Industrial park hierarchical response method and system based on federal alternating direction multiplier method

Abstract

The present application belongs to the technical field of demand response strategy, and provides an industrial park hierarchical response method and system based on a federal alternating direction multiplier method. An industrial park hierarchical response model is constructed, which takes into account the interests of load aggregators and industrial users and process flows. The industrial park hierarchical response model includes a load aggregator layer and an industrial user layer. The load aggregator layer optimizes a user-level response scheme with the minimum cost as the target. The industrial user layer optimizes an equipment-level response scheme with the maximum additional income as the target. Based on the industrial park hierarchical response model, a federal alternating direction multiplier method distributed algorithm is constructed. The industrial park hierarchical response model is solved by the federal alternating direction multiplier method distributed algorithm. The load aggregator layer and the industrial user layer dynamically update the compensation price until the optimal solution is converged. The sensitive data protection and the calculation efficiency can be improved, and a demand response scheme that takes into account the interests of multiple subjects and process flows can be obtained.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of demand response strategy, and particularly relates to an industrial park hierarchical response method and system based on a federal alternating direction multiplier method. BACKGROUND

[0002] With large-scale photovoltaic and wind turbine grid connection, the power system is affected by the uncertainty and volatility brought by new energy access. Demand response, as a demand side management technology, is an effective method to smooth the intermittent power fluctuations brought by new energy. Existing demand response includes price-based demand response and incentive-based demand response: price-based demand response guides users to adjust power demand by changing the price signal, including time-of-use electricity price, real-time electricity price and critical peak electricity price; in incentive-based demand response, load aggregators sign agreements with users, send response schemes to users according to grid demand and provide economic compensation afterwards. Industrial users not only can provide large response capacity, but also have high response reliability, and are the main participants of incentive-based demand response.

[0003] At present, there have been many studies on industrial users participating in demand response, mainly including cement, steel and electrolytic aluminum industries. However, when the existing research establishes a model of industrial users participating in demand response, it does not fully consider the constraints of industrial production processes and upstream and downstream intermediate products; when formulating demand response strategies, it only takes the economic benefits of industrial users as the target, ignoring the benefits of other stakeholders in demand response projects. For the optimal strategy solving problem of demand response participated by industrial users, the existing research mostly adopts centralized optimization method. However, incentive-based demand response projects are often carried out through load aggregators, and the centralized solving method requires industrial users to provide all data of production equipment to load aggregators, but their sensitive data may affect the enthusiasm of users participating in demand response. SUMMARY

[0004] In view of the deficiencies of the prior art, the application provides an industrial park hierarchical response method and system based on a federal alternating direction multiplier method, which aims to solve the problems mentioned in the background.

[0005] In a first aspect, the application provides an industrial park hierarchical response method based on a federal alternating direction multiplier method, including the following steps:

[0006] Step S1: Construct an industrial park hierarchical response model considering the interests of load aggregators and industrial users and process flow, the industrial park hierarchical response model including a load aggregator layer and an industrial user layer, the load aggregator layer optimizing a user-level response scheme with a minimum cost as an objective, and the industrial user layer optimizing an equipment-level response scheme with a maximum additional income as an objective; wherein the load aggregator layer includes a load aggregator platform responsible for the industrial park and industrial users of different industries participating in the response project in the industrial park, and the industrial user layer includes industrial equipment of different industries with adjustable loads;

[0007] Step S2: Based on the industrial park hierarchical response model, a federated alternating direction multiplier method distributed algorithm is constructed, and the global parameter updating, transmission information encryption and client iteration rules are improved to improve the sensitive data protection and calculation efficiency in the iteration process.

[0008] Step S3: The industrial park hierarchical response model is solved by the federated alternating direction multiplier method distributed algorithm, and the load aggregator layer and the industrial user layer dynamically update the compensation price until the optimal solution is converged, so as to obtain a response scheme considering the interests of load aggregators and industrial users.

[0009] Further, in step S1, when the industrial park hierarchical response model considering the interests of load aggregators and industrial users and process flow is used for an incentive demand response project, the load aggregator layer provides a user-level response scheme of the industrial park it is responsible for according to the requirements of the power grid, and economically compensates the industrial users participating in the response project, and the industrial user layer adjusts the production plan of the equipment according to the user-level response scheme to meet the requirements of the power grid.

[0010] Specifically, before formulating the user-level response scheme, the industrial user layer reports response characteristic parameters to the load aggregator layer, the response characteristic parameters including the maximum capacity of the adjustable load of the industrial user, the response duration, the response period and the acceptable compensation price range; then the load aggregator layer collects the response characteristic parameters reported by the industrial user layer, and based on the typical demand response scenarios investigated, constructs a user-level aggregation optimization model, and then optimizes and solves it, in the optimization and solving process, the industrial users participating in the response project are divided into several groups, one group of industrial users is an aggregation, the user response of each aggregation obtained is taken as the user-level response scheme, then the user-level response scheme is provided to the industrial users participating in the response project, and the industrial users participating in the response project are economically compensated, and the industrial user layer constructs an equipment-level response optimization model according to the user-level response scheme and its own characteristics, and solves it to obtain the response scheme of the industrial equipment with adjustable load.

[0011] Further, the user-level aggregation optimization model is constructed and then optimized and solved, specifically:

[0012] A user-level aggregation optimization model is established, which takes into account the interests of both the load aggregator and the industrial users and the process flow. The load aggregator layer collects the response characteristic parameters reported by the industrial user layer. Based on a typical demand response scenario, the key parameters of the user-level aggregation optimization model are determined. The user-level aggregation decision variable of the user-level aggregation optimization model is the aggregation of each industrial user, the response of the unit response time and the actual response capacity.

[0013] Setting 、 、 as decision variables. denotes the response state of the industrial user in the unit response time in the aggregation . denotes the starting response state of the industrial user in the unit response time in the aggregation . denotes the response capacity of the industrial user in the unit response time in the aggregation .

[0014] The user-level aggregation optimization model aims to minimize the total response compensation cost of the aggregation, as shown in equation (1):

[0015] (1);

[0016] In the formula: denotes the total response compensation cost of the user-level aggregation; denotes the weight factor of the aggregation , which is set according to the demand response scenario; denotes the compensation price given by the load aggregator to the industrial user , denotes that the compensation price is given by the load aggregator; denotes the total number of aggregations . denotes the total number of industrial users . denotes the total number of unit response times .

[0017] According to the distribution of the industrial user response period and the typical power grid demand response period, the constraint as shown in equation (2) is established to limit the industrial users to participate in aggregation only in their response period:

[0018] (2);

[0019] In the formula: Indicates industrial users Unit response time Can I participate in the response? At that time, the industrial user can participate in the response within the unit's response time; otherwise, they cannot respond.

[0020] When industrial users are assigned to participate in the response within an aggregate, the actual response capacity cannot exceed its maximum response capacity limit, and the aggregate must meet the response capacity constraints, as shown in formulas (3) and (4):

[0021] (3);

[0022] (4);

[0023] In the formula: Indicates industrial users Maximum response capacity; Represents an aggregate The target response capacity.

[0024] Furthermore, a device-level response optimization model is constructed and solved to obtain the response scheme of industrial equipment with adjustable load, including the following steps:

[0025] A response optimization model for fused magnesium equipment is established that balances the interests of multiple stakeholders, including load aggregators and industrial users, while also considering safety constraints. The constraints in the equipment-level response optimization scheduling process of the fused magnesium equipment-level response optimization model are as follows:

[0026] (i) The power of the magnesium melting electric arc furnace equipment is controlled within a safe range during the production process;

[0027] (ii) The output of the magnesium melting electric arc furnace equipment remains unchanged before and after participating in demand response;

[0028] Additional benefits that fused magnesium users gain from participating in demand response include: changes in electricity costs, equipment adjustment costs, changes in production efficiency, and demand response benefits.

[0029] Furthermore, constructing and solving a device-level response optimization model to obtain the response scheme for industrial equipment with adjustable loads also includes the following steps:

[0030] A cement equipment-level polymerization optimization model is established that takes into account the interests of multiple stakeholders, including load aggregators and industrial users, and considers the process flow. The complete production line of a cement plant is divided into four sub-processes: crushing process, kiln feed preparation process, clinker production process, and grinding process. The main production equipment in the crushing process, kiln feed preparation process, clinker production process, and grinding process are crusher, raw meal mill, clinker mill, and cement mill, respectively. Storage bins are set up between every two sub-processes.

