Non-intrusive load decomposition potential estimation driven demand response excitation pricing method

Abstract

The invention relates to a demand response excitation pricing method driven by non-intrusive load decomposition potential estimation. The method comprises the following steps: acquiring intelligent electric meter data of each user; according to the intelligent electric meter data of each user, non-intrusive load monitoring equipment based on a coupling hidden Markov model is adopted to decompose the total electricity consumption, and the total electricity consumption is converted into a stepped demand response curve; according to the stepped demand response curve, utilizing McCormick envelope to convert a profit maximization problem into a mixed integer linear programming optimization problem, and constructing a mixed integer linear programming model, the mixed integer linear programming model comprising a target function and constraint conditions; solving the mixed integer linear programming model to obtain a solving result; and determining a demand side management price incentive strategy according to a solving result, and pricing the electricity price of the user according to the demand side management price incentive strategy. The user side response capability can be enhanced, the demand response implementation effect is improved, and the economy and flexibility of power system operation are promoted.

Description

Technical Field

[0001] This application relates to the field of power technology, and in particular to a non-intrusive demand response incentive pricing method driven by load decomposition potential estimation. Background Technology

[0002] With increasing electrification and the generation of variable renewable energy, power systems face challenges such as heightened net load fluctuations and rising peak supply costs. Demand response (DR), as a mature tool, can improve reliability, reduce energy costs, smooth peak loads, and reduce emissions by shifting or reducing load during periods of system stress. DR projects are typically implemented by utilities or power companies, which aim to maximize their own profits. Price-based DR is one of the main forms, where power companies influence user electricity consumption behavior by designing pricing or incentive strategies.

[0003] Price-based time-of-use pricing (DR) relies on accurately reflecting demand elasticity (i.e., the sensitivity of electricity consumption to price changes). It is typically constructed using experimental or survey methods based on electricity consumption data under different electricity prices. For example, incentive-based DR systems have been designed and implemented for load service entities, using coupons to shift load from high-price periods to low-price periods; existing technologies analyze user acceptance of time-of-use pricing through selective experiments. However, these methods struggle to reflect the actual operating characteristics of equipment (electricity consumption is constrained by the physical characteristics of the equipment), leading to uncertainties in potential estimates, incentive strategies that may deviate from actual demand, and actual load reduction effects often falling short of expectations. User-side responsiveness is also poor, resulting in lower economic efficiency and flexibility in power system operation. Summary of the Invention

[0004] Therefore, it is necessary to provide a non-intrusive load decomposition potential estimation-driven demand response incentive pricing method that can effectively improve user-side response capabilities and the economy and flexibility of power system operation, addressing the aforementioned technical issues.

[0005] A non-intrusive demand response incentive pricing method driven by load decomposition potential estimation, the method comprising:

[0006] Collect data from each user's smart meter;

[0007] Based on the smart meter data provided by each user, a non-intrusive load monitoring device based on a coupled hidden Markov model is used to decompose the total electricity consumption and transform it into a stepped demand response curve.

[0008] Based on the stepped demand response curve, the profit maximization problem is transformed into a mixed-integer linear programming optimization problem using the McCormick envelope. A mixed-integer linear programming model is constructed, which includes an objective function and constraints.

[0009] The mixed-integer linear programming model is solved to obtain the solution results;

[0010] Based on the solution results, a demand-side management pricing incentive strategy is determined, and electricity prices for users are priced according to this strategy.

[0011] In one embodiment, the step of decomposing the total electricity consumption based on the smart meter data of each user, using a non-intrusive load monitoring device with a coupled Hidden Markov Model, and converting it into a stepped demand response curve includes:

[0012] For device e ∈ {1,...,E} in user b ∈ B, the usage state of device e in time period t is... Following a Markov chain, B represents the total number of users, E represents the total number of devices used by each user, 0 represents off, and 1 represents on;

[0013] Let the external environmental characteristics of time period t be c. t =(c 1,t ,...,c K,t K represents the number of external environmental features. The total active power observed by user b during time period t;

[0014] The initial state probability is:

[0015]

[0016] in, Let P(·) be the probability that user b's device e is in the usage state s during the initial time period 1. Let s be the usage state of device e of user b in the initial time period 1, and s be the device usage state, s∈{0,1}.

