High-energy-consumption load side resource aggregation regulation and control modeling method containing distributed light storage

Abstract

The invention relates to the technical field of load regulation and control, in particular to a distributed light storage-containing high-energy-consumption load side resource aggregation regulation and control modeling method. Comprising the following steps: S1, establishing a distributed light storage-containing regulation potential and regulation cost model of high-energy-consumption industry enterprise load resources: establishing a structure-geography-meteorology-power-regulation five-dimensional model comprising a structure model, a geography model, a meteorology model, a power model and a regulation model; s2, proposing a distributed optical storage-containing high-energy-consumption industry enterprise load resource aggregation method based on transverse federated learning; comprising the steps of encrypted entity alignment, feature data preprocessing, task division and clustering solution. The designed five-dimensional model of the enterprise load resources in the high-energy-consumption industry can accurately represent the real physical characteristics of the enterprise load resources, and the aggregation regulation and control requirements are met; the transverse federation learning based on the load regulation and control model is suitable for collaborative clustering of enterprises in high-energy-consumption industries, can meet privacy protection requirements, and provides an accurate clustering result.

Description

Technical Field

[0001] The present invention relates to the technical field of load regulation, and in particular to a modeling method for resource aggregation regulation of a high-energy-consuming load side including distributed photovoltaic storage. Background Art

[0002] With the acceleration of energy transformation, the application of distributed photovoltaic storage in high-energy-consuming industries and enterprises is becoming more and more extensive. However, the load resource aggregation and regulation of high-energy-consuming industries and enterprises with distributed photovoltaic storage face many challenges, which also prompts the continuous deepening of related research.

[0003] In the field of new energy power stations, especially photovoltaic power generation, current domestic and foreign research mainly focuses on extracting characteristic quantities such as photovoltaic inverter parameter sensitivity as clustering indicators, using clustering algorithms to obtain online weighted characteristic distances, and then realizing the classification of centralized photovoltaic inverters to derive the aggregation model of equivalent photovoltaic power stations. For example, in large photovoltaic power stations, by analyzing the output characteristics of the inverters, inverters with similar characteristics are classified into one category, and a simplified power station model is constructed to facilitate the evaluation and regulation of the overall operation status of the power station. However, the premise of this method is that the inherent characteristics of photovoltaic inverters are basically the same. The load resources of high-energy-consuming industry enterprises are widely distributed and have different characteristics. The distributed photovoltaic storage systems they contain have significant differences in equipment type, operation mode, capacity configuration, etc., which cannot meet the application conditions of this method, making it difficult for these research results on new energy power stations to be directly applied to the aggregation regulation of load resources of high-energy-consuming industry enterprises.

[0004] At present, relevant frontier explorations have been carried out at home and abroad for the aggregation and control of load resources of enterprises in high-energy-consuming industries with wide-area dispersion. The main idea is to comprehensively consider the electrical distance and power similarity between each distributed resource node, and to divide the clusters and perform equivalent aggregation modeling. In microgrids and virtual power plants with multiple distributed resources, this method has achieved certain application results. For example, in some microgrid projects, by calculating the electrical distance between different distributed power sources and loads, combined with power characteristics, they are divided into different clusters, and equivalent models are constructed to simplify system analysis and control. However, the existing results still have obvious shortcomings. On the one hand, the differences in physical characteristics (such as differences in load fluctuation characteristics caused by different production processes), operating rules (such as different production shifts and equipment start-stop rules of different enterprises), and photovoltaic and energy storage configurations (such as different photovoltaic module types, installation angles, energy storage capacities, and charging and discharging strategies) of load resources of enterprises in high-energy-consuming industries are not fully considered. These differences make the division of clusters and the determination of the number of clusters extremely complicated, making it difficult to accurately achieve effective aggregation control. On the other hand, there are a large number of information islands between enterprises in high-energy-consuming industries. Enterprises are reluctant to fully share detailed information about their own load resources for reasons of data security and commercial competition, which makes it difficult to effectively integrate data between enterprises, greatly hindering the aggregation of load resources and bringing more severe challenges to the aggregation of load resources in high-energy-consuming industries. Further in-depth research is urgently needed to break through these difficulties and achieve efficient and accurate aggregation control. In view of this, we propose a modeling method for resource aggregation control on the high-energy-consuming load side with distributed photovoltaic storage. Summary of the invention

[0005] The purpose of the present invention is to provide a high-energy-consuming load-side resource aggregation and regulation modeling method containing distributed photovoltaic storage to solve the problems raised in the above-mentioned background technology.

[0006] In order to solve the above technical problems, the purpose of the present invention is to provide a high-energy-consuming load-side resource aggregation control modeling method containing distributed photovoltaic storage, comprising the following steps:

[0007] S1. Establish the regulation potential and regulation cost model of the load resources of high-energy-consuming industry enterprises with distributed photovoltaic storage: In view of the regulation needs of high-energy-consuming industry enterprises with distributed photovoltaic storage to participate in the county power grid, a five-dimensional model of "structure-geography-meteorology-power-regulation" of the load resources of high-energy-consuming industry enterprises is established;

[0008] Among them, the "structure-geography-climate-power-regulation" five-dimensional model includes a structure model, a geography model, a meteorology model, a power model and a regulation model, and its establishment process is as follows:

[0009] First, the structural model is established according to the internal composition and structural relationship of the load resources of high-energy-consuming industry enterprises; secondly, the geographical model and meteorological model are established according to their geographical information; then, according to their electrical mechanism, their output model, i.e., power model, is established; finally, according to their output model and combined with their regulation characteristics, their regulation model is established;

[0010] S2. Propose a method for aggregating load resources of enterprises in high-energy-consuming industries with distributed photovoltaic storage based on horizontal federated learning; including:

[0011] S2.1. Alignment of encrypted entities: First, align the encrypted load-side resource entities of high-energy-consuming enterprises held by the power company and various high-energy-consuming enterprises, that is, find the load-side resource entities shared by the power company and various high-energy-consuming enterprises in the region;

[0012] S2.2, Feature data preprocessing: Secondly, the feature vector data of high energy-consuming industry enterprises shared by both parties are preprocessed, entities containing invalid data are eliminated, and valid data are normalized;

[0013] S2.3, Task division: Each high-energy-consuming industry enterprise uses its own structural model, geographical model, meteorological model, power model, regulation model and other characteristic data to cluster; the power company plays the role of a collaborator, responsible for determining the number of aggregates, initializing the center of the aggregate, and summarizing and updating the combination of model parameters;

[0014] S2.4, clustering solution: According to the task division in step S2.3, enterprises in high-energy-consuming industries cluster load resources.

[0015] As a further improvement of the technical solution, in the structural model, the load resource structure model of high energy-consuming industry enterprises mainly includes photovoltaic installation methods and energy storage configuration in addition to the adjustable load equipment contained in itself. The structural model S i The definition is as follows:

[0016] First, photovoltaic installation method I i :Distributed photovoltaic installation methods are divided into ground / roof, wall, etc.; among them, ground / roof installation methods can be divided into fixed brackets, single-axis tracking, dual-axis tracking, etc. according to their bracket types, and wall installation methods also include east, west, south, etc.; the distributed photovoltaic installation method model can be encoded using a two-digit integer variable;

[0017] Second, energy storage configuration E i:The structural model of the distributed photovoltaic storage system needs to be differentiated according to whether it is configured with energy storage and the energy storage configuration method; the energy storage configuration method includes three different access methods: DC side, AC side, and independent type; when the DC side is configured, it is a photovoltaic storage integrated grid-connected mode; when the AC side is configured, photovoltaic and energy storage are two independent systems, but the AC side grid-connected point is the same; the independent type means that there is no other photovoltaic system at the AC side grid-connected point of the energy storage system; the energy storage configuration model can be encoded using a two-digit integer variable.