[0031] When participating in demand response, the cement plant changes the power of the main production equipment of the sub-process to provide flexible load;

[0032] The main production equipment power of the cement plant, the bucket storage amount and the relationship between storage and production, the constraints are shown in equations (12)-(15):

[0033] (12);

[0034] (13);

[0035] (14);

[0036] (15);

[0037] In the formula: represents the total power of the cement plant in unit response time running, represents the cement; represents the main production equipment of the sub-process in unit response time consumption of rated power per unit output; represents the power of the main production equipment of the sub-process in unit response time after participating in response; represents the production rate of the sub-process in unit response time ; represents the supply rate of the bucket storage of the sub-process in unit response time to the next sub-process ; represents the ratio of the input material amount and the output material amount of the sub-process ; represents the storage material amount of the sub-process in unit response time ; represents the unit time length of the cement plant participating in demand response;

[0038] According to the actual production and safety requirements of the cement plant, the production rate of the main production equipment of each sub-process and the supply rate of the bucket storage to the main production equipment are limited within the safety range;

[0039] The product output of the main production equipment of the sub-process after participating in demand response at least meets the production plan, and the constraints are shown in equations (19) and (20):

[0040] (19); ​

[0041] wherein: represents the planned production of the cement plant;

[0042] (20);

[0043] wherein: represents the compensation price given by the cement user, indicates that the compensation price of the cement user is given by the cement user; represents the minimum value of the compensation price of the cement user; represents the maximum value of the compensation price of the cement user.

[0044] Further, in step S2, a federated alternating direction method of multipliers distributed algorithm is proposed based on the alternating direction method of multipliers and the federated learning principle;

[0045] The federated alternating direction method of multipliers distributed algorithm is established, and the corresponding augmented Lagrangian function is defined as shown in formula (28):

[0046] (28);

[0047] wherein: represents the augmented Lagrangian function of the client, and the client is an industrial user; represents the augmented Lagrangian function of the whole problem; represents the model parameter of the client ; represents the Lagrangian multiplier of the client ; represents the model parameter matrix; represents the Lagrangian multiplier matrix; represents the client, represents the model parameter, represents the Lagrangian multiplier; represents the weight of the client ; represents the weight of each client; represents the loss function of the client ; represents the loss function; represents the penalty factor of the client ; represents the penalty factor, and ; represents the total number of clients ;

[0048] The federated alternating direction method of multipliers distributed algorithm is iteratively solved as shown in formula (29):

[0049] (29);

[0050] In the formula: Indicates the first Global parameters for the next iteration Indicates the iteration round; Indicates the first The model parameter matrix for the next iteration; Indicates the first The Lagrange multiplier matrix of the next iteration; Indicates the client In the Model parameters for the next iteration; Indicates the client In the Lagrange multipliers in the next iteration; This indicates local parameters updated by the client. Denotes the Lagrange multiplier in the k-th iteration;

[0051] An improvement to the federated alternating direction multiplier method distributed algorithm:

[0052] Step S201: Set parameters , Indicates each iteration Each communication is performed once; settings Represents a specific set of iteration rounds, where the number of iterations... At that time, the central server receives data sent by the client and selects appropriate parameters. Significantly reduces communication rounds; the central server acts as a load aggregator.

[0053] Step S202: Set threshold parameters , This indicates the actual number of clients receiving data, prior to the central server receiving the data. After a client sends data, it no longer waits for other clients; that is, other clients do not participate in the update of global parameters in this iteration.

[0054] Step S203: When transmitting coupling parameters, the client avoids sending raw data directly and instead uses encrypted boundary information to enhance the protection of sensitive data of each subject.

[0055] Furthermore, in step S3, the hierarchical response model of the industrial park is solved using the federated alternating direction multiplier method distributed algorithm, specifically as follows:

[0056] In the industrial park hierarchical response model composed of the load aggregator layer and the industrial user layer, the load aggregator and the industrial user maximize the benefits by optimizing the demand response compensation price; meanwhile, the optimization variables of the load aggregator layer also include the response state variable, the response power and the response time of each industrial user; the optimization variables of the industrial user layer also include the response state, the response power and the response time of the power of each main production equipment;

[0057] The demand response compensation price paid by the load aggregator to the industrial user and the demand response income price obtained by the industrial user remain equal;

[0058] The specific calculation process of the federal alternating direction multiplier method distributed algorithm for solving the industrial park hierarchical response model is as follows:

[0059] Step S301: parameter initialization, setting the initial values of the load aggregator layer sub-problem, the industrial user layer sub-problem and the parameters of the federal alternating direction multiplier method distributed algorithm, and setting the iteration number to 0;

[0060] Step S302: calculating the industrial user layer sub-problem according to the solving result of the load aggregator layer sub-problem, obtaining the result of the device-level response optimization model and the numerical value used in the first iteration;

[0061] Step S303: updating the global variable, the Lagrange multiplier and the residual error respectively according to formula (29);

[0062] Step S304: calculating the residual error and judging whether it meets the convergence condition.

[0063] In a second aspect, the present application provides an industrial park hierarchical response system based on the federal alternating direction multiplier method, comprising:

[0064] The model module: an industrial park hierarchical response model considering the multi-agent benefits of the load aggregator and the industrial user and the process flow is established, the industrial park hierarchical response model includes the load aggregator layer and the industrial user layer, the load aggregator layer optimizes the user-level response scheme with the minimum cost as the target, and the industrial user layer optimizes the device-level response scheme with the maximum additional income as the target;

[0065] The algorithm module: based on the industrial park hierarchical response model, a federal alternating direction multiplier method distributed algorithm is constructed, the global parameter updating, the information transmission encryption and the client iteration rule are improved to improve the sensitive data protection and the calculation efficiency in the iteration process;

[0066] ​The optimization module: the industrial park hierarchical response model is solved by a federal alternating direction multiplier method distributed algorithm, the load aggregator layer and the industrial user layer dynamically update the compensation price until converging to an optimal solution, and a response scheme considering the interests of the load aggregator and the industrial user is obtained.

[0067] In a third aspect, the present application provides a computer readable storage medium storing a computer program, the computer program being executed by a processor to implement the industrial park hierarchical response method based on the federal alternating direction multiplier method.

[0068] In a fourth aspect, the present application provides an electronic device comprising at least one processor and a memory connected to the at least one processor in communication, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the industrial park hierarchical response method based on the federal alternating direction multiplier method.

[0069] The present application has the following beneficial effects:

[0070] (1) The industrial park hierarchical response method and system based on the federal alternating direction multiplier method establish an industrial park hierarchical response model considering the interests of the load aggregator and the industrial user and the process flow, the aggregator layer optimizes the user-level response scheme with the minimum cost as the target, and the user layer optimizes the equipment-level response scheme with the maximum additional income as the target; the federal alternating direction multiplier method distributed algorithm is proposed, the global parameter update, information encryption and client iteration rules are improved, and the sensitive data protection in the iteration process and the improvement of the calculation efficiency are realized.

[0071] (2) The federal alternating direction multiplier method distributed algorithm is used to solve the industrial park hierarchical response model, the load aggregator layer and the industrial user layer dynamically update the compensation price until converging to an optimal solution, and a response scheme considering the interests of the load aggregator and the industrial user is obtained, the sensitive data protection and the improvement of the calculation efficiency are realized, and a demand response scheme considering the interests of multiple subjects and the process flow is obtained.

[0072] (3) The user-level aggregation optimization model is designed as a mixed integer nonlinear programming problem, the response time and the response capacity of the user can change within a certain range, and the flexibility of the response strategy is improved. The cement equipment-level response model considers the correlation of the process flow before and after the production process and the intermediate product, and the electric smelting magnesium equipment-level response model considers the production safety constraint, and the obtained equipment-level response scheme is more practical. BRIEF DESCRIPTION OF DRAWINGS

[0073] The exemplary embodiments of the present application can be more completely understood by referring to the following drawings:

[0074] Figure 1A flow chart of the industrial park hierarchical response method based on the federal alternating direction multiplier method is provided for the embodiment of the present application.