[0017] The observed power follows an additive normal distribution:

[0018]

[0019] in, Let μ0 be the usage state vector of all devices for user b at time t, and μ0 be the basic average power. e Let e ​​be the average power consumption of device e. σ represents the usage status of user b's device e during time period t. 2 For noise variance, This indicates that the distribution follows a normal distribution;

[0020] Considering external environmental characteristics and inter-device coupling, the state transition probability is:

[0021]

[0022] in, This represents the probability of how external environmental characteristics affect the usage status of home appliances. This is the coupling factor between devices, and the scaling constant is used for normalization. Let P(·|·) be the usage state of user b's device e at time t+1, and let P(·|·) be the conditional probability. The usage status of user b's device e at time t. The usage status of user b's device k at time t. The correlation of usage status changes between different devices k and e at time period t and time period t+1;

[0023] The joint probability distribution is:

[0024]

[0025] Where, parameter λ={π e A e (·),M e,k ,μ0,μ e ,σ 2}, For user b, the total power sequence over the entire time period 1 to T is given. For user b, the usage status sequence throughout the time period 1 to T. Let q be the probability distribution of the usage status of user b's device e during the initial time period t=1, where T is the maximum value during the entire time period. e,1 π represents the usage state of device e during the initial time period 1. e Let A be the probability distribution of the usage state of device e in the initial time period t=1. e (·) represents the characteristic conditional transfer factor, M e,k The correlation of usage status changes between different devices k and e;

[0026] Solving the optimization problem using the simulated annealing algorithm: Where, q (b)* The optimal state sequence for user b is the total power sequence for user b over the entire time period 1 to T. The most likely sequence of home appliance switching states under parameter λ, q (b) Let q be all possible state sequences for user b, and let q be the device usage state vector for all users.

[0027] The posterior power of the equipment is estimated from the posterior mean:

[0028]

[0029] in, The model estimates the total power consumption of user b's device e during time period t;

[0030] From B users, the non-intrusive load monitoring device provides time-series data for each device e. Aggregating active power into the overall portfolio of the power company:

[0031]

[0032] Where, p e,t The total power consumption of all devices e during time period t;

[0033] All electrical devices are categorized into 6 classes, and the total power consumption of devices in the same class is added together to obtain the total power consumption h of the m-th class of electrical devices in time period t. m,t (m∈{1,...,6});

[0034] The total power consumption h of the m-th type of electrical equipment during time period t m,t (m∈{1,...,6}) and the corresponding price v that the user is willing to accept for not using the m-th type of electrical equipment. m The resulting set {(h m,t ,v t This forms a stepped demand response curve.

[0035] In one embodiment, the objective function is:

[0036]

[0037] Where S is the reference apparent power, Let the time interval set be {1,...,T}, and τ t For retail electricity price, λ t For the wholesale market LMP during time period t, d t Let w be the total load for time period t. t These are the variables introduced after the nonlinear term is linearized.

[0038] In one embodiment, the constraints include constraints representing demand response, demand aggregation and boundary constraints, and linear inequality constraints.

[0039] In one embodiment, the constraint representing the demand response is:

[0040] 0≤x m,t ≤h m,t z m,t m∈M

[0041]

[0042] z m,t ∈{0,1}m∈M

[0043] z m,t ≤z m-1,t m≥2

[0044] Where, x m,t Let z represent the power consumption of the m-th type of electrical equipment in time period t of the stepped demand response curve, where M is the number of equipment categories. m,t Let σ be an auxiliary binary variable for the m-th type of electrical equipment during time period t. t Let be the demand response incentive value for time period t. If the demand response incentive value is greater than the minimum response value σ, then... t ≥v m,t Then z m,t The value is 1, and x m,t ≤h m,t If σ t <v m,t Then z m,t x is 0 m,t =0 indicates that this portion of the load has not been reduced, v m,t The minimum response value for the m-th type of electrical equipment during time period t. As the upper limit of incentives, z m-1,t Let m be the auxiliary binary variable for the (m-1)th type of electrical equipment in time period t.

[0045] In one embodiment, the constraints of the demand aggregation and boundaries are:

[0046]

[0047] d t =p t -r t

[0048]

[0049] Where, r t To reduce or shift the load, p t This is the original electrical load.

