Power reuse optimization method, device, medium and product for virtual power plants participating in diversified power services

Abstract

The present application discloses a power reuse optimization method, device, medium and product for a virtual power plant participating in multiple power services, and relates to the field of virtual power plant optimization technology. The method includes: initializing the underlying decision variables of the ice storage system in the virtual power plant at each moment in a preset time period; utilizing a base load host power consumption prediction model, a dual-condition host power consumption prediction model, a power reuse optimization model for the virtual power plant participating in multiple power services and optimization model constraints to cyclically optimize the underlying decision variables at each moment in the initialized preset time period to obtain target underlying decision variables at each moment in the preset time period; utilizing the target underlying decision variables at each moment in the preset time period to control the operation of the ice storage system in the preset time period. The present application improves the net income of the virtual power plant when participating in multiple power services.

Description

Technical Field

[0001] The present application relates to the field of virtual power plant optimization technology, and in particular to a power reuse optimization method, device, medium and product for a virtual power plant participating in multiple power services. Background Art

[0002] As the global energy transition accelerates, virtual power plants, as the core vehicle of the Energy Internet, and their ability to provide diverse power services have become a key technical direction for smart grid development. The integration of cutting-edge deep learning technologies with power system optimization theory to build a collaborative optimization system for virtual power plants that adapts to the complex environment of diverse power services is a growing focus of both academic and industrial research.

[0003] As a hub connecting distributed energy resources and power services, virtual power plants (VPPs) have a crucial role in optimizing decision-making and operating efficiency, directly impacting the economic value of new power systems. Current optimization methods for VPPs in trading spot electricity, green power, and ancillary services primarily involve three key technical steps: resource property modeling, constructing a power service optimization model, and solving the optimization model.

[0004] When modeling resource characteristics, mechanistic modeling approaches are currently widely used, using mathematical equations to characterize the operating characteristics of equipment like photovoltaics and energy storage. However, due to equipment losses and environmental factors, cumulative losses can lead to distortion, causing the actual operating state of the equipment to deviate from the theoretically derived results. Traditional mechanistic modeling approaches are even less applicable for resources like ice storage systems, which lack rigorous and accurate mechanistic models. Nonlinear or linear fitting methods are often used to find operating characteristic curves, resulting in significant errors. Emerging machine learning-based resource modeling methods, such as backpropagation neural networks (BP) and long short-term memory (LSTM) algorithms, have been used for output forecasting. However, they suffer from limitations such as static modeling and fragmentation of heterogeneous sources. Consequently, purely data-driven models are insufficient to meet the resource characteristic modeling needs of virtual power plants.

[0005] In the construction of power service optimization models, mixed integer programming is mostly used. Although the optimization model of virtual power plants participating in separate power grids or auxiliary services has been developed to a relatively mature level, the power reuse optimization method of virtual power plants participating in multiple power services has not been considered.

[0006] In the optimization model solution method link, the mainstream method uses heuristic algorithms or commercial solvers for direct solution. Heuristic algorithms have defects such as uncontrollable solution quality, weak constraint processing capabilities, parameter sensitivity, and black box decision-making. Direct solution using commercial solvers is limited by model complexity and has problems such as high difficulty in solving high-dimensional models and low solution efficiency. Particularly prominent is the "splitting" problem: on the one hand, the prediction model is separated from the optimization decision, resulting in a cumulative error amplification effect; on the other hand, the data-driven method is decoupled from the physical mechanism model, resulting in a very high risk probability that the decision results will violate the grid security constraints. These defects seriously restrict the profit space of virtual power plants in diversified power services.

[0007] In summary, a power reuse optimization method for virtual power plants to participate in diversified power services is needed to solve the problem of reduced profits caused by the deviation between the virtual power plant's reliance on the mechanism model and the actual working conditions. Summary of the Invention

[0008] The purpose of this application is to provide a power reuse optimization method, device, medium and product for a virtual power plant to participate in diversified power services, so as to solve the problem of reduced profits caused by the virtual power plant relying solely on the mechanism model and the deviation from the actual working conditions.

[0009] To achieve the above objectives, this application provides the following solutions.

[0010] In a first aspect, the present application provides a method for optimizing power reuse in which a virtual power plant participates in multiple power services, comprising:

[0011] Initialize the underlying decision variables of the ice storage system in the virtual power plant at each moment in the preset time period;

[0012] Using the baseload host power consumption prediction model, the dual-operating host power consumption prediction model, the power reuse optimization model for virtual power plants participating in multiple power services, and the optimization model constraints, the underlying decision variables at each moment in the initialized preset time period are cyclically optimized to obtain the target underlying decision variables at each moment in the preset time period;

[0013] The target underlying decision variables at each moment in the preset time period are used to control the operation of the ice storage system in the preset time period.