[0075] Figure 2 An overall architecture diagram of the industrial park hierarchical response model is provided for the embodiment of the present application.

[0076] Figure 3 A schematic diagram of the principle of the federal alternating direction multiplier method distributed algorithm is provided for the embodiment of the present application.

[0077] Figure 4 A structural schematic diagram of the industrial park hierarchical response system based on the federal alternating direction multiplier method is provided for the embodiment of the present application.

[0078] Figure 5 A structural schematic diagram of the electronic device is provided for the embodiment of the present application.

[0079] Figure 6 A load aggregator layer user level aggregation optimization result graph is provided for the embodiment 1 of the present application, wherein:

[0080] Figure 6 a in the above table is the response situation of the aggregator 1;

[0081] Figure 6 b in the above table is the response situation of the aggregator 2;

[0082] Figure 6 c in the above table is the response situation of the aggregator 3.

[0083] Figure 7 An industrial user layer device level response optimization result graph is provided for the embodiment of the present application, wherein:

[0084] Figure 7 a in the above table is the FM user 30 device response situation;

[0085] Figure 7 b in the above table is the CM user 14 device response situation. DETAILED DESCRIPTION

[0086] In order to make the technical problems to be solved by the present application, the technical solutions and beneficial effects more clear and explicit, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0087] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs; the terms used herein are only for the purpose of describing specific embodiments, and are not intended to limit the present application.

[0088] The embodiment of the application provides an industrial park hierarchical response method based on a federal alternating direction multiplier method, and comprises the following steps:

[0089] Step S1: An industrial park hierarchical response model considering the interests of load aggregators and industrial users and a process is constructed, the industrial park hierarchical response model comprises a load aggregator layer and an industrial user layer, the load aggregator layer optimizes a user-level response scheme with the minimum cost as an objective, and the industrial user layer optimizes an equipment-level response scheme with the maximum additional income as an objective; wherein the load aggregator layer comprises a load aggregator platform responsible for the industrial park and industrial users of different industries participating in the response project in the industrial park, and the industrial user layer comprises industrial equipment of different industries with adjustable loads;

[0090] It can be understood that in a demand response project, the load aggregator and the industrial user are different interest subjects, both of which hope to maximize their economic benefits, and there is an obvious conflict of interests; the optimization objective of the upper load aggregator is to minimize the cost by reducing the compensation price and overall planning the user response, and the optimization objective of each industrial user in the lower layer is to maximize the additional income without affecting the production plan; in order to balance the interests of both parties, the different interest demands of the upper and lower subjects should be considered when the demand response scheme is formulated, and the response ability differences of industrial users of different industries should be considered, so as to obtain a response scheme considering the interests of the load aggregator and the industrial user;

[0091] Step S2: Based on the industrial park hierarchical response model, a federal alternating direction multiplier method distributed algorithm is constructed, and the global parameter updating, transmission information encryption and client iteration rules are improved to improve the sensitive data protection and calculation efficiency in the iteration process;

[0092] It can be understood that firstly, the federal alternating direction multiplier method distributed algorithm only averages parameters in some steps, and only needs to communicate between the client and the central server in these steps, thereby effectively reducing the communication rounds; secondly, the federal alternating direction multiplier method distributed algorithm selects part of the clients to participate in training in each step, thereby reducing the influence of the lag phenomenon in the clients; thirdly, when the coupling parameters are transmitted, the client avoids directly sending the original data and instead uses the encrypted intermediate data, thereby enhancing the protection of the sensitive data of each subject;

[0093] Step S3: The industrial park hierarchical response model is solved through the federal alternating direction multiplier method distributed algorithm, the load aggregator layer and the industrial user layer dynamically update the compensation price until the optimal solution is converged, and a response scheme considering the interests of the load aggregator and the industrial user is obtained.

[0094] The process of the industrial park hierarchical response method based on the federal alternating direction multiplier method is as shown in Figure 1As shown, the overall architecture of the hierarchical response model of the industrial park considering the interests of the load aggregator and industrial users and the process flow is as shown in Figure 2 .

[0095] In some embodiments, in step S1, when the hierarchical response model of the industrial park considering the interests of the load aggregator and industrial users and the process flow is used to carry out the incentive demand response project, the load aggregator layer provides a user-level response scheme of the industrial park under its responsibility according to the requirements of the power grid, and economically compensates the industrial users participating in the response project, and the industrial user layer adjusts the production plan of the equipment according to the user-level response scheme to meet the requirements of the power grid.

[0096] Specifically, before formulating the user-level response scheme, the industrial user layer reports response characteristic parameters to the load aggregator layer, the response characteristic parameters including the maximum capacity of the adjustable load of the industrial user, the response duration, the response period and the acceptable compensation price range; then the load aggregator layer collects the response characteristic parameters reported by the industrial user layer, and based on the typical demand response scenarios investigated, constructs a user-level aggregation optimization model, and then performs optimization solving, in which the industrial users participating in the response project are divided into several groups, one group of industrial users being one aggregation, the user response of each aggregation obtained is taken as the user-level response scheme, then the user-level response scheme is provided to the industrial users participating in the response project, and the industrial users participating in the response project are economically compensated, and the industrial user layer constructs an equipment-level response optimization model and solves it according to the user-level response scheme and in combination with its own characteristics, to obtain the response scheme of the industrial equipment of the adjustable load.

[0097] In some embodiments, the user-level aggregation optimization model is constructed and then optimization solving is performed, specifically as follows.

[0098] The user-level aggregation optimization model considering the interests of the load aggregator and industrial users and the process flow is established, the load aggregator layer collects the response characteristic parameters reported by the industrial user layer, and based on the typical demand response scenarios, determines the key parameters of the user-level aggregation optimization model, the user-level aggregation decision variables of the user-level aggregation optimization model being the belonging of each industrial user to an aggregation, the response in a unit response time and the actual response capacity;

[0099] The decision variables are set as , , . represents the response state of the industrial user in a unit response time in the aggregation . represents the response state of the industrial user in a unit response time in the aggregation start response state in, denotes industrial users in unit response time in the aggregation response capacity in, wherein denotes response state, denotes start response state, denotes response capacity, denotes industrial users, denotes aggregation, denotes unit response time, and is a state variable;

[0100] The specific definitions are as follows:

[0101] is a 0-1 variable, which indicates that the industrial user is assigned to the aggregation and participates in the response state at the unit response time ; otherwise, ;

[0102] is a 0-1 variable, which indicates that the industrial user is assigned to the aggregation and starts to participate in the response at the unit response time ; otherwise, ;

[0103] is a real variable, which indicates the capacity of the industrial user assigned to the aggregation and participating in the response at the unit response time ;

[0104] The user-level aggregation optimization model aims to minimize the total response compensation cost of the aggregation, as shown in equation (1):

[0105] (1);

[0106] In the formula: denotes the total response compensation cost of the user-level aggregation; denotes the weight factor of the aggregation , which is set according to the demand response scenario; denotes the compensation price given by the load aggregator to the industrial user , indicates that the compensation price is given by the load aggregator; ​​denotes the total number of industrial users total number, denotes the total number of industrial users denotes the total number of industrial users denotes the total number of industrial users denotes the unit response time total number,

[0107] According to the industrial user response time period distribution and the typical power grid demand response time period, constraints as shown in equation (2) are established to limit industrial users to only participate in aggregation during their response time period:

[0108] (2);

[0109] In the formula: denotes the total number of industrial users in the unit response time whether it can participate in response; when , the industrial user can participate in response at the unit response time, otherwise it cannot respond;

[0110] When industrial users are allocated to participate in response in the aggregation, the actual response capacity cannot exceed the maximum response capacity limit, and the aggregation as a whole needs to meet the response capacity constraint, as shown in equations (3) and (4):

[0111] (3);

[0112] (4);

[0113] In the formula: denotes the total number of industrial users maximum response capacity; denotes the target response capacity of the aggregation ;

[0114] The compensation price must be between the upper and lower limits of the price, as shown in equation (5):

[0115] (5);

[0116] In the formula: denotes the total number of industrial users minimum compensation price; denotes the total number of industrial users maximum compensation price;

[0117] Equations (1)-(5) constitute a user-level aggregation optimization model constructed by the load aggregator layer, which is a mixed integer nonlinear programming (MINLP) problem.