[0050] In one embodiment, the linear inequality constraint is:

[0051] w t ≥0

[0052]

[0053] in, For r t The upper boundary.

[0054] The aforementioned non-intrusive load decomposition potential estimation-driven demand response incentive pricing method collects smart meter data from each user. Based on this data, a non-intrusive load monitoring device using a coupled Hidden Markov Model (HMM) decomposes the total electricity consumption and transforms it into a stepped demand response curve. Using this stepped demand response curve, the profit maximization problem is transformed into a mixed-integer linear programming (MILP) optimization problem using the McCormick envelope. A MILP model is constructed, including an objective function and constraints. The MILP model is solved to obtain the solution results. Based on these results, a demand-side management (DSM) pricing incentive strategy is determined, and electricity prices are set for users according to this strategy. Therefore, applying NILM to analyze smart meter data identifies device-level electricity consumption characteristics and flexibility potential. Secondly, the NILM output is transformed into a stepped demand response curve. Finally, a MILP model is established to optimize demand response incentives, enhancing user-side response capabilities, improving the effectiveness of demand response implementation, and promoting the economy and flexibility of power system operation. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating a non-intrusive load decomposition potential estimation-driven demand response incentive pricing method in one embodiment.

[0056] Figure 2 This is a block diagram of a non-intrusive load decomposition potential estimation-driven demand response incentive pricing method in one embodiment.

[0057] Figure 3 This is a schematic diagram of the hourly power consumption decomposition results of an electric vehicle charging pile in one embodiment.

[0058] Figure 4 This is a schematic diagram of the aggregated demand response (DR) curve of a power company in a random one-hour period, as shown in one embodiment.

[0059] Figure 5 This is a schematic diagram comparing the daily settlement load under three different DR frameworks in one embodiment;

[0060] Figure 6 This is a schematic diagram comparing daily energy savings under three different DR frameworks in one embodiment;

[0061] Figure 7 This is a schematic diagram comparing daily profits under three different DR frameworks in one embodiment. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0063] In one embodiment, such as Figure 1 As shown, a non-intrusive demand response incentive pricing method driven by load decomposition potential estimation is provided, including the following steps:

[0064] Step S220: Collect smart meter data from each user.

[0065] Among these measures, the power company collects data from each user's smart meter.

[0066] Step S240: Based on the smart meter data of each user, a non-intrusive load monitoring device based on a coupled hidden Markov model is used to decompose the total electricity consumption and transform it into a stepped demand response curve.

[0067] Since smart meter data typically has a low sampling rate (≤1Hz), a Coupled Hidden Markov Model (CFHMM) is used to decompose the total household power. This method only requires steady-state external environmental features (such as date, temperature, day of the week, time of day, etc.) and does not require additional sensors.

[0068] Non-intrusive load monitoring (NILM) provides device-level load flexibility with finer granularity than traditional aggregated flexibility or single-resource demand response (such as temperature-controlled loads (TCLs) and electric vehicles (EVs)). After obtaining NILM results, these device-level power consumption curves are aggregated to determine flexible loads.

[0069] The end-user energy devices exhibit heterogeneity: some are reducible, some are transferable; some have high power consumption, while others are negligible. For demand response, power companies do not need to know the precise power consumption of each device. Therefore, devices can be categorized into six types based on typical demand response behavior patterns and willingness to accept (WTA), as shown in Table 1. Base load is classified as "Other" in the last row and assigned a WTA parameter of v6. By setting v6 to a large value, the optimization model treats it as non-reducible load.

[0070] Table 1. Equipment Grouping and WTA Parameters by Demand Response (DR) Behavior

[0071]

[0072] The total power consumption h of the m-th type of electrical equipment during time period t m,t (m∈{1,...,6}) and the corresponding acceptable price v for not using the m-th type of electrical equipment.m The resulting set {(h m,t ,v t This forms a stepped demand response curve, providing a basis for incentive strategies.

[0073] It should be understood that DR potential can be extracted from the results of the NILM algorithm.

[0074] Step S260: Based on the stepped demand response curve, the profit maximization problem is transformed into a mixed-integer linear programming optimization problem using the McCormick envelope. A mixed-integer linear programming model is constructed, which includes the objective function and constraints.