[0014] In one embodiment, the ice storage system includes: a cold station part, which includes: a cooling tower, a cooling water pump, a chilled water pump, a base load host and a dual-mode host; the underlying decision variables include: the speed ratio of the fan in the cooling tower, the decision variables of the base load host and the decision variables of the dual-mode host; the decision variables of the base load host include: chilled water flow, cooling water flow, instantaneous cooling power of the base load host, cooling water temperature flowing into the base load host, cooling water temperature flowing out of the base load host, chilled water temperature flowing into the base load host and chilled water temperature flowing out of the base load host; the decision variables of the dual-mode host include: ethylene glycol flow, cooling water flow, cooling water temperature flowing into the dual-mode host, cooling water temperature flowing out of the dual-mode host, ethylene glycol temperature flowing into the dual-mode host and ethylene glycol temperature flowing out of the dual-mode host.

[0015] In one embodiment, the process of determining the base station host power consumption prediction model includes:

[0016] Initialize the MHSA-PINN model; the MHSA-PINN model is a model that combines a multi-head self-attention mechanism with a physical information neural network;

[0017] Obtaining a first training feature dataset, a first training target dataset, and a working mechanism constraint set of the base host; the first training feature dataset includes: actual values of the training feature decision variables of the base host at multiple first historical moments; the first training target dataset includes: actual values of the power consumption of the base host at each first historical moment;

[0018] The MHSA-PINN model is trained using the first training feature data set, the first training target data set, and the working mechanism constraint set of the base host to obtain the base host power consumption prediction model.

[0019] In one embodiment, the process of determining the dual-mode host power consumption prediction model includes:

[0020] Initialize the MHSA-PINN model; the MHSA-PINN model is a model that combines a multi-head self-attention mechanism with a physical information neural network;

[0021] Obtaining a second training feature dataset, a second training target dataset, and a working mechanism constraint set of the dual-mode host; the second training feature dataset includes: actual values of the training feature decision variables of the dual-mode host at multiple second historical moments; the second training target dataset includes: actual values of the power consumption of the dual-mode host at each second historical moment;

[0022] The MHSA-PINN model is trained using the second training feature data set, the second training target data set, and the working mechanism constraint set of the dual-mode host to obtain the dual-mode host power consumption prediction model.

[0023] In one embodiment, the power reuse optimization model includes:

[0024] ;

[0025] ;

[0026] ;

[0027] ;

[0028] in, is the net benefit of the virtual power plant; is the total revenue of the virtual power plant; is the total cost of the virtual power plant; is the total number of moments in the preset period; for Moment , The electric power of distributed photovoltaics in the virtual power plant used to participate in green electricity trading; for The electricity price of green electricity trading at the moment; is the time interval; for Moment , Regulate electric power for ice storage system to participate in peak load auxiliary service; for The compensation price for peak-shaving ancillary services at the given moment; for Moment , is the electrical energy power from the grid; for The electricity selling price of the power grid at that moment.

[0029] In one embodiment, the optimization model constraints include:

[0030] ;

[0031] ;

[0032] ;

[0033] ;

[0034] ;

[0035] ;

[0036] ;

[0037] ;

[0038] ;

[0039] ;

[0040] ;

[0041] ;

[0042] ;

[0043] ;

[0044] in, is the power consumption of the cooling tower; is the speed ratio of the fan in the cooling tower; The maximum power consumption of the cooling tower; is the power consumption of the cooling water pump; is the density of water; is the acceleration due to gravity; is the cooling water flow rate; is the cooling water pump head; The working efficiency of the cooling water pump; is the motor efficiency; is the inverter efficiency; is the power consumption of the chilled water pump; is the chilled water flow rate; is the chilled water pump head; The working efficiency of the chilled water pump; for The ice storage capacity of the ice storage tank at any time; for The ice storage capacity of the ice storage tank at any time; is the heat loss coefficient of the ice storage tank; is ice storage efficiency; for The storage cooling power at all times; for The melting power of time; For ice melting efficiency; is the instantaneous heat exchange power of the plate heat exchanger; is the overall heat transfer coefficient of the plate heat exchanger; is the heat transfer area of the plate heat exchanger; is the heat transfer temperature difference of the fluid on both sides of the plate heat exchanger; It is the total delay of chilled water delivery; Chiller delay for chilled water delivery; Delay for riser pipes delivering chilled water; Plane delay for chilled water delivery; Chiller pipe line for chilled water delivery; The chiller delivery flow rate of chilled water; Riser pipe for chilled water delivery; The riser delivery flow rate of chilled water; A flat pipe system for chilled water transportation; is the plane delivery velocity of chilled water; The temperature rise of the water pump during the chilled water transportation process; is the number of chilled water pumps in series; The ratio of heat generated by the water pump to power consumption; is the specific heat of water; For the room The power consumption of central air conditioner; is the outdoor temperature; For the room The set temperature of the central air conditioner; is the energy efficiency ratio of central air conditioning; is the equivalent thermal resistance of the central air conditioner; is the power generation capacity of distributed photovoltaics; The electric power delivered to the ice storage system by distributed photovoltaics; is the power consumption of the ice storage system; is the power load caused by the terminal temperature control load; is the maximum power generation of distributed photovoltaics; is the power consumption of the base host; The power consumption of the dual-operating host; The cooling capacity provided for the base load host; The cooling capacity provided for direct cooling of dual-mode host; is the cooling loss caused by pipeline transmission; The cooling capacity required for the terminal temperature control load.