[0018] As a further improvement of the technical solution, in the geographical model, the load resource geographical model of high-energy-consuming industry enterprises refers to the geographical coordinates of the geometric center of the area where the enterprise is located, expressed by the longitude and latitude of the geometric center, that is:

[0019] G i =[L i ,B i ] (1)

[0020] In the formula, G i is the geographic model, L i is longitude, B i is latitude;

[0021] In the meteorological model, the load power consumption and distributed photovoltaic power generation in the load resources of high-energy-consuming industry enterprises are affected by meteorological factors, mainly including irradiance and temperature; therefore, the meteorological model can be expressed as:

[0022] W i =[R i ,T i ] (2)

[0023] Where W i is the meteorological model, R i is the irradiance, T i For temperature, the measured data are taken from the meteorological monitoring station in the area where the high energy-consuming industry enterprise is located.

[0024] As a further improvement of the technical solution, in the power model, the power model of the load resources of high-energy-consuming industry enterprises is mainly composed of an adjustable load model, a photovoltaic component model, an energy storage battery model and a comprehensive efficiency model, and the model is defined as follows:

[0025] The adjustable load model includes an adjustable power size constraint, a maximum duration constraint and an adjustable cost model, and the specific algorithm is:

[0026] A. Adjust power constraints:

[0027]

[0028] In the formula, They represent the upward power and downward power of the adjustable load i in time period t respectively; The flags respectively indicate whether the adjustable load i in period t participates in the upward or downward regulation. A value of 1 indicates that the load participates in the regulation, and a value of 0 indicates that the load does not participate in the regulation. They represent the maximum upward and downward adjustment capabilities of the adjustable load i during period t respectively; They are the start and end times of the adjustable upward and downward time periods, and there can be multiple adjustable time periods;

[0029] B. Maximum duration constraint:

[0030]

[0031] In the formula, [N s ,N e ] is the start and end time of the period that can be adjusted upward / downward, with upward adjustment being positive and downward adjustment being negative; It is the flag bit of the adjustable load i participating in the upward / downward adjustment in period t, which is positive for upward adjustment and negative for downward adjustment; For the period [N′ s ,N′ e ] maximum duration within ;

[0032] It should be noted that [N′ s ,N′ e ] may contain multiple N s ,N e ;

[0033] C. Adjustment cost model:

[0034]

[0035] In the formula, Respectively indicate an increase or decrease in unit price; They represent the upward and downward adjustment of the adjustable load i in period t respectively; N dr Indicates the number of adjustable loads.

[0036] As a further improvement of the technical solution, in the photovoltaic module model, the power generation of the photovoltaic module is mainly affected by the irradiance and temperature in the meteorological elements; the relationship between irradiance, temperature and power is as follows:

[0037]

[0038] Where P PV is the actual power, R is the actual irradiance; T is the actual temperature; R ref is the light irradiance under IEC standard conditions, which is 1000W / m 2 ; T refis the temperature under standard conditions, that is, 298K (25°C); P ref That is the maximum output under standard conditions; k is the power temperature coefficient.

[0039] As a further improvement of the technical solution, in the energy storage battery model, the energy storage battery model is mainly composed of the corresponding relationship and limit constraints between the charge and discharge power and its remaining capacity (state-of-charge, SOC); including:

[0040] a. Charge and discharge power constraints:

[0041]

[0042] In the formula, are the maximum discharge power and maximum charging power of unit e respectively; are the minimum discharge power and minimum charging power of unit e respectively (in kW); are the discharge power and charging power of unit e respectively; They are respectively the discharge state and the charge state. When the value is 1, it means discharge / charge, and when it is 0, it means no discharge / no charge;

[0043] b. Charge and discharge state constraints: The charge and discharge states are mutually exclusive, that is, they satisfy the following constraints:

[0044]

[0045] At the same time, the following state transition constraints are met:

[0046]

[0047] In the formula, represents the discharge state of unit e at time t+1, They are respectively the start-discharge state and the start-charge state. When the value is 1, it means that the discharge / charge is started at the current moment, and when it is 0, it means that there is no action at the current moment.

[0048] c.SOC constraint:

[0049]

[0050] In the formula, S e (t) is the SOC of unit e in period t; S e (t-1) is the SOC of unit e in the period t-1; is the active power of unit e in period t; t The duration of a period (unit: hours); are the maximum and minimum allowable values ​​of SOC respectively; S 0is the initial capacity of SOC; S e (0), S e (t) represents the SOC at the start and end time of an operation cycle t∈[0,T], respectively;

[0051] d. Continuous charge and discharge constraints: In order to avoid frequent switching of energy storage charge and discharge states, charging is only performed when the SOC is below a certain threshold, and discharging is performed when the SOC is above a certain threshold, that is, the following constraints are met:

[0052]

[0053] In the formula, They are respectively the charge threshold and the discharge threshold.

[0054] As a further improvement of the technical solution, in the comprehensive efficiency model, the comprehensive efficiency model of the load resources of high-energy-consuming industry enterprises takes into account the comprehensive efficiency of various links such as photovoltaic module efficiency, battery charging and discharging efficiency, inverter efficiency, and distribution system efficiency:

[0055]

[0056] Where P i is the total power of the high energy consumption industry enterprise, P LD (t) is the power load power of the high energy consumption industry enterprise, η 1 , η 2 , η 3 , η 4 They are the comprehensive efficiency coefficients of photovoltaic modules and their inverters, energy storage discharge, energy storage charging, and distribution systems.

[0057] As a further improvement of the technical solution, in the control model, the control model of the load resources of high-energy-consuming industry enterprises refers to the upper and lower limits of the active control response capability of the superior control center and the corresponding response cost. The model is defined as follows:

[0058] 1) Upper and lower limits of response capability, including:

[0059] a. Upper limit of response capacity: The upper limit of load resource response capacity of enterprises in high energy consumption industries is determined by the maximum downward adjustment power of the adjustable load and the increaseable discharge power of the energy storage system, that is:

[0060]

[0061] b. Lower limit of response capability: The lower limit of the load resource response capability of enterprises in high energy consumption industries is determined by the maximum increase power of the adjustable load, the real-time output of the photovoltaic system and the increaseable charging power of the energy storage system, that is:

[0062]

[0063] In the formula, They are the upper and lower limits of the load resource response capabilities of enterprises in high energy-consuming industries;

[0064] 2) Response costs, including:

[0065] a. Response cost of increasing power: The response cost of increasing power of load resources in high-energy-consuming industries mainly comes from energy storage output and adjustable load reduction. Photovoltaic does not have the ability to increase output. Therefore, the response cost of increasing power is expressed as:

[0066]

[0067] b. Downward response cost: The response cost of the load resources of high-energy-consuming industry enterprises to reduce power is mainly due to the additional cost caused by the increase in electricity consumption of adjustable loads and the power generation loss caused by the reduction of photovoltaic output. There is no additional cost for energy storage to reduce output. Therefore, the downward response cost is expressed as:

[0068]

[0069] In the formula, They are the response costs of increasing and decreasing the power of load resources in high energy-consuming industries; is the electricity cost of the energy storage system; is the on-grid electricity price of the photovoltaic system; ΔP PV- (t) is the abandoned power of the PV system; Respectively indicate an increase or decrease in unit price; They respectively represent the upward and downward adjustment amounts of the adjustable load i in period t.