[0075] Among them, potential is described by a stepped demand response curve, and the profit maximization problem is transformed into an easily solvable MILP by using the McCormick envelope.

[0076] Among them, the DR incentive strategy is modeled as a mixed integer linear programming (MILP) problem, and the NILM potential is used as a demand response constraint to construct a profit maximization MILP model for power companies.

[0077] The demand response (DR) incentive strategy is formulated as a MILP (Multi-Level Liberalization) optimization problem. The DR potential based on MILM is introduced as a constraint representing the demand response. Other constraints include demand aggregation and boundary conditions. The initial objective function is to maximize the power company's profit during the planning period and avoid losses during periods of high marginal price (LMP), expressed as: Profit includes electricity sales revenue and incentive costs. The initially constructed objective function contains a bilinear term σ. t r t This leads to the problem exhibiting non-convex and nonlinear characteristics, making it difficult to solve. Therefore, the McCormick envelope method is used to linearize this bilinear term. A new decision variable w is introduced. t and utilize boundary conditions and Then, replace the bilinear term in the initially constructed objective function with w. t And by adding the following four linear inequality constraints, the linearized objective function is:

[0078] This problem is a mixed integer linear programming (MILP) problem, which can be solved using readily available solvers.

[0079] Step S280: Solve the mixed-integer linear programming model and obtain the solution results.

[0080] Step S300: Based on the solution results, determine the demand-side management price incentive strategy, and price the user's electricity according to the demand-side management price incentive strategy.

[0081] The process involves decomposing total electricity consumption into device-level curves using non-intrusive load monitoring (NILM) devices. Different devices are then grouped and assigned a WTA (Minimum Compensation Required by Users to Provide Grid Services, such as Peak Shaving, Provision of Reserves, or Adaptation to Renewable Energy Fluctuations) to infer user flexibility and aggregate demand response (DR) potential. Finally, a mixed-integer linear programming (MILP) model combined with NILM potential is used to maximize power company profits.

[0082] The aforementioned non-intrusive load decomposition potential estimation-driven demand response incentive pricing method applies NILM analysis to smart meter data to identify equipment-level electricity consumption characteristics and flexibility potential. Secondly, it transforms the NILM output into a stepped demand response curve. Finally, it establishes a MILP model to optimize DR incentives, which can enhance user-side response capabilities, improve the effectiveness of demand response implementation, and promote the economy and flexibility of power system operation.

[0083] In one embodiment, based on smart meter data from each user, a non-intrusive load monitoring device employing a coupled Hidden Markov Model decomposes the total electricity consumption and transforms it into a stepped demand response curve, including:

[0084] For device e ∈ {1,...,E} in user b ∈ B, the usage state of device e in time period t is... Following a Markov chain, B represents the total number of users, E represents the total number of devices used by each user, 0 represents off, and 1 represents on;

[0085] Let the external environmental characteristics of time period t be c. t =(c 1,t ,…,c K,t K represents the number of external environmental features. The total active power observed by user b during time period t;

[0086] The initial state probability is:

[0087]

[0088] in, Let P(·) be the probability that user b's device e is in the usage state s during the initial time period 1. Let s be the usage state of device e of user b in the initial time period 1, and s be the device usage state, s∈{0,1}.

[0089] The observed power follows an additive normal distribution:

[0090]

[0091] in, Let μ0 be the usage state vector of all devices for user b at time t, and μ0 be the basic average power. e Let e ​​be the average power consumption of device e. σ represents the usage status of user b's device e during time period t. 2 For noise variance, This indicates that the distribution follows a normal distribution;

[0092] Considering external environmental characteristics and inter-device coupling, the state transition probability is:

[0093]

[0094] in, As a feature condition transfer factor, it represents the probability of the influence of external environmental characteristics on the usage status of home appliances. This is the coupling factor between devices, and the scaling constant is used for normalization. Let P(·|·) be the usage state of user b's device e at time t+1, and let P(·|·) be the conditional probability. The usage status of user b's device e at time t. The usage status of user b's device k at time t. As a cross-device coupling factor, it is used to describe the correlation of usage state changes between different devices k and e at time period t and time period t+1 (e.g., the linkage between washing machines and dryers).