[0045] In one embodiment, the optimization process of the underlying decision variables at each moment in a preset time period under any current number of cycles includes:

[0046] Determine whether the current loop is the first time;

[0047] If the current loop is the first time, then:

[0048] Determine the initial predicted value of the objective function based on the baseload host power consumption prediction model, the dual-mode host power consumption prediction model, the power reuse optimization model, the optimization model constraints, and the underlying decision variables at each moment in the initialized preset time period;

[0049] Determine the initial predicted value of the objective function as the initial optimal value of the objective function;

[0050] If the current loop is not the first time, then:

[0051] Obtain the underlying decision variables at each moment in the preset time period under the current number of cycles, and determine the predicted value of the objective function under the current number of cycles based on the baseload host power consumption prediction model, the dual-mode host power consumption prediction model, the power reuse optimization model, the optimization model constraints, and the underlying decision variables at each moment in the preset time period under the current number of cycles;

[0052] Determine whether a stopping condition is met; the stopping condition is that the current number of cycles reaches the maximum number of cycles or the function difference under the current number of cycles is less than a preset value; the function difference under the current number of cycles is the difference between the predicted value of the objective function under the current number of cycles and the optimal value of the objective function under the previous number of cycles;

[0053] If the stopping condition is met, the underlying decision variables at each moment in the preset time period under the current number of cycles are determined as the target underlying decision variables at each moment in the preset time period.

[0054] In the second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored on the memory and runnable on the processor, wherein the processor executes the computer program to implement any of the above-mentioned methods for optimizing the power reuse of a virtual power plant participating in diversified power services.

[0055] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for optimizing the power reuse of a virtual power plant participating in diversified power services.

[0056] In a fourth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the power reuse optimization method for a virtual power plant participating in multiple power services as described in any of the above items.

[0057] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0058] The present application discloses a power reuse optimization method, device, medium and product for a virtual power plant to participate in multiple power services. First, the underlying decision variables of the ice storage system in the virtual power plant at each moment in a preset time period are initialized; then, the underlying decision variables at each moment in the initialized preset time period are cyclically optimized using the baseload host power consumption prediction model, the dual-condition host power consumption prediction model, the power reuse optimization model for the virtual power plant to participate in multiple power services and the optimization model constraints, to obtain the target underlying decision variables at each moment in the preset time period; finally, the target underlying decision variables at each moment in the preset time period are used to control the operation of the ice storage system in the preset time period. The present application optimizes the underlying decision variables by integrating the baseload host power consumption prediction model, the dual-condition host power consumption prediction model, the power reuse optimization model for the virtual power plant to participate in multiple power services and the optimization model constraints, thereby solving the problem of reduced profits caused by the virtual power plant relying solely on the mechanism model and the deviation from the actual working conditions, and improving the net income of the virtual power plant when participating in multiple power services. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0060] Figure 1 A flowchart of a power reuse optimization method for a virtual power plant participating in multiple power services provided in one embodiment of the present application.

[0061] Figure 2 Architecture diagram of the power reuse optimization method for virtual power plants to participate in diversified power services.

[0062] Figure 3 Schematic diagram of an exemplary application scenario of a power reuse optimization method for virtual power plants participating in diversified power services.

[0063] Figure 4 Schematic diagram of the ice storage system structure.

[0064] Figure 5 Schematic diagram of the MHSA-PINN model structure.

[0065] Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0067] The purpose of this application is to provide a power reuse optimization method, device, medium and product for a virtual power plant to participate in diversified power services, aiming to solve the problem of reduced profits of virtual power plants.

[0068] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0069] In an exemplary embodiment, Figure 1 and Figure 2 As shown in FIG, a power reuse optimization method for a virtual power plant to participate in multiple power services is provided. The power reuse optimization method using a virtual power plant to participate in multiple power services is applied to Figure 3 In the environment shown, the virtual power plant participates in the power reuse optimization method of multiple power services, including steps 1 to 3.

[0070] Step 1: Initialize the underlying decision variables of the ice storage system in the virtual power plant at each moment in the preset period.