[0118] In some embodiments, the device-level response optimization model is constructed and solved to obtain a response scheme of the industrial device with adjustable load, including the following steps:

[0119] An electrically fused magnesium device-level response optimization model is established, which takes into account the interests of both the load aggregator and the industrial user and considers safety constraints. The constraints considered in the device-level response optimization scheduling process of the electrically fused magnesium device-level response optimization model are:

[0120] To ensure production safety and product quality, the power of the magnesium arc furnace must be controlled within a safe range during production to prevent accidents caused by excessive power and to avoid affecting the quality of magnesium oxide due to insufficient power;

[0121] To ensure that the production task of the electrically fused magnesium user is not affected, the product output of the magnesium arc furnace device remains unchanged before and after participating in demand response;

[0122] The upper and lower limits of the compensation price are shown in formula (6):

[0123] (6);

[0124] In the formula: represents the compensation price given by the electrically fused magnesium user, is used to identify that the compensation price is given by the user, represents electrically fused magnesium; represents the minimum value of the compensation price of the electrically fused magnesium user; represents the maximum value of the compensation price of the electrically fused magnesium user;

[0125] Considering the actual operation of the electrically fused magnesium plant, the adjustment of the device will generate adjustment costs when participating in response, and the product revenue of the electrically fused magnesium plant and the revenue obtained by participating in demand response are also considered;

[0126] The purpose of the industrial high-load capacity load participating in demand response is to obtain additional revenue from the load aggregator while ensuring production, so the additional revenue obtained by the electrically fused magnesium user after participating in demand response is composed of four parts, including the change in electricity cost , device adjustment cost , production benefit change and demand response revenue ;

[0127] The calculation process of the additional revenue obtained by the electrically fused magnesium user after participating in demand response is shown in formula (7);

[0128] The calculation process of the change in electricity cost of the electrically fused magnesium plant is shown in formula (8);

[0129] The calculation process of the electrically fused magnesium device adjustment cost is shown in formula (9), and the electrically fused magnesium device operating power will generate a corresponding cost every time it changes.

[0130] The calculation process of the change in production efficiency of the fused magnesium plant is shown in formula (10). Since the constraint ensures that the output of the fused magnesium plant remains unchanged, the production efficiency does not change.

[0131] The calculation process for the revenue obtained by the fused magnesium plant from participating in demand response is shown in formula (11);

[0132] (7);

[0133] (8);

[0134] (9);

[0135] (10);

[0136] (11);

[0137] In the formula: This represents the total benefit of fused magnesium users participating in demand response; and The variables are 0 and 1, representing the first and second digits respectively. A magnesium melting electric arc furnace at a unit response time The upward and downward states, when and When the molten magnesium electric arc furnace is in power-up mode, and The electric arc furnace for molten magnesium is in a power-down mode. Indicates a magnesium melting electric arc furnace; Indicates the first A magnesium melting electric arc furnace at a unit response time Operating power after participating in the response; Indicates the first A magnesium melting electric arc furnace at a unit response time Operating power before participating in the response; This indicates the unit duration of cement plant participation in demand response; and These represent the actual and planned output of fused magnesium users, respectively. This indicates the unit industrial electricity price; This indicates the cost of adjusting the power of a magnesium melting electric arc furnace once. This indicates the unit economic benefit of fused magnesium products; This represents the revenue per unit of response capacity for fused magnesium users participating in demand response; This indicates the start time of the actual response period for fused magnesium users. This indicates the end time of the actual response period for fused magnesium users. indicates the start of demand response, indicates the end of demand response; indicates a magnesium arc furnace total number; unit response time The total number of the unit (hour) is 24;

[0138] In some embodiments, the device-level response optimization model is constructed and solved to obtain a response scheme of the industrial device with adjustable load, further comprising the following steps:

[0139] A cement device-level aggregation optimization model is established, which takes into account the interests of load aggregators and industrial users and considers the process flow. The complete production line of the cement plant is divided into four sub-processes: crushing (C) process, kiln feed preparation (KFP) process, clinker production (CP) process, and finish grinding (FG) process. A silo reserve is provided between each two sub-processes to store intermediate products. When the demand response demand is issued from the grid side, the cement production line provides flexible load by changing the main production equipment (i.e., the industrial device with adjustable load) of the sub-process;

[0140] The main production equipment of the crushing process, the kiln feed preparation process, the clinker production process, and the finish grinding process is a crusher, a raw mill, a clinker mill, and a cement mill, respectively. The operating power of each sub-process is determined by the production rate. The intermediate products produced by each sub-process are all stored in the corresponding silo reserve. Therefore, the power of the main production equipment, the amount of silo reserve, and the relationship between storage and production are constrained as shown in equations (12)-(15):

[0141] (12);

[0142] (13);

[0143] (14);

[0144] (15);

[0145] In the formula: indicates the total power of the cement plant in unit response time running, indicates the cement; indicates the main production equipment of the sub-process in unit response time unit production consumption rated power, unit (kWh / ton); indicates the main production equipment of the sub-process in unit response time Power after the response; Subroutines Unit response time Productivity, in units (tons / h); Subroutines Barrel storage in unit response time Next subprocess The material supply rate; Subroutines The ratio of input raw material quantity to output raw material quantity; Subroutines Unit response time The amount of raw materials stored, in ton.

[0146] Based on the actual production and safety requirements of the cement plant, it is necessary to limit the production rate of the main production equipment in each sub-process and the supply rate of the silo reserves to the main production equipment within a safe range, as shown in the constraint formulas (16) and (17):

[0147] (16);

[0148] (17);

[0149] In the formula: and These represent the minimum and maximum productivity values ​​of the main production equipment in subprocess s, respectively. and Representing sub-procedures Next subprocess The minimum and maximum values ​​of the supply rate of the barrel and silo reserves to the main production equipment;

[0150] The storage capacity of the silos between each two subprocesses is limited. If the storage capacity of the silos is too large, it may interfere with the normal production process of the cement plant. If the storage capacity of the silos is too small, it may affect the output of the cement plant. Therefore, the storage capacity of each silo is constrained as shown in formula (18):

[0151] (18);

[0152] In the formula: and Representing sub-procedures The minimum and maximum values ​​of barrel storage capacity;

[0153] To ensure production targets are met, cement plants participating in demand response must ensure that the main production equipment in their subprocesses produces at least the required output after the response, as shown in formulas (19) and (20).

[0154] (19);

[0155] wherein: represents the planned production of the cement plant;

[0156] (20);

[0157] wherein: represents the compensation price given by the cement user, indicates that the compensation price of the cement user is given by the cement user; represents the minimum value of the compensation price of the cement user; represents the maximum value of the compensation price of the cement user;

[0158] For the economic benefits of the cement plant participating in demand response, the change amount of electricity cost , the equipment adjustment cost , the change amount of production benefits and the demand response income are also considered. Since the industrial equipment of the adjustable load of the cement plant has high flexibility, the equipment adjustment cost of the present embodiment is 0, so the additional income of the cement plant participating in demand response , and the calculation process is shown in formulas (21)-(25):

[0159] (21);

[0160] (22);

[0161] (23);

[0162] (24);

[0163] (25);

[0164] wherein: represents the unit economic benefit of the cement plant product; represents the unit response capacity income of the cement plant user participating in demand response; represents the actual response time period start time of the cement user, represents the actual response time period end time of the cement user; represents the rated production rate of the sub-process ;

[0165] Based on the established response characteristic model of the adjustable load industrial equipment in the electrically fused magnesium and cement plant, while considering the actual response capacity of each user obtained by the optimization and solution of the load aggregator layer, the adjustable load industrial equipment also needs to meet the target capacity constraint.

[0166] To maximize the response enthusiasm of industrial users, the equipment-level response optimization model takes the maximum additional income obtained by each industrial user participating in demand response as the objective function, and the calculation process is shown in formulas (26) and (27):

[0167] (26);

[0168] (27);

[0169] Formulas (6)-(11) and (26) constitute the electric magnesium equipment-level response optimization model, which is a MINLP problem; and formulas (12)-(25) and (27) constitute the cement equipment-level aggregation optimization model, which is a linear programming problem.