[0095] The joint probability distribution is:

[0096]

[0097] Where, parameter λ={π e A e (·),M e,k ,μ0,μ e ,σ 2 The specific values ​​of each parameter in λ are learned from the training data using the expectation-maximization algorithm (i.e., other constraints). For user b, the total power sequence over the entire time period 1 to T is given. For user b, the usage status sequence throughout the time period 1 to T. Let q be the probability distribution of the usage status of user b's device e during the initial time period t=1, where T is the maximum value during the entire time period. e,1 π represents the usage state of device e during the initial time period 1. e Let A be the probability distribution of the usage state of device e in the initial time period t=1. e (·) represents the characteristic conditional transfer factor, Me,k The correlation of usage status changes between different devices k and e;

[0098] Solving the optimization problem using the simulated annealing algorithm: Where, q (b)* The optimal state sequence for user b is the total power sequence for user b over the entire time period 1 to T. The most likely sequence of home appliance switching states under parameter λ, q (b) Let q be all possible state sequences for user b, and let q be the device usage state vector for all users.

[0099] The posterior power of the equipment is estimated from the posterior mean:

[0100]

[0101] in, The model estimates the total power consumption of user b's device e during time period t;

[0102] From B users, the non-intrusive load monitoring device provides time-series data for each device e. Aggregating active power into the overall portfolio of the power company:

[0103]

[0104] Where, p e,t The total power consumption of all devices e during time period t;

[0105] All electrical equipment is categorized into 6 types, and the total power consumption of equipment in the same type is added together to obtain the total power consumption h of each type of equipment in time period t. m,t (m∈{1,...,6});

[0106] The total power consumption h of the m-th type of electrical equipment during time period t m,t (m∈{1,...,6}) and the corresponding acceptable price v for not using the m-th type of electrical equipment. m The resulting set {(h m,t ,v t This forms a stepped demand response curve.

[0107] In one embodiment, the objective function is:

[0108]

[0109] Where S is the reference apparent power, Let the time interval set be {1,...,T}, and τ t For retail electricity price, λ t For the wholesale market LMP during time period t, dt Let w be the total load for time period t. t These are the variables introduced after the nonlinear term is linearized.

[0110] In one embodiment, constraints include constraints representing demand response, demand aggregation and boundary constraints, and linear inequality constraints.

[0111] In one embodiment, the constraint representing the demand response is:

[0112] 0≤x m,t ≤h m,t z m,t m∈M

[0113]

[0114] z m,t ∈{0,1}m∈M

[0115] z m,t ≤z m-1,t m≥2

[0116] Where, x m,t Let z represent the power consumption of the m-th type of electrical equipment in time period t of the stepped demand response curve, where M is the number of equipment categories. m,t Let σ be an auxiliary binary variable for the m-th type of electrical equipment during time period t. t Let be the demand response incentive value for time period t. If the demand response incentive value is greater than the minimum response value σ, then... t ≥v m,t Then z m,t The value is 1, and x m,t ≤h m,t If σ t <v m,t Then z m,t x is 0 m,t =0 indicates that this portion of the load has not been reduced, v m,t The minimum response value for the m-th type of electrical equipment during time period t. As the upper limit of incentives, z m-1,t Let m be the auxiliary binary variable for the (m-1)th type of electrical equipment in time period t.

[0117] In one embodiment, the constraints on demand aggregation and boundaries are:

[0118]

[0119] d t =p t -r t

[0120]

[0121] Where, r t To reduce or shift the load, p t This represents the original total electrical load.

[0122] In one embodiment, the linear inequality constraint is:

[0123] w t ≥0

[0124]

[0125] in, For r t The upper boundary.

[0126] In one embodiment, to verify the effectiveness of the non-intrusive load decomposition potential estimation-driven demand response incentive pricing method of this application, a bridge was built between NILM technology and DR project design.

[0127] The proposed non-intrusive load decomposition potential estimation-driven demand response incentive pricing method was tested on publicly available datasets. Smart meter data and nodal marginal prices (LMPs) were obtained from the Pecan Street dataset, specifically using smart meter readings (1-minute resolution) and weekly average LMP data from independent system operators in the study area. Experiments were conducted on an Intel Core i7-13850HX platform (2.10GHz CPU, 64GB RAM, 64-bit Windows system), and the MILP model was solved using Pyomo with the glpk solver.