[0071] As an optional implementation, Figure 4 As shown, the ice storage system includes: a cold station part, which includes: a cooling tower, a cooling water pump, a chilled water pump, a base load host and a dual-mode host; the underlying decision variables include: the speed ratio of the fan in the cooling tower, the decision variables of the base load host and the decision variables of the dual-mode host; the decision variables of the base load host include: chilled water flow, cooling water flow, instantaneous cooling power of the base load host, cooling water temperature flowing into the base load host, cooling water temperature flowing out of the base load host, chilled water temperature flowing into the base load host and chilled water temperature flowing out of the base load host; the decision variables of the dual-mode host include: ethylene glycol flow, cooling water flow, cooling water temperature flowing into the dual-mode host, cooling water temperature flowing out of the dual-mode host, ethylene glycol temperature flowing into the dual-mode host and ethylene glycol temperature flowing out of the dual-mode host.

[0072] Specifically, in addition to the cold station, the ice storage system also includes a pipeline network and terminal temperature control loads. In addition to the cooling tower, cooling water pumps, chilled water pumps, baseload mainframes, and dual-mode mainframes, the cold station also includes an ice storage tank and a plate heat exchanger. The cold station consists of three circuits: a cooling water circuit, an ethylene glycol circuit, and a chilled water circuit. The cooling tower cools municipal tap water and high-temperature cooling water to produce low-temperature cooling water. This low-temperature cooling water is pumped to the chillers (including the baseload mainframe and dual-mode mainframes) via cooling water pumps. The chillers absorb the cold energy of the low-temperature cooling water through condensation of the refrigerant in the condenser and return the high-temperature cooling water to the cooling tower, completing the cooling water circuit. The dual-mode main unit in the chiller releases cold energy through refrigerant evaporation in the evaporator, cooling high-temperature ethylene glycol to low-temperature ethylene glycol. Depending on the on / off status of the ice-making control valve, the low-temperature ethylene glycol flows to the plate heat exchanger directly for refrigeration or to the ice storage tank for ice production. Finally, the high-temperature ethylene glycol is returned to the dual-mode main unit, forming an ethylene glycol cycle. The baseload main unit in the chiller releases cold energy through refrigerant evaporation in the evaporator. The plate heat exchanger and ice-melting refrigeration plate heat exchanger absorb the cold energy released by the low-temperature ethylene glycol, cooling the high-temperature chilled water to low-temperature chilled water. The chilled water pump transports the low-temperature chilled water through the pipeline network to the terminal temperature-controlled load. After releasing the cold energy, it becomes high-temperature chilled water and is returned to the baseload main unit and plate heat exchanger, forming a chilled water cycle.

[0073] The pipe network consists of transmission pipes that transport low-temperature chilled water over long distances and return high-temperature chilled water.

[0074] The terminal temperature control load part is composed of central air conditioning. By forming an adjustable temperature range according to the comfort temperature set by the user, it is reflected as an adjustable range of load cooling power, which in turn affects the adjustable electric power capacity of the entire ice storage system.

[0075] Step 2: Using the baseload host power consumption prediction model, the dual-condition host power consumption prediction model, the power reuse optimization model for virtual power plants participating in diversified power services, and the optimization model constraints, the underlying decision variables at each moment in the initialized preset time period are cyclically optimized to obtain the target underlying decision variables at each moment in the preset time period.

[0076] As an optional implementation manner, the process of determining the base host power consumption prediction model includes steps 211 to 213.

[0077] Step 211: Initialize the MHSA-PINN model; the MHSA-PINN model is a model that combines a multi-head self-attention mechanism with a physical information neural network.

[0078] Step 212: Obtain a first training feature data set, a first training target data set, and a working mechanism constraint set of the base host; the first training feature data set includes: actual values of the training feature decision variables of the base host at multiple first historical moments; the first training target data set includes: actual values of the power consumption of the base host at each first historical moment.

[0079] Step 213: Using the first training feature data set, the first training target data set, and the working mechanism constraint set of the base host, the MHSA-PINN model is trained to obtain a base host power consumption prediction model.

[0080] Specifically, the actual values of the training feature decision variables of the base host at each first historical moment in the first training feature data set are input into the MHSA-PINN model in the form of a matrix as an input feature matrix, and the actual values of the power consumption of the base host at each first historical moment in the first training target data set are input into the MHSA-PINN model in the form of a matrix as an input target matrix.

[0081] As an optional implementation manner, the process of determining the dual-operating-mode host power consumption prediction model includes steps 221 to 223.

[0082] Step 221: Initialize the MHSA-PINN model; the MHSA-PINN model is a model that combines a multi-head self-attention mechanism with a physical information neural network.

[0083] Step 222: Obtain a second training feature data set, a second training target data set, and a working mechanism constraint set of the dual-condition host; the second training feature data set includes: the actual values of the training feature decision variables of the dual-condition host at multiple second historical moments; the second training target data set includes: the actual value of the power consumption of the dual-condition host at each second historical moment.