[0070] As a further improvement of the technical solution, in step S2.2, the algorithm for normalizing the valid data is:

[0071]

[0072] In the formula, x i represents a characteristic quantity data provided by power companies or high-energy-consuming enterprises, x max and x min Respectively represent the maximum and minimum values ​​of the feature quantity data.

[0073] As a further improvement of the technical solution, in step S2.4, the specific process of clustering load resources of enterprises in high energy consumption industries includes the following steps:

[0074] S2.4.1. Initialize the clustering result error threshold to δ;

[0075] S2.4.2. Initialize the number of aggregates: The power company initially defines the number of aggregates, M;

[0076] S2.4.3. Initialize cluster centers: Each high-energy-consuming industry enterprise initializes a set of cluster centers based on the entity and feature data it has. The initialization cluster center of high-energy-consuming industry enterprise j is:

[0077] c jk =[S i ,W i ,P i ,ΔP i ,f i ] (19)

[0078]

[0079] In the formula, k is the number of the cluster center, k∈Z M , Z M represents a set of integers from 1 to M, j is the number of enterprises in high energy consumption industries, j∈Z L , L is the number of enterprises in high-energy-consuming industries participating in horizontal federated learning; S i is the structural model, W i is the meteorological model, P i is the power model, ΔP i is the control model, f i To regulate response costs; is the average value of each cluster center point;

[0080] After each high-energy-consuming industry enterprise initializes the cluster center, it is encrypted and uniformly submitted to the partner power company for aggregation. The power company calculates the average value of each cluster center point. As the final initialization cluster center, it is sent to each participant;

[0081] S2.4.4, separate clustering calculation: After receiving the cluster center issued by the power company, each participant decrypts it to obtain the current cluster center, traverses the load resources they have, and calculates the L2 paradigm distance from each high-energy-consuming industry enterprise to each cluster center; the L2 paradigm distance d from high-energy-consuming industry enterprise i to the kth cluster center ik The calculation is as follows:

[0082]

[0083] Where M i is the load regulation model for the i-th high energy consumption industry enterprise;

[0084] Determine the cluster to which each high-energy-consuming industry enterprise belongs based on the L2 paradigm distance, that is, find the cluster center point closest to the high-energy-consuming industry enterprise, and add the high-energy-consuming industry enterprise to the cluster represented by the cluster center point, so as to achieve the update of the cluster;

[0085] S2.4.5. Update cluster centers separately: Each participant calculates the average value of the characteristic vectors of all high-energy-consuming industry enterprises in each aggregate for the updated aggregate, and uses it as the updated cluster center of the aggregate. After encryption, it is uniformly submitted to the partner power company for summary update;

[0086] S2.4.6. The power company summarizes and updates the cluster center: After receiving the updated cluster centers submitted by each participant, the power company calculates the average value of each new cluster center as the final updated cluster center and sends it to each participant;

[0087] S2.4.7, iteratively execute steps S2.4.4 to S2.4.6 until the average distance between the feature vectors of the M cluster centers obtained from the two previous iterations is less than a certain threshold, that is, the clustering process has converged and the cluster centers no longer change; if the maximum number of iterations is reached and there is still no convergence, continue to step S2.4.8;

[0088] S2.4.8, modify the number of aggregates and repeat steps S2.4.3 to S2.4.7; if the results of step S2.4.7 do not converge after multiple attempts at the number of aggregates, you can relax the threshold, that is, return to step S2.4.1 to modify the clustering result error threshold, and then repeat steps S2.4.2 to S2.4.8 until a convergent result is obtained.

[0089] Compared with the prior art, the present invention has the following beneficial effects:

[0090] 1. In the modeling method of high-energy-consuming load-side resource aggregation and regulation with distributed photovoltaic storage, the constructed five-dimensional model of "structure-geography-meteorology-power-regulation" comprehensively and deeply covers the key elements of load resources of high-energy-consuming industry enterprises. The five-dimensional model of load resources of high-energy-consuming industry enterprises can accurately characterize their real physical characteristics. The five-dimensional models cooperate with each other and accurately characterize the real physical characteristics of load resources of high-energy-consuming industry enterprises from multiple dimensions, providing a solid model foundation for precise aggregation and regulation, and meeting the complex aggregation and regulation requirements;

[0091] 2. In the modeling method of resource aggregation and control on the high-energy-consuming load side with distributed photovoltaic storage, the aggregation method based on horizontal federated learning is suitable for collaborative clustering of enterprises in various high-energy-consuming industries. It can meet the requirements of privacy protection and provide accurate clustering results. It provides a reliable method support for the reasonable aggregation of load resources of enterprises in high-energy-consuming industries, breaks the obstacles of information islands between enterprises to aggregation and control, and improves the feasibility and effectiveness of the aggregation and control of resources on the entire high-energy-consuming load side. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] Figure 1 is a flow chart of an exemplary regulatory modeling method in the present invention;

[0093] Figure 2 is an exemplary five-dimensional model structure diagram of the present invention;

[0094] Figure 3 The following is an exemplary clustering solution flow chart of the present invention. DETAILED DESCRIPTION

[0095] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0096] Example 1

[0097] like Figure 1-Figure 3 As shown, this embodiment provides a high-energy-consuming load-side resource aggregation control modeling method including distributed photovoltaic storage, including the following steps:

[0098] S1. Establish the regulation potential and regulation cost model of the load resources of high-energy-consuming industry enterprises with distributed photovoltaic storage: In view of the regulation needs of high-energy-consuming industry enterprises with distributed photovoltaic storage to participate in the county power grid, a five-dimensional model of "structure-geography-meteorology-power-regulation" of the load resources of high-energy-consuming industry enterprises is established;

[0099] Among them, the five-dimensional model of "structure-geography-climate-power-regulation" includes structure model, geography model, meteorology model, power model and regulation model, such as Figure 2 As shown, the establishment process is:

[0100] First, the structural model is established according to the internal composition and structural relationship of the load resources of high-energy-consuming industry enterprises; secondly, the geographical model and meteorological model are established according to their geographical information; then, according to their electrical mechanism, their output model, i.e., power model, is established; finally, according to their output model and combined with their regulation characteristics, their regulation model is established;

[0101] In this step, in the structural model, the load resource structure model of high-energy-consuming industry enterprises mainly includes photovoltaic installation methods and energy storage configuration in addition to the adjustable load equipment contained in itself. The structural model S i The definition is as follows:

[0102] First, photovoltaic installation method I i :Distributed photovoltaic installation methods are divided into ground / roof, wall, etc.; among them, ground / roof installation methods can be divided into fixed brackets, single-axis tracking, dual-axis tracking, etc. according to their bracket types, and wall installation methods also include east, west, south, etc.; the distributed photovoltaic installation method model can be encoded with a two-digit integer variable (that is, a specific two-digit integer is used to uniquely represent different installation method categories, so as to distinguish and calculate them in the model);

[0103] Second, energy storage configuration E i :The structural model of the distributed photovoltaic storage system needs to be differentiated according to whether it is configured with energy storage and the energy storage configuration method; the energy storage configuration method includes three different access methods: DC side, AC side, and independent type; when the DC side is configured, it is a photovoltaic storage integrated grid-connected mode; when the AC side is configured, photovoltaic and energy storage are two independent systems, but the AC side grid-connected point is the same; the independent type means that there is no other photovoltaic system at the AC side grid-connected point of the energy storage system; the energy storage configuration model can be encoded with a two-digit integer variable (the purpose is to distinguish different energy storage configuration situations).