[0170] In some embodiments, in step S2, a federated alternating direction method of multipliers (FedADMM) distributed algorithm is proposed based on the alternating direction method of multipliers (ADMM) and the federated learning (FL) principle;

[0171] The traditional distributed optimization method needs to directly exchange parameter calculation results between sub-problem subjects, which is easy to cause sensitive information leakage and affect the participation willingness of demand response subjects. Therefore, the federated learning framework is used for local optimization of each sub-problem subject to protect the sensitive data of the load aggregator and each industrial user.

[0172] The task of federated learning is to learn the optimal model parameters to minimize the global loss function value, and the alternating direction method of multipliers is an effective algorithm for solving large-scale distributed optimization problems. According to the basic principle, a federated alternating direction method of multipliers method is proposed for optimization calculation, and the principle schematic diagram is shown in Figure 3 .

[0173] The federated alternating direction method of multipliers distributed algorithm is established, and the corresponding augmented Lagrangian function definition is shown in formula (28):

[0174] (28);

[0175] In the formula, Lc represents the augmented Lagrangian function of the client, and the client is an industrial user; L represents the augmented Lagrangian function of the whole problem; Lc represents the augmented Lagrangian function of the client, and the client is an industrial user; model parameters; denote the client Lagrange multipliers; denote the model parameter matrix; denote the Lagrange multiplier matrix; denote the client, denote the model parameters, denote the Lagrange multipliers; denote the weight of the client , denote the weight of each client; denote the client loss function, denote the loss function; denote the penalty factor of the client , denote the penalty factor, and ; denote the total number of clients ;

[0176] Therefore, on the basis of determining the initial value, the established federated alternating direction multiplier method distributed algorithm is solved by iteration as shown in formula (29):

[0177] (29);

[0178] In the formula: denote the global parameters of the iteration, denote the iteration round; denote the model parameter matrix of the iteration; denote the Lagrange multiplier matrix of the iteration; denote the model parameters of the client in the iteration; denote the Lagrange multipliers of the client in the iteration; denote the local parameters updated by the client; denote the Lagrange multipliers of the k iteration;

[0179] Directly applying the above federal alternating direction multiplier method distributed algorithm for solving may encounter problems in communication cost and computing time. First, the algorithm needs to be updated three times repeatedly at each iteration, requiring the client and the central server to communicate at each iteration, greatly increasing the communication cost. Second, some clients may delay the transmission of necessary parameters to the central server due to insufficient communication transmission resources or limited server computing capacity, resulting in a significant increase in the overall solution time of the algorithm. Third, the direct transmission of coupling parameters between the central server and the clients is not conducive to protecting the internal sensitive information of each subject.

[0180] To solve the above problems, the federal alternating direction multiplier method distributed algorithm is improved as follows:

[0181] (1) Design the federal alternating direction multiplier method distributed algorithm to average parameters only in certain steps. The client and the central server only need to communicate in these steps, effectively reducing the communication rounds. Specifically:

[0182] Set the parameter , to indicate that the client communicates once every iteration ; Set to indicate a specific set of iteration rounds. When the iteration number , the central server receives the data sent by the client, selects the appropriate parameter , and significantly reduces the communication rounds. The central server is the load aggregator.

[0183] (2) The federal alternating direction multiplier method distributed algorithm selects part of the clients to participate in training at each step, which can reduce the impact of the dropout phenomenon in the clients. Specifically:

[0184] Set the threshold parameter , to indicate the number of clients that actually receive data. After the central server receives the data sent by the first clients, it no longer waits for other clients, i.e., other clients are considered to be dropouts at this iteration and do not participate in the update of the global parameters. The selected clients in the iteration are denoted as , , and the selected clients are denoted as

[0185] (3) When transmitting coupling parameters, the client avoids directly sending raw data and uses encrypted intermediate data , enhancing the protection of sensitive data of each subject. denotes the encrypted boundary information.

[0186] Further, the federated alternating direction multiplier method distributed algorithm allows the central server to record data previously sent by clients other than For any client At iteration number Let be the largest integer such that and Then is the last iteration round for which the client The central server uses data previously sent by clients other than and data sent by clients in at this time.

[0187] In some embodiments, in step S3, the industrial park hierarchical response model is solved by the federated alternating direction multiplier method distributed algorithm, specifically:

[0188] In the industrial park hierarchical response model composed of the load aggregator layer and the industrial user layer, the load aggregator and the industrial user both maximize the benefits by optimizing the demand response compensation price; at the same time, the optimization variables of the load aggregator layer also include the response state variable, the response power and the response time of each industrial user; the optimization variables of the industrial user layer also include the response state, the response power and the response time of each main production equipment power;

[0189] Since the demand response compensation price paid by the load aggregator to the industrial user and the demand response income price obtained by the industrial user need to be kept equal, an auxiliary variable is set , which represents the auxiliary variable vector, respectively represents the auxiliary variable of industrial user 1, industrial user 2 and industrial user , and a consistency constraint is added, as shown in formula (30):

[0190] (30);

[0191] In the formula: respectively represent the compensation price given by the load aggregator to industrial user 1, industrial user 2, industrial user in the load aggregator layer sub-problem; respectively represent the compensation price given by the industrial user to industrial user 1, industrial user 2, industrial user in the industrial user layer sub-problem;

[0192] Therefore, the auxiliary variable becomes a global variable, and formula (30) becomes a coupling constraint between the industrial user layer and the device-level response optimization model;

[0193] The optimization model for the load aggregation quotient sub-problem, i.e. the user-level aggregation optimization model, is shown in formula (31):

[0194] (31);

[0195] In the formula: Industrial users in the load aggregation quotient layer subproblem Lagrange multipliers; Indicates industrial users The actual compensation price is the global parameter jointly optimized by the load aggregator and the industrial user; Indicates load aggregator Penalty factor in the FedADMM algorithm; This represents the compensation price vector provided by the load aggregator;

[0196] The user-level aggregation optimization model is a mixed integer nonlinear programming (MINLP) problem. The objective function contains nonlinear terms, and the McCormick envelope method is used to linearly relax the nonlinear terms in the model.

[0197] The optimization model for the industrial user-level sub-problem, namely the device-level response optimization model, is shown in formula (32):

[0198] (32);

[0199] In the formula: Indicates industrial users The change in electricity costs Indicates industrial users Equipment adjustment costs, Indicates industrial users Changes in production efficiency Indicates industrial users Demand response benefits; Industrial users in the industrial user layer subproblem Lagrange multipliers; Indicates industrial users Penalty factor in the FedADMM algorithm;

[0200] The optimization model for fused magnesium user equipment is a MINLP problem with constraints containing nonlinear terms. The McCormick envelope method is also used to linearly relax the nonlinear terms.

[0201] The specific calculation process of the federated alternating direction multiplier method distributed algorithm for solving the hierarchical response model of industrial parks is as follows:

[0202] (1) Parameter initialization, set the initial value of each parameter of the load aggregator layer sub-problem, each user sub-problem of the industrial user layer, and the distributed algorithm of the federal alternating direction multiplier method, and set the iteration number to 0.

[0203] (2) First, calculate the upper sub-problem of the load aggregator according to formula (31), and send the solution (including whether each user participates in response, the time of participating in response, and the response capacity) to each industrial user; then, each electric magnesium and cement user calculates the lower sub-problem of each user according to formula (32) respectively; thus, the result of the device-level response optimization model and the numerical value used in the first iteration are obtained.

[0204] (3) The global variable, Lagrange multiplier, and residual error are updated according to formula (29).

[0205] (4) Calculate the residual error and judge whether it meets the convergence condition, as shown in formula (33):

[0206] (33).

[0207] In the formula: is the residual error vector of the th iteration; is the auxiliary variable vector of the th iteration; is the auxiliary variable vector of the th iteration; is the residual error convergence threshold.