[0084] Specifically, the actual values of the training feature decision variables of the dual-condition host at each second historical moment in the second training feature data set are input into the MHSA-PINN model in the form of a matrix as an input feature matrix, and the actual values of the power consumption of the dual-condition host at each second historical moment in the second training target data set are input into the MHSA-PINN model in the form of a matrix as an input target matrix.

[0085] Step 223: Using the second training feature data set, the second training target data set, and the working mechanism constraint set of the dual-mode host, the MHSA-PINN model is trained to obtain a dual-mode host power consumption prediction model.

[0086] Specifically, such as Figure 5As shown in Figure 1, the MHSA-PINN model includes an encoder module and a knowledge-data hybrid residual module. The encoder module is implemented based on the Multi-Head Self-Attention (MHSA) mechanism, and the knowledge-data hybrid residual module is implemented based on the Physics-Informed Neural Networks (PINN). The encoder module is responsible for identifying the positional feature information of the input feature sequence (corresponding to the decision variables of the baseload host in the input baseload host power consumption prediction model, corresponding to the decision variables of the dual-condition host in the input dual-condition host power consumption prediction model) and the output target fitting sequence (corresponding to the predicted value of the baseload host's electric power in the input baseload host power consumption prediction model). The knowledge-data hybrid residual module is responsible for comparing the target fitting sequence output by the encoder module with the true target sequence (corresponding to the actual value of the baseload host's electric power in the input baseload host power consumption prediction model, corresponding to the actual value of the electric power of the dual-condition host in the input dual-condition host power consumption prediction model). At the same time, the collaborative mechanism model (corresponding to the working mechanism constraint set of the baseload host in the baseload host power consumption prediction model, corresponding to the working mechanism constraint set of the dual-condition host in the dual-condition host power consumption prediction model) considers the fitting accuracy, and corrects the internal hyperparameters of the encoder module according to the accuracy feedback to improve the fitting accuracy.

[0087] Generally speaking, the encoder module calculates and captures the correlation between each feature and each position of the input data as follows:

[0088] (1)

[0089] (2)

[0090] (3)

[0091] (4)

[0092] (5)

[0093] (6)

[0094] (7)

[0095] in, is the query matrix; is the input feature matrix, the matrix composed of the input feature sequence; is the time step length of the input feature sequence; is the number of features in the input feature sequence; is the query weight matrix; is the bond matrix; is the key weight matrix; is the value matrix; is the weight matrix of the value; is the attention weight matrix; is the softmax function; is transposed; is the dimension of the query matrix; is the output of the self-attention mechanism; For the Output of the head; For the The query matrix of the head; For the The key matrix of the head; For the The value matrix of the head; is the output of the multi-head self-attention mechanism; For splicing operation; is the output of the first head; This is the output of the second head; is the weight matrix.

[0096] For the self-attention mechanism, the matrix First, generate the query matrix through linear transformation , key matrix Sum Matrix , then, use the dot product to calculate the similarity between the query and the key, and generate the attention weight matrix , the output of the self-attention mechanism By combining the attention weight matrix with the value matrix Multiply and get the output Contains the relationship information between different positions in a single input feature sequence; for the multi-head self-attention mechanism, the matrix It is decomposed into multiple heads, each head calculates self-attention independently, and the outputs of all heads are then concatenated and linearly transformed to generate the final output of the multi-head self-attention mechanism. Contains the relationship information between different positions in all input feature sequences.

[0097] In the knowledge-data mixed residual module, the calculation formula of the knowledge-data mixed residual is:

[0098] (8)

[0099] (9)

[0100] in, is the knowledge-data mixing residual; is the allocation weight coefficient of the data residual; is the data residual; is the allocation weight coefficient of knowledge residual; is the knowledge residual; is the input target matrix, i.e. the real target sequence; Fit a sequence to the target.

[0101] For the data residuals, the goal is to minimize the fitted target matrix generated by the MHSA-PINN model With the input target matrix The deviation between them; for the knowledge residual, its goal is to enable the MHSA-PINN model to reasonably configure the weight parameters based on the mechanism model, thereby reducing the influence of input features that violate physical principles and improving the accuracy of expression fitting. In the context of fitting the baseload host power consumption prediction model and the dual-condition host power consumption prediction model, the physical constraints are represented by energy conservation. The knowledge residual part in the knowledge-data hybrid residual module in the baseload host power consumption prediction model is filled with the working mechanism constraint set of the baseload host, and the knowledge residual part in the knowledge-data hybrid residual module in the dual-condition host power consumption prediction model is filled with the working mechanism constraint set of the dual-condition host. The working mechanism constraint set of the baseload host is shown in Equation (10), and the working mechanism constraint set of the dual-condition host is shown in Equation (11).