[0104] Furthermore, in the geographical model, the load resource geographical model of high-energy-consuming industry enterprises refers to the geographical coordinates of the geometric center of the area where the enterprise is located, expressed by the longitude and latitude of the geometric center, that is:

[0105] G i =[L i ,B i ] (1)

[0106] In the formula, G i is the geographic model, L i is longitude, B i is latitude; these two parameters can be used to determine the location information of high-energy-consuming industry enterprises in geographic space, providing a basis for subsequent consideration of the impact of geographical factors on load resources. For example, climate, light and other conditions may be different in different geographical locations, which in turn affects the load characteristics of the enterprise (such as distributed photovoltaic power generation).

[0107] Furthermore, in the meteorological model, the load power consumption and distributed photovoltaic power generation in the load resources of high-energy-consuming industry enterprises are affected by meteorological factors, mainly including irradiance and temperature; therefore, the meteorological model can be expressed as:

[0108] Wi =[R i ,T i ] (2)

[0109] Where W i is the meteorological model, R i is the irradiance, T i For temperature, the measured data of the meteorological monitoring station in the area where the high-energy-consuming industry enterprise is located are taken. Irradiance and temperature are important meteorological factors that affect load power consumption (such as temperature affects the power load of equipment such as air conditioners) and distributed photovoltaic power generation (irradiance directly affects power generation). Using measured data can more accurately reflect the actual situation and make the model more in line with the actual operating status.

[0110] Furthermore, in the power model, the power model of the load resources of high-energy-consuming industry enterprises is mainly composed of an adjustable load model, a photovoltaic component model, an energy storage battery model, and a comprehensive efficiency model. The model is defined as follows:

[0111] First, the adjustable load model includes the adjustable power size constraint, the maximum duration constraint and the adjustable cost model. The specific algorithm is:

[0112] A. Adjust power constraints:

[0113]

[0114] In the formula, They represent the upward power and downward power of the adjustable load i in time period t respectively. They are variables whose reasonable ranges need to be determined through constraints, and reflect the change in the adjustable power of the load; The flags respectively indicate whether the adjustable load i in period t participates in the upward or downward regulation. A value of 1 indicates that it participates in the regulation, and a value of 0 indicates that it does not participate in the regulation. It is used to clarify whether the load participates in the power regulation operation in different periods. They represent the maximum upward and downward adjustment capabilities of the adjustable load i in time period t, respectively, which limits the upper limit of the load adjustment power and is the maximum adjustment range determined based on the characteristics of the load equipment itself and the operating conditions; They are the start and end times of the adjustable periods, respectively. There can be multiple adjustable periods, which are used to define the time ranges within which the load can be adjusted up or down, taking into account the different feasibility and needs of load adjustment in different periods in actual operation;

[0115] B. Maximum duration constraint:

[0116]

[0117] In the formula, [N s ,N e] is the start and end time of the adjustable up / down period (positive for upward adjustment and negative for downward adjustment), which determines the range of time accumulation calculation within a specific adjustable period; It is the flag bit of the adjustable load i participating in the upward / downward adjustment in period t (positive for upward adjustment and negative for downward adjustment). It is similar to the flag bit in the previous power size constraint adjustment, and is used to mark the load adjustment state in each period to calculate the duration; For the period [N′ s ,N′ e ] maximum duration within ;

[0118] It should be noted that [N′ s ,N′ e ] may contain multiple N s ,N e , which means considering the maximum duration limit of load regulation within a larger time interval to prevent excessive load regulation from affecting equipment life or normal operation;

[0119] C. Adjustment cost model:

[0120]

[0121] In the formula, They represent the unit price of upward adjustment and downward adjustment respectively, and are the key coefficients for calculating the adjustment cost, reflecting the cost difference caused by the adjustment of unit power; They represent the upward and downward adjustment of the adjustable load i in period t, which are related to the upward and downward adjustment power in the previous adjustment power size constraint and are the actual adjustment power change amount for calculating the cost; N dr Represents the number of adjustable loads, which is used to determine the total number of loads participating in the cost calculation.

[0122] Second, in the photovoltaic module model, the power generation of photovoltaic modules is mainly affected by the irradiance and temperature in the meteorological elements; the relationship between irradiance, temperature and power is as follows:

[0123]

[0124] Where P PV is the actual power, that is, the power generated by the photovoltaic module under the actual irradiance R and temperature T. It is a variable that needs to be calculated by the model, reflecting the actual power generation of the photovoltaic module; R is the actual irradiance, which is the key meteorological parameter affecting the power generation of the photovoltaic module. Its size directly determines the amount of light energy that the photovoltaic module can receive, thereby affecting the power generation; T is the actual temperature. The temperature has an impact on the power generation efficiency of the photovoltaic module. The temperature term in the formula reflects its influence on the power; P ref is the light irradiance under IEC standard conditions, which is 1000W / m2 ; T ref is the temperature under standard conditions, that is, 298K (25°C); P ref That is, the maximum output under standard conditions; these standard parameters are used to compare and convert the actual situation with the standard state, so as to accurately calculate the actual power under different irradiances and temperatures; k is the power temperature coefficient, which describes the rate at which power changes with temperature. Different photovoltaic modules have different power temperature coefficients, which reflects the sensitivity of photovoltaic modules to temperature changes.

[0125] Third, in the energy storage battery model, the energy storage battery model is mainly composed of the corresponding relationship and limit constraints between the charge and discharge power and its remaining capacity (state-of-charge, SOC); including:

[0126] a. Charge and discharge power constraints:

[0127]

[0128] In the formula, They are the maximum discharge power and maximum charging power of unit e, respectively, which limit the power upper limit of the energy storage battery during the discharge and charging process to prevent damage or safety problems caused by exceeding the rated power of the equipment; are the minimum discharge power and minimum charging power (in kW) of unit e, respectively, which determines the lower limit of the charging and discharging power of the energy storage battery, which may be related to the technical characteristics, safety requirements or operation strategy of the equipment; are the discharge power and charging power of unit e, respectively, and are the actual charging and discharging power variables of the energy storage battery at different times t, and their magnitudes must meet the above power constraints; They are discharge state and charge state respectively. When the value is 1, it means discharge / charge, and when it is 0, it means no discharge / no charge. It is used to clarify the working state of the energy storage battery at each moment (charging, discharging or idle) so as to accurately apply power constraints.