[0208] In some embodiments, the present embodiment provides an industrial park hierarchical response system 200 based on the federal alternating direction multiplier method, as shown in FIG. 2, which includes a model module 210, an algorithm module 220, and an optimization module 230: Figure 4

[0209] The model module 210: establishes an industrial park hierarchical response model that takes into account the interests of the load aggregator and the industrial users and considers the process flow, the industrial park hierarchical response model includes a load aggregator layer and an industrial user layer, the load aggregator layer optimizes the user-level response scheme with the minimum cost as the target, and the industrial user layer optimizes the device-level response scheme with the maximum additional income as the target.

[0210] The algorithm module 220: based on the industrial park hierarchical response model, constructs a distributed algorithm of the federal alternating direction multiplier method, and improves the global parameter update, information transmission encryption, and client iteration rules to improve the protection of sensitive data and computing efficiency in the iteration process.

[0211] ​​​The optimization module 230: solve the hierarchical response model of the industrial park by the federal alternating direction multiplier method distributed algorithm, dynamically update the compensation price of the load aggregator layer and the industrial user layer until the optimal solution is converged, and obtain the response scheme considering the interests of the load aggregator and the industrial user.

[0212] In some embodiments, the embodiments of the present application provide a computer readable storage medium, which can include a storage program area and a storage data area, wherein the storage program area can store an operating system and a computer program required by at least one function, the computer program is executed by a processor to implement an industrial park hierarchical response method based on the federal alternating direction multiplier method; the storage data area can store data created by an industrial park hierarchical response system based on the federal alternating direction multiplier method, and the like.

[0213] In addition, the computer readable storage medium can include a high-speed random access memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state memory device.

[0214] The computer readable storage medium can optionally include a memory remotely arranged relative to the processor, and these remote memories can be connected to the industrial park hierarchical response system based on the federal alternating direction multiplier method through a network, examples of the network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0215] In some embodiments, the embodiments of the present application provide an electronic device, such as Figure 5 As shown, the electronic device includes a processor 310 and a memory 320, and can further include an input device 330 and an output device 340.

[0216] The processor 310, the memory 320, the input device 330, and the output device 340 can be connected by a bus or other means, Figure 5 For example, by a bus connection;

[0217] The memory 320 is a computer readable storage medium of the embodiments, and the processor 310 executes various function applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 320, that is, implements the industrial park hierarchical response method based on the federal alternating direction multiplier method; the input device 330 can receive input digital or character information, and generate key signal input related to user settings and function control of the industrial park hierarchical response system based on the federal alternating direction multiplier method; the output device 340 can include a display device such as a display screen;

[0218] The electronic device is applied to an industrial park hierarchical response system based on a federal alternating direction multiplier method, and is used for a client, and comprises at least one processor, and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute an industrial park hierarchical response method based on the federal alternating direction multiplier method.

[0219] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and necessary universal hardware platforms, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions essentially or in other words, the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of the embodiments or some parts of the embodiments.

[0220] Embodiment 1

[0221] The actual data of an industrial park in a certain place is taken as an example for calculation and analysis. The maximum capacity of the adjustable load of the industrial users in various industries, the response time, and the acceptable compensation price range are shown in Table 1; the response characteristic parameters of the typical adjustable equipment of the electric smelting magnesium and cement users are shown in Tables 2 and 3.

[0222] Table 1 Response parameters of industrial users

[0223]

[0224] Table 2 Response parameters of electric smelting magnesium equipment

[0225]

[0226] Table 3 Response parameters of cement equipment

[0227]

[0228] According to the actual production, multiple demand response time periods of the industrial park are determined, such as 11:00-15:00 and 16:00-20:00, to obtain a preset typical scenario response time period set. Taking the 11:00-15:00 time period as an example for specific analysis, the response time is set to 4h, and the unit response time of the optimization model is 1h.

[0229] Considering the user response credibility and equipment failure and other sudden situations, in order to meet the requirements of the power grid side to the greatest extent, a certain capacity of the aggregator needs to be set as a backup. Therefore, the number of user-level aggregators in the load aggregator layer sub-problem is set to 2. is 3, the response period is 11:00-15:00, and the target response capacity of each aggregation body is = 20 MW, = 30 MW, = 40 MW, and the compensation price of the load aggregator is given according to the expected range reported by each user.

[0230] Based on the above basic data and parameters, the industrial park hierarchical distributed optimization model established in this embodiment is solved by calling Gurobi. The allocation of industrial users in the aggregation body and the actual response situation, as well as the overall response situation of the aggregation body, are as shown in Figure 6 , wherein Figure 6 a in the aggregation body 1 response situation, Figure 6 b in the aggregation body 2 response situation, Figure 6 c in the aggregation body 3 response situation.

[0231] It can be seen from Figure 6 that the industrial users are aggregated into three aggregation bodies according to their respective response characteristics and costs, the overall response capacity of each aggregation body meets the constraint condition, and all the screened users are within their response period and the response capacity and length do not exceed their maximum limit. Only part of the users are selected to participate in this demand response in this scenario, and the participation ratio of the electric melting magnesium and cement industry users is similar. The users with lower response compensation prices in each industry are preferentially distributed to the aggregation body to participate in the response, while the users with higher response compensation prices are not screened. The number of users distributed in the aggregation body increases with the increase of the target response capacity, and the target capacity of the aggregation body 3 is the largest, and a total of 9 users are distributed to the aggregation body.

[0232] The actual response capacity and compensation cost of each aggregation body when participating in demand response are shown in Table 4.

[0233] Table 4 Actual response capacity and cost of user-level aggregation body

[0234]

[0235] It can be seen from Figure 6 and Table 4 that the overall actual response capacity of the three aggregation bodies is almost equal to the target capacity, and there is no common situation that the actual response capacity is significantly greater than the target capacity, which shows that the model established in this embodiment has high load flexibility and can well meet the response demand of the power grid. The greater the target response capacity of the aggregation body, the higher the response compensation cost, and the target response capacity of the aggregation body 3 is the largest, and its compensation cost is the highest, which is 244736 yuan. In addition, the users with lower response compensation prices will be preferentially distributed to the aggregation body with higher weight factor. According to the historical demand response scenario, the weight factor is reasonably adjusted, so that the compensation cost of the aggregation body meeting the demand response target with higher frequency is lower, thereby reducing the related expenditure of the load aggregator.

[0236] based on the device-level optimization calculation results, Figure 7 The response of the typical industrial users (fused magnesium users 30 and cement users 14, hereinafter referred to as FM users 30 and CM users 14, and demand response is referred to as DR) is shown. Since the fused magnesium industry in the industrial park generally adopts a 24h continuous production mode, the load of its main production equipment generally remains at the same level within a day, with small fluctuations. For example, Figure 7 As shown in a of FIG. 11, the FM user 30 reduces the power of the electric arc furnace 3 during the period of 16:00-20:00 and increases the power of the electric arc furnace 3 during the period of 11:00-15:00 to complete the response task issued by the load aggregator. Limited by the number of adjustments and adjustment costs, the user 14 adjusts the power of the electric arc furnace 2 times within a production cycle (24h), and meets the user response task under the premise of ensuring production safety and production tasks.

[0237] Due to the production characteristics and the setting of the barrel storage link of the cement industry, the production load can be adjusted within a large range without causing intermediate product overflow, and some sub-process production can be interrupted to provide flexible load. For example, Figure 7 As shown in b of FIG. 11, the CM user 14 interrupts the C process in this production cycle, interrupts the KFP process during the periods of 0:00-7:00, 13:00-14:00 and 18:00-23:00, and interrupts the FG process during the periods of 0:00-3:00, 9:00-13:00 and 14:00-24:00, and relies on the remaining intermediate products of the barrel storage link of the last production cycle to maintain production. Since the clinker grinding equipment is not suitable for interruption, the user 14 increases the CP production power to complete the response task issued by the load aggregator. Overall, all the cement users participating in this demand response have similar device response situations as the typical user 14: interrupting the C process first, and mainly increasing the power of the KFP and CP process equipment. In addition, in order to obtain more additional product benefits, the cement user hopes to increase the production load as much as possible, which will significantly affect the production plan. Therefore, the model limits the excessive increase of the cement production load by the electricity cost, so that it completes the response plan issued by the load aggregator under the premise of the least impact on the production plan.