[0102] (10)

[0103] (11)

[0104] in, The cooling capacity of cooling water exchanged through the base load host; is the specific heat of water; is the quality of cooling water; The temperature of cooling water flowing out of the base load host; is the cooling water temperature flowing into the base load host; It is the cooling capacity of chilled water exchanged through the base load host; is the quality of chilled water; is the chilled water temperature flowing into the base load host; is the chilled water temperature flowing out of the base load host; is the power consumption of the base host; is the actual cooling coefficient of the base load host; is the minimum cooling coefficient of the base load host; is the maximum cooling coefficient of the base load host; It is the cooling capacity exchanged by cooling water through dual-mode host; The cooling water temperature flowing out of the dual-mode main engine; is the cooling water temperature flowing into the dual-mode main engine; It is the cooling capacity exchanged by ethylene glycol through dual-mode main engine; is the specific heat of ethylene glycol; is the mass of ethylene glycol; is the glycol temperature flowing into the dual-mode main engine; is the glycol temperature flowing out of the dual-mode main engine; The power consumption of the dual-operating host; is the actual refrigeration coefficient of the dual-mode host; is the minimum cooling coefficient of the dual-mode host; It is the maximum cooling coefficient of the dual-mode host.

[0105] The expression of the base host power consumption prediction model is:

[0106] (12)

[0107] in, is the predicted value of the power consumption of the base host; is a mapping function of a base host power consumption prediction model; is the instantaneous cooling power of the base load host; is the cooling water flow rate; is the chilled water flow rate.

[0108] The expression of the dual-condition main engine power consumption prediction model is:

[0109] (13)

[0110] in, is the predicted value of power consumption of the dual-operating host; It is the mapping function of the dual-operating-condition host power consumption prediction model; It is the instantaneous cooling power of the dual-mode host.

[0111] Step 213 specifically includes: the encoder module first decomposes the first training feature dataset into multiple single feature sequences and maps them to different subspaces, capturing the dependencies between the feature sequences at different positions. The encoder module then maps the predicted value of the base host's power consumption through a feed-forward fully connected layer. The knowledge-data hybrid residual module compares the predicted value of the base host's power consumption with the training target dataset and calculates the mean squared error to obtain the data residual. The knowledge-data hybrid residual module calculates the knowledge residual based on the base host's working mechanism constraint set, the first training feature dataset, and the first training target dataset. The knowledge-data hybrid residual module performs a weighted sum of the data residual and the knowledge residual to obtain the knowledge-data hybrid residual, which is returned to the encoder module for parameter update. The MHSA-PINN model is trained in multiple cycles to ultimately obtain and save the base host power consumption prediction model.

[0112] Step 223 specifically includes: the encoder module first decomposes the second training feature dataset into multiple single feature sequences and maps them to different subspaces, capturing the dependencies between the feature sequences at different positions. The encoder module then maps the predicted power consumption value of the dual-mode main engine through a feed-forward fully connected layer. The knowledge-data hybrid residual module compares the predicted power consumption value of the dual-mode main engine with the training target dataset and calculates the mean squared error to obtain the data residual. The knowledge-data hybrid residual module calculates the knowledge residual based on the dual-mode main engine's working mechanism constraint set, the second training feature dataset, and the second training target dataset. The knowledge-data hybrid residual module performs a weighted sum of the data residual and the knowledge residual to obtain the knowledge-data hybrid residual, which is returned to the encoder module for parameter update. The MHSA-PINN model is trained in multiple cycles to ultimately obtain and save the dual-mode main engine power consumption prediction model.

[0113] As an optional implementation, the power reuse optimization model includes:

[0114] (14)

[0115] (15)

[0116] (16)

[0117] (17)

[0118] in, is the objective function; is the net benefit of the virtual power plant; is the total revenue of the virtual power plant; is the total cost of the virtual power plant; is the total number of moments in the preset period; for Moment , The electric power of distributed photovoltaics in the virtual power plant used to participate in green electricity trading; for The electricity price of green electricity trading at the moment; is the time interval; for Moment , Regulate electric power for ice storage system to participate in peak load auxiliary service; for The compensation price for peak-shaving ancillary services at the given moment; for Moment , is the electrical energy power from the grid; for The electricity selling price of the power grid at that moment.

[0119] Specifically, the objective function is to maximize the net benefits of virtual power plants participating in diversified power services.