[0129] b. Charge and discharge state constraints: The charge and discharge states are mutually exclusive, that is, the energy storage battery cannot be charged and discharged at the same time, to ensure the uniqueness and rationality of the working state of the energy storage battery, that is, to meet the following constraints:

[0130]

[0131] At the same time, the following state transition constraints are met:

[0132]

[0133] In the formula, represents the discharge state of unit e at time t+1, They are the start-up discharge state and the start-up charge state, respectively. When the value is 1, it means that the discharge / charge is started at the current moment, and when it is 0, it means that no action is taken at the current moment. Formula (9) describes the transfer relationship between the discharge states of the energy storage battery at different moments, taking into account the influence of operations such as start-up discharge and stop-discharge on the state change. Formula (10) further limits the rationality of the start-up charge and discharge operation, prevents the simultaneous start-up of discharge and charge operations, and ensures the correctness of the operation logic of the energy storage battery.

[0134] c.SOC constraint:

[0135]

[0136] In the formula, S e (t) is the SOC of unit e in period t, which is a key variable reflecting the state of charge of the energy storage battery. Its value changes with the charging and discharging process and is an important indicator for the operation management and control of the energy storage battery. e (t-1) is the SOC of unit e in the t-1 period, which is used to calculate the change of SOC in a period and reflects the dynamic change process of SOC; is the active power of unit e in period t, taking into account the impact of charging and discharging power on SOC. When discharging, positive power will reduce SOC, and when charging, negative power will increase SOC; δ t The duration of a period (unit: hour), which is used to convert the power and SOC changes on a time scale to make the calculation consistent in the time dimension; They are the maximum and minimum allowable values ​​of SOC, which limit the reasonable range of SOC and prevent SOC from being too high (which may cause problems such as overcharging) or too low (which may affect battery life and performance); S 0 The initial capacity of SOC determines the charge state of the energy storage battery at the initial moment and is the starting point of the entire SOC calculation process. e (0), S e (T) represents the SOC at the start and end time of an operation cycle t∈[0,T] respectively;

[0137] d. Continuous charge and discharge constraints: In order to avoid frequent switching of energy storage charge and discharge states, charging is only performed when the SOC is below a certain threshold, and discharging is performed when the SOC is above a certain threshold, that is, the following constraints are met:

[0138]

[0139] In the formula, They are the allowed charging threshold and the allowed discharging threshold, respectively, which are used to control the charging and discharging operation timing of the energy storage battery to avoid frequent switching of the charging and discharging states; for example, charging is performed when the SOC is lower than the allowed charging threshold, and discharging is performed when the SOC is higher than the allowed discharging threshold, so as to extend the service life of the energy storage battery and improve the operating efficiency.

[0140] Fourth, in the comprehensive efficiency model, the comprehensive efficiency model of load resources of high-energy-consuming industry enterprises takes into account the comprehensive efficiency of various links such as photovoltaic module efficiency, battery charging and discharging efficiency, inverter efficiency, and distribution system efficiency:

[0141]

[0142] Where P i P is the total power of the high-energy-consuming industry enterprise at the time after comprehensively considering multiple factors such as photovoltaic, energy storage, and load. It is a key variable reflecting the overall power balance and operating status of the enterprise. LD (t) is the power load power of the high energy consumption industry enterprise, η 1 , η 2 , η 3 , η 4 They are the comprehensive efficiency coefficients of photovoltaic modules and their inverters, energy storage discharge, energy storage charging, and distribution systems. They are used to consider the efficiency losses in each link during the energy conversion and transmission process, and to accurately calculate the actual contribution of each part of the power to the total power at the gateway by multiplying the corresponding power terms.

[0143] Fifth, in the control model, the control model of load resources of high-energy-consuming industry enterprises refers to the upper and lower limits of the active control response capability of the superior control center and the corresponding response cost. The model is defined as follows:

[0144] 1) Upper and lower limits of response capability, including:

[0145] a. Upper limit of response capacity: The upper limit of load resource response capacity of enterprises in high energy consumption industries is determined by the maximum downward adjustment power of the adjustable load and the increaseable discharge power of the energy storage system, that is:

[0146]

[0147] b. Lower limit of response capability: The lower limit of the load resource response capability of enterprises in high energy consumption industries is determined by the maximum increase power of the adjustable load, the real-time output of the photovoltaic system and the increaseable charging power of the energy storage system, that is:

[0148]

[0149] In the formula, They are the upper and lower limits of the load resource response capabilities of enterprises in high energy-consuming industries;

[0150] in, is the upper limit of the load resource response capacity of enterprises in high-energy-consuming industries, that is, the maximum capacity of enterprises to adjust power upward at time t, taking into account the maximum downward adjustment power of adjustable loads ( Here, the power reduction is used to reversely calculate the upper limit of the upward adjustment capacity, which can be understood as the released load power can be used for upward adjustment), the increaseable discharge power of the energy storage system ( That is, the difference between the maximum discharge power and the current discharge power, indicating that the discharge power that can be increased can be used for upward adjustment) and the increaseable charging power of the energy storage system ( Charging power can be used to balance system power and is indirectly considered as part of the upward regulation capability);

[0151] is the lower limit of the load resource response capacity of enterprises in high-energy-consuming industries, that is, the minimum capacity of enterprises to adjust power downward at time t, taking into account the maximum upward power of adjustable loads The real-time output of the photovoltaic system (P PV (t)), the increaseable charging power of the energy storage system ( That is, the difference between the maximum charging power and the current charging power, indicating that the additional charging power can be used to absorb excess power and is regarded as part of the downward adjustment capacity) and the increaseable discharge power of the energy storage system ( The discharge power can be used to balance the system power and is indirectly considered as part of the down-regulation capability).

[0152] 2) Response costs, including:

[0153] a. Response cost of increasing power: The response cost of increasing power of load resources in high-energy-consuming industries mainly comes from energy storage output and adjustable load reduction. Photovoltaic does not have the ability to increase output. Therefore, the response cost of increasing power is expressed as:

[0154]

[0155] b. Downward response cost: The response cost of the load resources of high-energy-consuming industry enterprises to reduce power is mainly due to the additional cost caused by the increase in electricity consumption of adjustable loads and the power generation loss caused by the reduction of photovoltaic output. There is no additional cost for energy storage to reduce output. Therefore, the downward response cost is expressed as:

[0156]

[0157] In the formula, They are the response costs of increasing and decreasing the power of load resources in high energy-consuming industries; is the electricity cost of the energy storage system; is the on-grid electricity price of the photovoltaic system; ΔP PV- (t) is the abandoned power of the PV system; Respectively indicate an increase or decrease in unit price; They respectively represent the upward and downward adjustment amounts of the adjustable load i in period t.

[0158] S2. Propose a method for aggregating load resources of enterprises in high-energy-consuming industries with distributed photovoltaic storage based on horizontal federated learning; including:

[0159] S2.1. Alignment of encrypted entities: First, align the encrypted load-side resource entities of high-energy-consuming enterprises held by the power company and various high-energy-consuming enterprises, that is, find the load-side resource entities shared by the power company and various high-energy-consuming enterprises in the region;

[0160] S2.2, Feature data preprocessing: Secondly, the feature vector data of high energy-consuming industry enterprises shared by both parties are preprocessed, entities containing invalid data are eliminated, and valid data are normalized;

[0161] Among them, the algorithm for normalizing valid data is:

[0162]

[0163] In the formula, x i represents a characteristic quantity data provided by power companies or high-energy-consuming enterprises, x max and x min Respectively represent the maximum and minimum values ​​of the feature quantity data.