[0128] A coupled hidden Markov model (CFHMM) was trained based on 297 days of smart meter data and tested on the data from the last week. A comparative analysis was conducted on the decomposition results for six types of electrical equipment. Figure 3 This paper presents a comparison between the actual hourly power consumption of electric vehicle charging stations on the test dataset and the NILM decomposition results. The root mean square error (RMSE) is used to evaluate performance, quantifying the deviation between predicted and actual power. The overall RMSE is 3.31 MWh, indicating that CFHMM has reasonable accuracy in device-level load decomposition. Although there are superimposed signal errors caused by multiple devices operating simultaneously, it meets the practical requirements for DR potential estimation.

[0129] Based on the equipment-level load data obtained from NILM decomposition, a stepped demand response curve was constructed. The Willingness to Accept (WTA) parameter was set as v = [v1,...,v6] = [5,10,15,20,25,100] (yuan / MWh). The price that users are willing to accept for not using Category 6 electrical equipment, v6 = 100, was set as an extremely high value, stemming from the incentive ceiling in this case study. Constraints per megawatt-hour. Note that these WTAs are rough approximations used in this case study. Figure 4 The DR curve of the aggregate is shown.

[0130] Next, the stepped demand response curve is incorporated into the profit maximization problem of MILP power companies for analysis. Three scenarios are set up: (1) Scenario 1: baseline case without DR; (2) Scenario 2: using a uniform elasticity model; (3) Scenario 3: applying the proposed framework. In Scenario 2, a uniform elasticity model is used, specifically: the WTA parameter adopts the average value. The demand response (DR) potential is approximately one-quarter of the total electricity consumption, and its demand response (DR) stepped response curve is modeled as a single step, i.e. For responsive loads. Electricity consumption comparisons under three scenarios: Figure 5 As shown in the figure. The results show that after introducing DR, the electricity consumption is significantly reduced compared to the baseline scenario, and the proposed framework (Scenario 3) achieves a greater load reduction than the uniform elastic model.

[0131] In addition, the daily electricity savings and profit results are as follows: Figure 6 and Figure 7 As shown in the figure, data analysis demonstrates that demand response (DR) not only resulted in substantial daily electricity savings but also significantly improved power company profits. The method presented in this application outperforms the traditional uniform elasticity model in both energy-saving efficiency and profitability. Particularly noteworthy is that on day 3, without DR, the power company would have faced severe losses due to extremely high system load, a significant increase in the nodal marginal price (LMP), and a fixed retail price; however, in scenario 3, DR successfully reduced / transferred the load, allowing the power company to avoid this major loss.

[0132] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0133] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0134] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A non-intrusive demand response incentive pricing method driven by load decomposition potential estimation, characterized in that, The method includes: Collect data from each user's smart meter; Based on the smart meter data provided by each user, a non-intrusive load monitoring device based on a coupled hidden Markov model is used to decompose the total electricity consumption and transform it into a stepped demand response curve. Based on the stepped demand response curve, the profit maximization problem is transformed into a mixed-integer linear programming optimization problem using the McCormick envelope. A mixed-integer linear programming model is constructed, which includes an objective function and constraints. The mixed-integer linear programming model is solved to obtain the solution results; Based on the solution results, a demand-side management pricing incentive strategy is determined, and electricity prices for users are priced according to this strategy.