[0120] As an optional implementation, optimizing model constraints includes:

[0121] (18)

[0122] (19)

[0123] (20)

[0124] (twenty one)

[0125] (twenty two)

[0126] (twenty three)

[0127] (twenty four)

[0128] (25)

[0129] (26)

[0130] (27)

[0131] (28)

[0132] (29)

[0133] (30)

[0134] (31)

[0135] in, is the power consumption of the cooling tower; is the speed ratio of the fan in the cooling tower; The maximum power consumption of the cooling tower; is the power consumption of the cooling water pump; is the density of water; is the acceleration due to gravity; is the cooling water flow rate; is the cooling water pump head; The working efficiency of the cooling water pump; is the motor efficiency; is the inverter efficiency; is the power consumption of the chilled water pump; is the chilled water flow rate; is the chilled water pump head; The working efficiency of the chilled water pump; for The ice storage capacity of the ice storage tank at any time; for The ice storage capacity of the ice storage tank at any time; is the heat loss coefficient of the ice storage tank; is ice storage efficiency; for The storage cooling power at all times; for The melting power of time; For ice melting efficiency; is the instantaneous heat exchange power of the plate heat exchanger; is the overall heat transfer coefficient of the plate heat exchanger; is the heat transfer area of the plate heat exchanger; is the heat transfer temperature difference of the fluid on both sides of the plate heat exchanger; It is the total delay of chilled water delivery; Chiller delay for chilled water delivery; Delay for riser pipes delivering chilled water; Plane delay for chilled water delivery; Chiller pipe line for chilled water delivery; The chiller delivery flow rate of chilled water; Riser pipe for chilled water delivery; The riser delivery flow rate of chilled water; A flat pipe system for chilled water transportation; is the plane delivery velocity of chilled water; The temperature rise of the water pump during the chilled water transportation process; is the number of chilled water pumps in series; The ratio of heat generated by the water pump to power consumption; is the specific heat of water; For the room The power consumption of central air conditioner; is the outdoor temperature; For the room The set temperature of the central air conditioner; is the energy efficiency ratio of central air conditioning; is the equivalent thermal resistance of the central air conditioner; is the power generation capacity of distributed photovoltaics; The electric power delivered to the ice storage system by distributed photovoltaics; is the power consumption of the ice storage system; is the power load caused by the terminal temperature control load; is the maximum power generation of distributed photovoltaics; is the power consumption of the base host; The power consumption of the dual-operating host; The cooling capacity provided for the base load host; The cooling capacity provided for direct cooling of dual-mode host; is the cooling loss caused by pipeline transmission; The cooling capacity required for the terminal temperature control load.

[0136] Specifically, Equation (18) represents the working model of the cooling tower, Equation (19) represents the working model of the cooling water pump, Equation (20) represents the working model of the chilled water pump, Equation (21) represents the working model of the ice storage tank, Equation (22) represents the working model of the plate heat exchanger, the pipe network model is composed of the pipe network time delay and temperature rise represented by Equations (23) and (24), and the terminal temperature control load model is represented by Equation (25). Equation (26) is the complementary synergy constraint between the distributed photovoltaic and ice storage systems, and Equation (27) is the power reuse constraint. Equations (26) and (27) together indicate that distributed photovoltaic power generation can supply electricity to the ice storage system and participate in green power trading and electricity sales; Equations (28), (29) and (31) are the energy interaction constraints between the virtual power plant and the multi-market, indicating that the ice storage system needs to purchase electricity from the grid to meet the load demand, and the ice storage system participates in peak load auxiliary services through ice storage and ice melting; Equation (30) is the total power consumption constraint of the ice storage system.

[0137] Specifically, the electric energy produced by distributed photovoltaics participates in peak-shaving auxiliary services by being transmitted to the ice storage system; at the same time, part of the green electricity produced participates in green electricity trading, realizing the coordinated complementarity and power reuse of distributed photovoltaics and ice storage systems.

[0138] As an optional implementation, the optimization process of the underlying decision variables at each moment in a preset time period under any current number of cycles includes:

[0139] Determine whether the current loop is the first time.

[0140] If the current loop is the first time, then:

[0141] Based on the baseload host power consumption prediction model, the dual-mode host power consumption prediction model, the power reuse optimization model, the optimization model constraints and the underlying decision variables at each moment in the initialized preset time period, the initial prediction value of the objective function is determined.

[0142] The initial predicted value of the objective function is determined as the initial optimal value of the objective function.

[0143] If the current loop is not the first time, then:

[0144] Obtain the underlying decision variables at each moment in the preset time period under the current number of cycles, and determine the predicted value of the objective function under the current number of cycles based on the base load host power consumption prediction model, the dual-condition host power consumption prediction model, the power reuse optimization model, the optimization model constraints, and the underlying decision variables at each moment in the preset time period under the current number of cycles.

[0145] Determine whether the stopping condition is met; the stopping condition is that the current number of loops reaches the maximum number of loops or the function difference under the current number of loops is less than the preset value; the function difference under the current number of loops is the difference between the predicted value of the objective function under the current number of loops and the optimal value of the objective function under the previous number of loops.