[0164] S2.3, Task division: Each high-energy-consuming industry enterprise uses its own structural model, geographical model, meteorological model, power model, regulation model and other characteristic data to cluster; the power company plays the role of a collaborator, responsible for determining the number of aggregates, initializing the center of the aggregate, and summarizing and updating the combination of model parameters;

[0165] S2.4, Clustering solution: Figure 3 As shown, according to the task division of step S2.3, the specific process of clustering load resources of high-energy-consuming industry enterprises includes the following steps:

[0166] S2.4.1. Initialize the clustering result error threshold to δ;

[0167] S2.4.2. Initialize the number of aggregates: The power company initially defines the number of aggregates, M;

[0168] S2.4.3. Initialize cluster centers: Each high-energy-consuming industry enterprise initializes a set of cluster centers based on the entity and feature data it has. The initialization cluster center of high-energy-consuming industry enterprise j is:

[0169] c jk=[S i ,W i ,P i ,ΔP i ,f i ] (19)

[0170]

[0171] In the formula, k is the number of the cluster center, k∈Z M , Z M represents a set of integers from 1 to M, j is the number of enterprises in high energy consumption industries, j∈Z L , L is the number of enterprises in high-energy-consuming industries participating in horizontal federated learning; S i is the structural model, W i is the meteorological model, P i is the power model, ΔP i is the control model, f i To regulate response costs; is the average value of each cluster center point;

[0172] After each high-energy-consuming industry enterprise initializes the cluster center, it is encrypted and uniformly submitted to the partner power company for aggregation. The power company calculates the average value of each cluster center point. As the final initialization cluster center, it is sent to each participant;

[0173] S2.4.4, separate clustering calculation: After receiving the cluster center issued by the power company, each participant decrypts it to obtain the current cluster center, traverses the load resources they have, and calculates the L2 paradigm distance from each high-energy-consuming industry enterprise to each cluster center; the L2 paradigm distance d from high-energy-consuming industry enterprise i to the kth cluster center ik The calculation is as follows:

[0174]

[0175] Where M i is the load regulation model for the i-th high energy consumption industry enterprise;

[0176] Determine the cluster to which each high-energy-consuming industry enterprise belongs based on the L2 paradigm distance, that is, find the cluster center point closest to the high-energy-consuming industry enterprise, and add the high-energy-consuming industry enterprise to the cluster represented by the cluster center point, so as to achieve the update of the cluster;

[0177] S2.4.5. Update cluster centers separately: Each participant calculates the average value of the characteristic vectors of all high-energy-consuming industry enterprises in each aggregate for the updated aggregate, and uses it as the updated cluster center of the aggregate. After encryption, it is uniformly submitted to the partner power company for summary update;

[0178] S2.4.6. The power company summarizes and updates the cluster center: After receiving the updated cluster centers submitted by each participant, the power company calculates the average value of each new cluster center as the final updated cluster center and sends it to each participant;

[0179] S2.4.7, iteratively execute steps S2.4.4 to S2.4.6 until the average distance between the feature vectors of the M cluster centers obtained from the two previous iterations is less than a certain threshold, that is, the clustering process has converged and the cluster centers no longer change; if the maximum number of iterations is reached and there is still no convergence, continue to step S2.4.8;

[0180] S2.4.8, modify the number of aggregates and repeat steps S2.4.3 to S2.4.7; if the results of step S2.4.7 do not converge after multiple attempts at the number of aggregates, you can relax the threshold, that is, return to step S2.4.1 to modify the clustering result error threshold, and then repeat steps S2.4.2 to S2.4.8 until a convergent result is obtained.

[0181] This technical solution aims at the problem of how to aggregate the load resources of high-energy-consuming industry enterprises with distributed photovoltaic storage to participate in the hierarchical and regional regulation of county power grids under privacy protection, and proposes an aggregated regulation modeling method based on load regulation model and horizontal federated learning. Firstly, a method combining mechanism model and data-driven is adopted to establish the regulation potential and regulation cost model of the load resources of high-energy-consuming industry enterprises with distributed photovoltaic storage. Secondly, based on the load regulation model, a method for aggregating the load resources of high-energy-consuming industry enterprises with distributed photovoltaic storage based on horizontal federated learning is proposed, which is of great significance for high-energy-consuming industry enterprises to participate in the regulation of county power grids under privacy protection.

[0182] This technical solution can accurately characterize physical properties, including:

[0183] 1) The constructed five-dimensional model of "structure-geography-climate-power-regulation" comprehensively and deeply covers the key elements of load resources of enterprises in high-energy-consuming industries; in the structural model, the photovoltaic installation methods (such as different types of ground / roof, wall and bracket differences) and energy storage configuration (different access methods) are considered in detail, and the layout and connection methods of energy equipment within the enterprise are accurately described, laying the foundation for analyzing its operating characteristics;

[0184] 2) The geographic model locates the longitude and latitude of the area where the enterprise is located, combined with the irradiance and temperature data in the meteorological model (both taken from the actual measurements of the regional meteorological monitoring station), which can accurately reflect the differences in the natural environment affecting enterprises in different geographical locations, and then reflect the changing laws of their load resources under the influence of meteorological factors. For example, different light and temperature in different regions lead to different photovoltaic power generation and load power demand;

[0185] 3) Power model Comprehensive adjustable load model (including adjustable power size constraints, maximum duration constraints and adjustment cost model), photovoltaic module model (considering the impact of irradiance and temperature on power generation), energy storage battery model (involving multiple constraints such as charging and discharging power, SOC, etc.) and comprehensive efficiency model (considering the efficiency of each link), meticulously depicts the various characteristics of enterprise load resources at the power level, including load regulation capability, power generation change, energy storage charging and discharging characteristics, and overall energy conversion and transmission efficiency;

[0186] 4) The control model clarifies the upper and lower limits of the enterprise's active control response capability and the response cost to the superior control center, so that the behavior and capabilities of the enterprise in power grid control can be accurately quantified. In summary, the five-dimensional model cooperates with each other to accurately characterize the real physical characteristics of the load resources of high-energy-consuming industry enterprises from multiple dimensions, providing a solid model foundation for precise aggregate control and meeting complex control requirements.

[0187] It is also possible to achieve collaborative clustering under privacy protection, including:

[0188] 1) The aggregation method based on horizontal federated learning provides an effective collaborative mechanism for enterprises in various high-energy-consuming industries, while fully protecting the privacy of enterprise data. In the encrypted entity alignment stage, encryption technology is used to ensure the security of data when enterprises and power companies are looking for common load-side resource entities, avoiding the leakage of sensitive information.

[0189] 2) In the feature data preprocessing phase, the common entity data is normalized to make the data comparable without exposing the specific values ​​of the original data, thus providing a suitable data format for subsequent cluster analysis;

[0190] 3) In the task division, enterprises use their own data for clustering, and power companies are responsible for determining and summarizing key parameters. Each party performs its own duties, and the data is processed locally and the clustering process is completed through limited encrypted interactions.