[0238] The optimization results of the compensation price when the interests of both the load aggregator and the industrial users are balanced are shown in Table 5. It can be seen that neither all of the participants in the demand response choose the minimum value of the compensation price to minimize the cost of the load aggregator, nor all of them choose the maximum value to maximize the income of the users, but within the range of the compensation price reported by the users, a final price that can reach a consensus is found, which shows that the hierarchical distributed model established in this embodiment can take into account the interests of both the load aggregator and the responding users. In a production cycle, the cost of the load aggregator participating in the demand response and the additional income of the cement and electric smelting magnesium users are shown in Table 6.

[0239] Table 5 Optimization results of the compensation price

[0240]

[0241] Table 6 Compensation cost of the load aggregator and additional income of each user

[0242]

[0243] To prove the effectiveness of the algorithm in reducing communication cost and protecting sensitive data, under the condition that the model parameters and the computing environment are the same, the hierarchical distributed model is solved by ADMM, and the iteration times, communication rounds and boundary parameters transmitted by FedADMM and ADMM are shown in Table 7.

[0244] Table 7 Comparison of solving effects of different algorithms

[0245]

[0246] It can be seen that although the iteration times of FedADMM are 24 more than those of ADMM, the communication rounds of FedADMM are less than those of ADMM, which shows that the method proposed in this embodiment can effectively reduce the communication cost. In terms of computing efficiency, the solving time of FedADMM is 36.3% of that of ADMM, which saves 1074 seconds of time and significantly improves the efficiency of the distributed solving method. The main reason for the difference in calculation time between the two is that the central server is responsible for solving the MINLP problem of the load aggregator agent. Each time the central server communicates with the client, the central server needs to recalculate the problem, and due to the nonlinearity of the problem, the solver needs a long time to solve. Therefore, the number of times the central server solves the MINLP problem will significantly affect the overall calculation time of the algorithm.

[0247] In terms of sensitive data protection, the ADMM transmits the coupling boundary parameter iteration information between the central server and the client in plaintext form. In this process, the actual boundary parameters are directly transmitted through the communication link, and there is a risk of sensitive information leakage, which may affect the willingness of industrial users to participate in demand response. However, in the federal alternating direction multiplier method distributed algorithm proposed in the embodiment, each client performs encrypted calculation on the coupling boundary parameters before transmitting them to the central server for data aggregation. Therefore, in the entire information exchange process of the optimization calculation, the risk of data leakage is effectively reduced, the isolation of key information in the transmission and calculation process is realized, and the sensitivity and security of sensitive data in the distributed collaborative computing environment are significantly improved.

[0248] The above merely describes preferred embodiments of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A hierarchical response method for industrial parks based on the federal alternating direction multiplier method, characterized in that, Includes the following steps: Step S1: Construct a hierarchical response model for industrial parks that takes into account the interests of multiple stakeholders, including load aggregators and industrial users, and considers the process flow. The hierarchical response model for industrial parks includes a load aggregator layer and an industrial user layer. The load aggregator layer optimizes the user-level response scheme with the goal of minimizing costs, while the industrial user layer optimizes the equipment-level response scheme with the goal of maximizing additional benefits. The load aggregator layer includes: the load aggregator platform responsible for the industrial park and industrial users from different industries participating in the response project within the industrial park; the industrial user layer includes: industrial equipment with adjustable loads from industrial users in different industries. Step S2: Based on the hierarchical response model of the industrial park, a federated alternating direction multiplier method distributed algorithm is constructed. By improving the global parameter update, transmission information encryption and client iteration rules, the sensitive data protection and computational efficiency of the iteration process are improved. Among them, a distributed algorithm based on the alternating direction multiplier method and the federated learning principle is proposed; A distributed algorithm based on the federated alternating direction multiplier method is established, and the corresponding augmented Lagrangian function is defined as shown in formula (28): (28)； In the formula: This represents the augmented Lagrangian function of the client, which is an industrial user. The augmented Lagrangian function representing the entire problem; Indicates client Model parameters; Indicates client Lagrange multipliers; Represents the model parameter matrix; Represents the Lagrange multiplier matrix; Indicates the client, Indicates model parameters, Represents the Lagrange multipliers; Indicates client The weight, This indicates the weight of each client; Indicates the client loss function Represents the loss function; Indicates the client The penalty factor Represents the penalty factor, and ; Indicates client total; The federated alternating direction multiplier method distributed algorithm is solved iteratively as shown in formula (29): (29)； In the formula: Indicates the first Global parameters for the next iteration Indicates the iteration round; Indicates the first The model parameter matrix for the next iteration; Indicates the first The Lagrange multiplier matrix of the next iteration; Indicates client In the Model parameters for the next iteration; Indicates client In the Lagrange multipliers in the next iteration; This indicates local parameters updated by the client. Denotes the Lagrange multiplier in the k-th iteration; An improvement to the federated alternating direction multiplier method distributed algorithm: Step S201: Set parameters , Indicates each iteration Each communication is performed once; settings Represents the set of iteration rounds, where the number of iterations is... At that time, the central server receives data sent by the client and selects appropriate parameters. Reduce communication rounds; the central server acts as a load aggregator. Step S202: Set threshold parameters , This indicates the actual number of clients receiving data, prior to the central server receiving the data. After a client sends data, it no longer waits for other clients; that is, other clients do not participate in the update of global parameters in this iteration. Step S203: When transmitting coupling parameters, the client avoids sending raw data directly and instead uses encrypted boundary information to enhance the protection of sensitive data of each subject. Step S3: Solve the hierarchical response model of the industrial park using the distributed algorithm of the federated alternating direction multiplier method. The compensation price of the load aggregator layer and the industrial user layer is dynamically updated until it converges to the optimal solution, thus obtaining a response scheme that takes into account the interests of both load aggregators and industrial users.

2. The hierarchical response method for industrial parks based on the federal alternating direction multiplier method as described in claim 1, characterized in that: In step S1, the hierarchical response model for industrial parks, which takes into account the interests of multiple stakeholders including load aggregators and industrial users and considers the process flow, is used to carry out incentive-based demand response projects. The load aggregator layer provides user-level response plans for the industrial parks under its responsibility according to the grid requirements and provides economic compensation to the industrial users participating in the response project. The industrial user layer adjusts the production plan of the equipment according to the user-level response plan to meet the grid requirements. Specifically, before formulating a user-level response plan, the industrial user layer reports response characteristic parameters to the load aggregator layer. These parameters include the industrial user's maximum adjustable load capacity, response duration, available response period, and acceptable compensation price range. The load aggregator layer then collects these parameters and, based on typical demand response scenarios surveyed, constructs a user-level aggregation optimization model. During this optimization process, the industrial users participating in the response project are divided into several groups, with each group forming an aggregate. The user response status of each aggregate is used as the user-level response plan. This plan is then provided to the participating industrial users, who receive economic compensation. Based on the user-level response plan and their own characteristics, the industrial user layer constructs and solves an equipment-level response optimization model to obtain the response plan for the adjustable load industrial equipment.

3. The hierarchical response method for industrial parks based on the federal alternating direction multiplier method as described in claim 2, characterized in that: Construct a user-level aggregation optimization model, and then perform optimization and solution, specifically as follows: Establish a user-level aggregation optimization model that takes into account the interests of multiple stakeholders, including load aggregators and industrial users, and considers the process flow. The load aggregator layer collects response characteristic parameters reported by the industrial user layer. Based on typical demand response scenarios, the key parameters of the user-level aggregation optimization model are determined. The user-level aggregation decision variables of the user-level aggregation optimization model are the aggregation status of each industrial user, the response status per unit response time, and the actual response capacity. set up , , For decision variables; Indicates industrial users Unit response time In the aggregate The response status in; Indicates industrial users Unit response time In the aggregate The start response status in the middle; Indicates industrial users Unit response time In the aggregate The response capacity in; The user-level aggregation optimization model aims to minimize the total response compensation cost of the aggregation, as shown in formula (1): (1)； In the formula: This represents the total response compensation cost for the user-level aggregate. Represents an aggregate The weighting factors are set according to the demand response scenario; This indicates the industrial users provided by the load aggregator. The compensation price, The compensation price is provided by the load aggregator; Represents an aggregate total, Indicates the total number; Indicates industrial users total; Indicates unit response time total; Based on the distribution of industrial user response periods and typical grid demand response periods, constraints as shown in Equation (2) are established to restrict industrial users to participate in aggregation only during their available response periods: (2)； In the formula: Indicates industrial users Unit response time Can I participate in the response? At that time, the industrial user can participate in the response within the unit's response time; otherwise, they cannot respond. When industrial users are assigned to participate in the response within an aggregate, the actual response capacity cannot exceed its maximum response capacity limit, and the aggregate must meet the response capacity constraints, as shown in formulas (3) and (4): (3)； (4)； In the formula: Indicates industrial users Maximum response capacity; Represents an aggregate The target response capacity.