2. The method according to claim 1, characterized in that, The process involves using a non-intrusive load monitoring device with coupled Hidden Markov Models to decompose the total electricity consumption based on the smart meter data from each user, and then transforming it into a stepped demand response curve. This includes: For device e ∈ {1,...,E} in user b ∈ B, the usage state of device e in time period t is... Following a Markov chain, B represents the total number of users, E represents the total number of devices used by each user, 0 represents off, and 1 represents on; Let the external environmental characteristics of time period t be c. t =(c 1,t ,...,c K,t K represents the number of external environmental features. The total active power observed by user b during time period t; The initial state probability is: in, Let P(·) be the probability that user b's device e is in the usage state s during the initial time period 1. Let s be the usage state of device e of user b in the initial time period 1, and s be the device usage state, s∈{0,1}. The observed power follows an additive normal distribution: in, Let μ0 be the usage state vector of all devices for user b at time t, and μ0 be the basic average power. e Let e ​​be the average power consumption of device e. σ represents the usage status of user b's device e during time period t. 2 For noise variance, This indicates that the distribution follows a normal distribution; Considering external environmental characteristics and inter-device coupling, the state transition probability is: in, This represents the probability of how external environmental characteristics affect the usage status of home appliances. This is the coupling factor between devices, and the scaling constant is used for normalization. Let P(·|·) be the usage state of user b's device e at time t+1, and let P(·|·) be the conditional probability. The usage status of user b's device e at time t. The usage status of user b's device k at time t. The correlation of usage status changes between different devices k and e at time period t and time period t+1; The joint probability distribution is: Where, parameter λ={π e A e (·),M e,k ,μ0,μ e ,σ 2 }, For user b, the total power sequence over the entire time period 1 to T is given. For user b, the usage status sequence throughout the time period 1 to T. Let q be the probability distribution of the usage status of user b's device e during the initial time period t=1, where T is the maximum value during the entire time period. e,1 π represents the usage state of device e during the initial time period 1. e Let A be the probability distribution of the usage state of device e in the initial time period t=1. e (·) represents the characteristic conditional transfer factor, M e,k The correlation of usage status changes between different devices k and e; Solving the optimization problem using the simulated annealing algorithm: Where, q (b)* The optimal state sequence for user b is the total power sequence for user b over the entire time period 1 to T. The most likely sequence of home appliance switching states under parameter λ, q (b) Let q be all possible state sequences for user b, and let q be the device usage state vector for all users. The posterior power of the equipment is estimated from the posterior mean: in, The model estimates the total power consumption of user b's device e during time period t; From B users, the non-intrusive load monitoring device provides time-series data for each device e. Aggregating active power into the overall portfolio of the power company: Where, p e,t The total power consumption of all devices e during time period t; All electrical devices are categorized into 6 classes, and the total power consumption of devices in the same class is added together to obtain the total power consumption h of the m-th class of electrical devices in time period t. m,t (m∈{1,...,6}); The total power consumption h of the m-th type of electrical equipment during time period t m,t (m∈{1,...,6}) and the corresponding price v that the user is willing to accept for not using the m-th type of electrical equipment. m The resulting set {(h m,t ,v t This forms a stepped demand response curve.

3. The method according to claim 2, characterized in that, The objective function is: Where S is the reference apparent power, Let the time interval set be {1,...,T}, and τ t For retail electricity price, λ t For the wholesale market LMP during time period t, d t Let w be the total load for time period t. t These are the variables introduced after the nonlinear term is linearized.

4. The method according to claim 3, characterized in that, The constraints include constraints representing demand response, constraints on demand aggregation and boundaries, and linear inequality constraints.

5. The method according to claim 4, characterized in that, The constraint representing the demand response is: 0≤x m,t ≤h m,t With m,t m∈M With m,t ∈{0,1}m∈M With m,t ≤of m-1,t m≥2 Where, x m,t Let z represent the power consumption of the m-th type of electrical equipment in time period t of the stepped demand response curve, where M is the number of equipment categories. m,t Let σ be an auxiliary binary variable for the m-th type of electrical equipment during time period t. t Let be the demand response incentive value for time period t. If the demand response incentive value is greater than the minimum response value σ, then... t ≥v m,t Then z m,t The value is 1, and x m,t ≤h m,t If σ t <v m,t Then z m,t x is 0 m,t =0 indicates that this portion of the load has not been reduced, v m,t The minimum response value for the m-th type of electrical equipment during time period t. As the upper limit of incentives, z m-1,t Let m be the auxiliary binary variable for the (m-1)th type of electrical equipment in time period t.

6. The method according to claim 5, characterized in that, The constraints on the aggregation and boundaries of the requirements are as follows: d t =p t -r t Where, r t To reduce or shift the load, p t This is the original electrical load.

7. The method according to claim 6, characterized in that, The linear inequality constraint is: w t ≥0 in, For r t The upper boundary.

Images