[0191] 4) In the whole clustering solution process, the cluster center is calculated iteratively for many times, and the necessary information is exchanged between enterprises in an encrypted manner, finally realizing the accurate division of the aggregate. This method not only makes full use of the local data resources of each enterprise, but also avoids direct sharing of data, effectively protects the privacy of enterprises, and can provide accurate clustering results, providing a reliable method support for the reasonable aggregation of load resources of enterprises in high-energy-consuming industries, breaking the obstacles of information islands between enterprises to aggregation regulation, and improving the feasibility and effectiveness of the aggregation regulation of resources on the entire high-energy-consuming load side.

[0192] Those skilled in the art will appreciate that the process of implementing all or part of the steps of the above embodiments may be completed by hardware, or may be completed by instructing related hardware through a program.

[0193] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A modeling method for resource aggregation and control of high energy consumption load side with distributed photovoltaic storage, characterized in that: The steps include: S1. Establish the regulation potential and regulation cost model of the load resources of high-energy-consuming enterprises with distributed photovoltaic storage: In view of the demand for the aggregation of load resources of high-energy-consuming enterprises with distributed photovoltaic storage to participate in the regulation of county power grids, establish a "structure-geography-meteorology-power-regulation" five-dimensional model of the load resources of high-energy-consuming enterprises; Among them, the "structure-geography-climate-power-regulation" five-dimensional model includes a structure model, a geography model, a meteorology model, a power model, and a regulation model. Its establishment process is as follows: First, the structural model is established according to the internal composition and structural relationship of the load resources of high-energy-consuming industry enterprises; secondly, the geographical model and meteorological model are established according to their geographical information; then, according to their electrical mechanism, their output model, i.e., power model, is established; finally, according to their output model and combined with their regulation characteristics, their regulation model is established; S2. Propose a method for aggregating load resources of enterprises in high-energy-consuming industries with distributed photovoltaic storage based on horizontal federated learning; including: S2.

1. Alignment of encrypted entities: Align the encrypted load-side resource entities of high-energy-consuming enterprises held by the power company and the high-energy-consuming enterprises, that is, find the load-side resource entities shared by the power company and the high-energy-consuming enterprises in the region; S2.

2. Feature data preprocessing: preprocess the feature vector data of high-energy-consuming industry enterprises shared by both parties, remove entities containing invalid data, and normalize the valid data; S2.3, Task division: Each high-energy-consuming industry enterprise uses the characteristic data of its own structural model, geographical model, meteorological model, power model, and control model to cluster; the power company acts as a collaborator and is responsible for determining the number of aggregates, initializing the center of the aggregate, and summarizing and updating the combination of model parameters; S2.4, clustering solution: According to the task division in step S2.3, enterprises in high-energy-consuming industries cluster load resources.

2. The resource aggregation control modeling method for high energy consumption load side with distributed photovoltaic storage according to claim 1 is characterized in that: In the structural model, the load resource structure model of high-energy-consuming industry enterprises mainly includes photovoltaic installation methods and energy storage configuration in addition to the adjustable load equipment contained in itself. The structural model S i The definition is as follows: First, photovoltaic installation method I i :Distributed photovoltaic installation methods are divided into ground / roof and wall; among them, ground / roof installation methods can be divided into fixed brackets, single-axis tracking, and dual-axis tracking according to their bracket types, and wall installation methods also include east, west, and south. Second, energy storage configuration E i :The structural model of the distributed photovoltaic storage system needs to be differentiated according to whether it is configured with energy storage and the energy storage configuration method; the energy storage configuration method includes three different access methods: DC side, AC side, and independent type; when the DC side is configured, it is a photovoltaic storage integrated grid-connected mode; when the AC side is configured, photovoltaic and energy storage are two independent systems, but the AC side grid connection point is the same; the independent type refers to a photovoltaic system with no other AC side grid connection point of the energy storage system.

3. The resource aggregation control modeling method for high energy consumption load side with distributed photovoltaic storage according to claim 2 is characterized in that: In the geographic model, the load resource geographic model of high-energy-consuming industry enterprises refers to the geographic coordinates of the geometric center of the area where the enterprise is located, expressed by the longitude and latitude of the geometric center, that is: G i =[L i ,B i ] (1) In the formula, G i is the geographic model, L i is longitude, B i is latitude; In the meteorological model, the load power consumption and distributed photovoltaic power generation in the load resources of high-energy-consuming industry enterprises are affected by meteorological factors, mainly including irradiance and temperature; therefore, the meteorological model can be expressed as: W i =[R i ,T i ] (2) Where W i is the meteorological model, R i is the irradiance, T i For temperature, the measured data are taken from the meteorological monitoring station in the area where the high energy-consuming industry enterprise is located.

4. The resource aggregation control modeling method for high energy consumption load side with distributed photovoltaic storage according to claim 3 is characterized by: In the power model, the power model of the load resources of high-energy-consuming industry enterprises is mainly composed of an adjustable load model, a photovoltaic component model, an energy storage battery model, and a comprehensive efficiency model. The model is defined as follows: The adjustable load model includes an adjustable power size constraint, a maximum duration constraint and an adjustable cost model, and the specific algorithm is: A. Adjust power constraints: In the formula, They represent the upward power and downward power of the adjustable load i in time period t respectively; The flags respectively indicate whether the adjustable load i in period t participates in the upward or downward regulation. A value of 1 indicates that it participates in the regulation, and a value of 0 indicates that it does not participate in the regulation. They represent the maximum upward and downward adjustment capabilities of the adjustable load i during period t respectively; They are the start and end times of the adjustable upward and downward time periods, and there can be multiple adjustable time periods; B. Maximum duration constraint: In the formula, [N s ,N e ] is the start and end time of the period that can be adjusted upward / downward, with upward adjustment being positive and downward adjustment being negative; It is the flag bit of the adjustable load i participating in the upward / downward adjustment in period t, which is positive for upward adjustment and negative for downward adjustment; For the period [N′ s ,N′ e ] within the maximum duration; C. Adjustment cost model: In the formula, Respectively indicate an increase or decrease in unit price; They represent the upward and downward adjustment of the adjustable load i in period t respectively; N dr Indicates the number of adjustable loads.

5. The resource aggregation control modeling method for high energy consumption load side with distributed photovoltaic storage according to claim 4 is characterized in that: In the photovoltaic module model, the power generation of photovoltaic modules is mainly affected by the irradiance and temperature in the meteorological elements; the relationship between irradiance, temperature and power is as follows: Where P PV is the actual power, R is the actual irradiance; T is the actual temperature; R ref is the light irradiance under IEC standard conditions; T ref is the temperature under standard conditions; P ref That is the maximum output under standard conditions; k is the power temperature coefficient.

