzzy C-means clustering on the two historical data respectively; Step S2: The load aggregator obtains the energy cost tolerance and equipment information of residential users from each home energy management system, calculates the user willingness to pay of each residential user, and calculates the energy allocated to each residential user based on the user willingness to pay and the real-time electricity price, i.e., allocated energy; Step S3: For each residential user, the home energy management system calculates the minimum demand of the residential user. If the allocated energy is less than the minimum demand of the current residential user, the energy cost tolerance of the residential user is increased and the allocated energy is calculated again until the allocated energy is greater than or equal to the minimum demand. Step S4, the load aggregator uses mixed integer linear programming to perform demand response scheduling in residential communities; The specific process of step S1 is as follows: S11. Classify the historical data according to the labels, where the load of non-dispatchable equipment is recorded as NSA load and the load of controllable equipment is recorded as CAs load; S12. Fuzzy C-means clustering is performed on NSAs load and CAs load respectively. The number of clusters for both sets of historical data is 3. The cluster centers obtained after clustering are α is the device category, β is the different power levels of the cluster, β∈[1,3]; The specific process of step S2 is as follows: S21. For each residential user i, obtain the energy cost tolerance of residential user i in time slot t And the number of residential users’ two types of equipment clustered into three power levels is recorded as S22. Calculate the user willingness to pay for residential user i Among them, ω1, ω2, and ω3 are normalized parameters, and the operating power of the dispatchable equipment SAs is The running time is S23. Calculate the energy allocated to each residential user every hour, i.e., the allocated energy RTP t It is the real-time electricity price; In step S3, if the energy is allocated Less than the minimum demand of residential users Then increase the energy cost tolerance by a fixed amount and recalculate the allocated energy until the allocated energy is met. In step S4, the constraints of the mixed integer linear programming adopted include: Constraints for the nth NSAs of residential user i in time slot t: in is the fixed load of the nth NSA of user i; Constraints for the e-th CAs of residential user i in time slot t: in, is the minimum load of the e-th CAs of residential user i, is the maximum load of the e-th CAs of residential user i; The fixed consumption pattern vector of the s-th SAs of residential user i is the cyclic shift, the scheduling plan of the sth SAs in is a binary integer vector, Calculate the power of the residential user in time slot t Calculate the sum of energy consumption of all residential users, that is, the amount of energy purchased from the grid by the residential community To minimize the total energy consumption of residential communities Demand response scheduling is performed for the objective function.

Images