4. The hierarchical response method for industrial parks based on the federal alternating direction multiplier method as described in claim 3, characterized in that: Constructing and solving a device-level response optimization model yields the response scheme for industrial equipment with adjustable loads, including the following steps: A response optimization model for fused magnesium equipment is established that balances the interests of multiple stakeholders, including load aggregators and industrial users, while also considering safety constraints. The constraints in the equipment-level response optimization scheduling process of the fused magnesium equipment-level response optimization model are as follows: (i) The power of the magnesium melting electric arc furnace equipment is controlled within a safe range during the production process; (ii) The output of the magnesium melting electric arc furnace equipment remains unchanged before and after participating in demand response; Additional benefits that fused magnesium users gain from participating in demand response include: changes in electricity costs, equipment adjustment costs, changes in production efficiency, and demand response benefits.

5. The hierarchical response method for industrial parks based on the federal alternating direction multiplier method as described in claim 4, characterized in that: The process of constructing and solving a device-level response optimization model to obtain the response scheme for industrial equipment with adjustable loads also includes the following steps: A cement equipment-level polymerization optimization model is established that takes into account the interests of multiple stakeholders, including load aggregators and industrial users, and considers the process flow. The complete production line of a cement plant is divided into four sub-processes: crushing process, kiln feed preparation process, clinker production process, and grinding process. The main production equipment in the crushing process, kiln feed preparation process, clinker production process, and grinding process are crusher, raw meal mill, clinker mill, and cement mill, respectively. Storage bins are set up between every two sub-processes. When participating in demand response, cement plants provide flexible loads by changing the power of the main production equipment in subprocesses; The power of the main production equipment, the storage capacity of the silos, and the relationship between storage and production in the cement plant are constrained by formulas (12)-(15): (12)； (13)； (14)； (15)； In the formula: Indicates the cement plant's unit response time Total operating power It refers to cement; Subroutines The main production equipment's unit response time The rated power consumption per unit of output; Subroutines The main production equipment's unit response time Power after the response; Subroutines Unit response time Productivity; Subroutines Barrel storage in unit response time Next subprocess The material supply rate; Subroutines The ratio of input raw material quantity to output raw material quantity; Subroutines Unit response time The amount of raw materials stored; This indicates the unit duration of cement plant participation in demand response; Based on the actual production and safety requirements of the cement plant, the production rate of the main production equipment in each sub-process and the supply rate of the silo reserves to the main production equipment are limited to a safe range. The main production equipment of the subprocess must produce enough products to meet the production plan after participating in demand response, as shown in formulas (19) and (20): (19)； In the formula: This indicates the planned output of cement plant products; (20)； In the formula: This indicates the compensation price offered by the cement user. The compensation price for cement users is determined by the cement users themselves. This represents the minimum compensation price for cement users; This indicates the maximum compensation price for cement users.

6. The hierarchical response method for industrial parks based on the federal alternating direction multiplier method as described in claim 5, characterized in that: In step S3, the hierarchical response model of the industrial park is solved using the federated alternating direction multiplier method distributed algorithm, specifically as follows: In the hierarchical response model of the industrial park, which consists of a load aggregator layer and an industrial user layer, both load aggregators and industrial users maximize their profits by optimizing the demand response compensation price. Meanwhile, the optimization variables of the load aggregator layer also include response state variables, response power and response time of each industrial user; the optimization variables of the industrial user layer also include response state, response power and response time of each major production equipment. The demand response compensation price paid by the load aggregator to the industrial user and the demand response revenue price received by the industrial user remain equal; The specific calculation process of the federated alternating direction multiplier method distributed algorithm for solving the hierarchical response model of industrial parks is as follows: Step S301: Parameter initialization. Set the initial values ​​for each parameter of the load aggregation quotient sub-problem, the industrial user sub-problem, and the federated alternating direction multiplier method distributed algorithm, and set the iteration number. Set to 0; Step S302: Calculate the industrial user layer sub-problem based on the solution results of the load aggregation quotient layer sub-problem, and obtain the results of the equipment-level response optimization model and the first... The values ​​used in the next iteration; Step S303: The global variables, Lagrange multipliers, and residuals are updated using formula (29); Step S304: Calculate the residuals and determine whether they meet the convergence conditions.

7. A hierarchical response system for industrial parks based on the federal alternating direction multiplier method, characterized in that, include: Model Module: Establish a hierarchical response model for industrial parks that takes into account the interests of multiple stakeholders, including load aggregators and industrial users, and considers the process flow. The hierarchical response model for industrial parks includes a load aggregator layer and an industrial user layer. The load aggregator layer optimizes the user-level response scheme with the goal of minimizing costs, while the industrial user layer optimizes the equipment-level response scheme with the goal of maximizing additional benefits. The load aggregator layer includes: the load aggregator platform responsible for the industrial park and industrial users from different industries participating in the response project within the industrial park; the industrial user layer includes: industrial equipment with adjustable loads from industrial users in different industries. Algorithm module: Based on the hierarchical response model of industrial parks, a federated alternating direction multiplier method distributed algorithm is constructed. By improving the global parameter update, transmission information encryption and client iteration rules, the sensitive data protection and computational efficiency of the iteration process are improved. Among them, a distributed algorithm based on the alternating direction multiplier method and the federated learning principle is proposed; A distributed algorithm based on the federated alternating direction multiplier method is established, and the corresponding augmented Lagrangian function is defined as shown in formula (28): (28)； In the formula: This represents the augmented Lagrangian function of the client, which is an industrial user. The augmented Lagrangian function representing the entire problem; Indicates client Model parameters; Indicates client Lagrange multipliers; Represents the model parameter matrix; Represents the Lagrange multiplier matrix; Indicates the client, Indicates model parameters, Represents the Lagrange multipliers; Indicates client The weight, This indicates the weight of each client; Indicates client loss function Represents the loss function; Indicates client The penalty factor Represents the penalty factor, and ; Indicates client total; The federated alternating direction multiplier method distributed algorithm is solved iteratively as shown in formula (29): (29)； In the formula: Indicates the first Global parameters for the next iteration Indicates the iteration round; Indicates the first The model parameter matrix for the next iteration; Indicates the first The Lagrange multiplier matrix of the next iteration; Indicates client In the Model parameters for the next iteration; Indicates client In the Lagrange multipliers in the next iteration; This indicates local parameters updated by the client. Denotes the Lagrange multiplier in the k-th iteration; An improvement to the federated alternating direction multiplier method distributed algorithm: Step S201: Set parameters , Indicates each iteration Each communication is performed once; settings Represents the set of iteration rounds, where the number of iterations is... At that time, the central server receives data sent by the client and selects appropriate parameters. Reduce communication rounds; the central server acts as a load aggregator. Step S202: Set threshold parameters , This indicates the actual number of clients receiving data, prior to the central server receiving the data. After a client sends data, it no longer waits for other clients; that is, other clients do not participate in the update of global parameters in this iteration. Step S203: When transmitting coupling parameters, the client avoids sending raw data directly and instead uses encrypted boundary information to enhance the protection of sensitive data of each subject. Optimization module: The distributed algorithm of the federated alternating direction multiplier method is used to solve the hierarchical response model of the industrial park. The compensation price of the load aggregator layer and the industrial user layer is dynamically updated until convergence to the optimal solution, so as to obtain a response scheme that takes into account the interests of the load aggregator and the industrial users.

8. A computer-readable storage medium, characterized in that: The device contains a computer program that, when executed by a processor, implements the method described in any one of claims 1 to 6.

9. An electronic device, characterized in that: The method includes at least one processor and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 6.

Images