6. The resource aggregation control modeling method for high energy consumption load side with distributed photovoltaic storage according to claim 5 is characterized in that: In the energy storage battery model, the energy storage battery model is mainly composed of the corresponding relationship and limit constraints between the charge and discharge power and its remaining capacity SOC; including: a. Charge and discharge power constraints: In the formula, are the maximum discharge power and maximum charging power of unit e respectively; are the minimum discharge power and minimum charging power of unit e respectively; are the discharge power and charging power of unit e respectively; They are respectively the discharge state and the charge state. When the value is 1, it means discharge / charge, and when it is 0, it means no discharge / no charge; b. Charge and discharge state constraints: The charge and discharge states are mutually exclusive, that is, they satisfy the following constraints: At the same time, the following state transition constraints are met: In the formula, represents the discharge state of unit e at time t+1, They are respectively the start-discharge state and the start-charge state. When the value is 1, it means that the discharge / charge is started at the current moment, and when it is 0, it means that there is no action at the current moment. c.SOC constraint: In the formula, S e (t) is the SOC of unit e in period t; S e (t-1) is the SOC of unit e in the period t-1; is the active power of unit e in period t; t is the duration of a period; are the maximum and minimum allowable values ​​of SOC respectively; S0 is the initial capacity of SOC; S e (0), S e (T) represents the SOC at the start and end time of an operation cycle t∈[0,T] respectively; d. Continuous charge and discharge constraints: In order to avoid frequent switching of energy storage charge and discharge states, charging is only performed when the SOC is below a certain threshold, and discharging is performed when the SOC is above a certain threshold, that is, the following constraints are met: In the formula, They are respectively the charge threshold and the discharge threshold.

7. The resource aggregation control modeling method for high energy consumption load side with distributed photovoltaic storage according to claim 6 is characterized by: In the comprehensive efficiency model, the comprehensive efficiency model of load resources of high-energy-consuming industry enterprises takes into account the comprehensive efficiency of photovoltaic module efficiency, battery charging and discharging efficiency, inverter efficiency, and distribution system efficiency: Where P i is the total power of the high energy consumption industry, P LD (t) is the power load power of the high energy consumption industry enterprise, η1, η2, η3, η4 are the comprehensive efficiency coefficients of photovoltaic modules and their inverters, energy storage discharge, energy storage charging, and distribution systems respectively.

8. The resource aggregation control modeling method for high energy consumption load side with distributed photovoltaic storage according to claim 7 is characterized in that: In the control model, the control model of load resources of high-energy-consuming industry enterprises refers to the upper and lower limits of the active control response capability of the superior control center and the corresponding response cost. The model is defined as follows: 1) Upper and lower limits of response capability, including: a. Upper limit of response capacity: The upper limit of load resource response capacity of enterprises in high energy consumption industries is determined by the maximum downward adjustment power of the adjustable load and the increaseable discharge power of the energy storage system, that is: b. Lower limit of response capability: The lower limit of the load resource response capability of enterprises in high energy consumption industries is determined by the maximum increase power of the adjustable load, the real-time output of the photovoltaic system and the increaseable charging power of the energy storage system, that is: In the formula, They are the upper and lower limits of the load resource response capabilities of enterprises in high energy-consuming industries; 2) Response costs, including: a. Response cost of increasing power: The response cost of increasing power of load resources in high-energy-consuming industries mainly comes from energy storage output and adjustable load reduction. Photovoltaic does not have the ability to increase output. Therefore, the response cost of increasing power is expressed as: b. Downward response cost: The response cost of the load resources of high-energy-consuming industry enterprises to reduce power is mainly due to the additional cost caused by the increase in electricity consumption of adjustable loads and the power generation loss caused by the reduction of photovoltaic output. There is no additional cost for energy storage to reduce output. Therefore, the downward response cost is expressed as: In the formula, They are the response costs of increasing and decreasing the power of load resources in high energy-consuming industries; is the electricity cost of the energy storage system; is the on-grid electricity price of the photovoltaic system; ΔP PV- (t) is the abandoned power of the PV system; Respectively indicate an increase or decrease in unit price; They respectively represent the upward and downward adjustment amounts of the adjustable load i in period t.

9. The resource aggregation control modeling method for high energy consumption load side with distributed photovoltaic storage according to claim 1 is characterized by: In step S2.2, the algorithm for normalizing the valid data is: In the formula, x i represents a characteristic quantity data provided by power companies or high-energy-consuming enterprises, x max and x min Respectively represent the maximum and minimum values ​​of the feature quantity data.

10. The resource aggregation control modeling method for high energy consumption load side with distributed photovoltaic storage according to claim 8 is characterized in that: In step S2.4, the specific process of clustering load resources of enterprises in high energy consumption industries includes the following steps: S2.4.

1. Initialize the clustering result error threshold to δ; S2.4.

2. Initialize the number of aggregates: The power company initially defines the number of aggregates, M; S2.4.

3. Initialize cluster centers: Each high-energy-consuming industry enterprise initializes a set of cluster centers based on the entity and feature data it has. The initialization cluster center of high-energy-consuming industry enterprise j is: c jk =[S i ,W i ,P i ,ΔP i ,f i ] (19) In the formula, k is the number of the cluster center, k∈Z M , Z M represents a set of integers from 1 to M, j is the number of enterprises in high energy consumption industries, j∈Z L , L is the number of enterprises in high-energy-consuming industries participating in horizontal federated learning; S i is the structural model, W i is the meteorological model, P i is the power model, ΔP i is the control model, f i To regulate response costs; is the average value of each cluster center point; After each high-energy-consuming industry enterprise initializes the cluster center, it is encrypted and uniformly submitted to the partner power company for aggregation. The power company calculates the average value of each cluster center point. As the final initialization cluster center, it is sent to each participant; S2.4.4, separate clustering calculation: After receiving the cluster center issued by the power company, each participant decrypts it to obtain the current cluster center, traverses the load resources they have, and calculates the L2 paradigm distance from each high-energy-consuming industry enterprise to each cluster center; the L2 paradigm distance d from high-energy-consuming industry enterprise i to the kth cluster center ik The calculation is as follows: Where M i is the load regulation model for the i-th high energy consumption industry enterprise; Determine the cluster to which each high-energy-consuming industry enterprise belongs based on the L2 paradigm distance, that is, find the cluster center point closest to the high-energy-consuming industry enterprise, and add the high-energy-consuming industry enterprise to the cluster represented by the cluster center point, so as to achieve the update of the cluster; S2.4.

5. Update cluster centers separately: Each participant calculates the average value of the characteristic vectors of all high-energy-consuming industry enterprises in each aggregate for the updated aggregate, and uses it as the updated cluster center of the aggregate. After encryption, it is uniformly submitted to the partner power company for summary update; S2.4.

6. The power company summarizes and updates the cluster center: After receiving the updated cluster centers submitted by each participant, the power company calculates the average value of each new cluster center as the final updated cluster center and sends it to each participant; S2.4.7, iteratively execute steps S2.4.4 to S2.4.6 until the average distance between the feature vectors of the M cluster centers obtained from the two previous iterations is less than a certain threshold, that is, the clustering process has converged and the cluster centers no longer change; if the maximum number of iterations is reached and there is still no convergence, continue to step S2.4.8; S2.4.8, modify the number of aggregates and repeat steps S2.4.3 to S2.4.7; if the results of step S2.4.7 do not converge after multiple attempts at the number of aggregates, you can relax the threshold, that is, return to step S2.4.1 to modify the clustering result error threshold, and then repeat steps S2.4.2 to S2.4.8 until a convergent result is obtained.

Images