[0146] If the stopping condition is met, the underlying decision variables at each moment in the preset time period under the current number of cycles are determined as the target underlying decision variables at each moment in the preset time period.

[0147] Step 3: Use the target underlying decision variables at each moment in the preset period to control the operation of the ice storage system in the preset period.

[0148] It can be seen from the above embodiments that the advantages of the present application are as follows.

[0149] (1) The present invention proposes a refined modeling method for ice storage system in virtual power plant based on MHSA-PINN. For the component structure with clear and perfect mechanism model, a traditional mechanism model is established. For the component structure without clear and perfect mechanism model (i.e. base load host and dual-operating mode host), an MHSA-PINN is invented. The encoder module based on multi-head self-attention mechanism inside MHSA-PINN is used to capture the feature relationship of input historical data, search for reasonable optimal relationship model, break through the shackles of traditional fuzzy mechanism model, and search for more accurate relationship model. The knowledge inside MHSA-PINN is used to The knowledge-data hybrid residual module evaluates and optimizes the encoder output model, providing feedback to the encoder module based on the multi-head self-attention mechanism to optimize internal hyperparameters, thereby enabling MHAS-PINN to output a more optimal model. The knowledge residual characterizes the output model's adherence to traditional mechanism constraints, ensuring that the output model adheres to the most basic working mechanism constraints. The data residual characterizes the output model's accuracy in representing resources. The weighted sum of the knowledge residual and the data residual constitutes the knowledge-data hybrid residual. By properly assigning weights, the degree to which the MHSA-PINN output relationship model focuses on the working mechanism and accuracy can be controlled. Overall, the refined modeling method for the internal ice storage system of a virtual power plant based on deep learning MHSA-PINN breaks through the constraints of traditional modeling methods based on working mechanisms, achieving greater simplicity and accuracy while ensuring compliance with the basic working mechanism constraints.

[0150] (2) The power reuse optimization method for the virtual power plant participating in the diversified power services of the present application first establishes the complementary and coordinated relationship constraints of the various resources and power reuse constraints within the virtual power plant, so that the various resources and energies can flow and complement each other, so that the various resources and energies can participate in the diversified power services simultaneously or gradually, thereby improving the energy utilization efficiency; then, the energy interaction constraints for the virtual power plant participating in the diversified power services are established, so that the virtual power plant can participate in various power energy transactions such as green power transactions and peak-shaving auxiliary services at the same time, supplemented by the complementary and coordinated resources and power reuse within the virtual power plant, which can maximize the reduction of the virtual power plant operation and maintenance costs, increase the profits of the virtual power plant participating in the diversified power services, reduce carbon emissions, and protect the environment.

[0151] (3) When optimizing the underlying decision variables, this application calls the MHSA-PINN model multiple times during the optimization solution process. Compared with the traditional optimization solution method, the MHSA-PINN model is embedded in the loop optimization process, which simplifies the solution steps, improves the solution efficiency, and makes the optimized target underlying decision variables more accurate.

[0152] In an exemplary embodiment, a computer device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a power reuse optimization method for a virtual power plant participating in multiple power services.

[0153] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, a power reuse optimization method for a virtual power plant participating in multiple power services is implemented.

[0154] In an exemplary embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements a power reuse optimization method for a virtual power plant participating in multiple power services.

[0155] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a power reuse optimization method for a virtual power plant to participate in multiple power services is implemented.

[0156] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0157] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0158] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0159] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0160] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0161] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for optimizing power reuse of a virtual power plant participating in multiple power services, characterized in that: The power reuse optimization method of the virtual power plant participating in multiple power services includes: Initialize the underlying decision variables of the ice storage system in the virtual power plant at each moment in the preset time period; Using the baseload host power consumption prediction model, the dual-operating host power consumption prediction model, the power reuse optimization model for virtual power plants participating in multiple power services, and the optimization model constraints, the underlying decision variables at each moment in the initialized preset time period are cyclically optimized to obtain the target underlying decision variables at each moment in the preset time period; Using the target underlying decision variables at each moment in the preset time period, the ice storage system is controlled to operate during the preset time period; The base load host power consumption prediction model and the dual-operating-mode host power consumption prediction model are both obtained by training the MHSA-PINN model; The MHSA-PINN model consists of an encoder module and a knowledge-data hybrid residual module. The encoder module is implemented based on a multi-head self-attention mechanism, and the knowledge-data hybrid residual module is implemented based on a physical information neural network. The encoder module is responsible for identifying the position feature information of the input feature sequence and outputting the target fitting sequence; the position feature information of the input feature sequence is the underlying decision variable of the base load host or the underlying decision variable of the dual-mode host; the output target fitting sequence is the predicted value of the electric power of the base load host or the predicted value of the electric power of the dual-mode host; The knowledge-data hybrid residual module is responsible for comparing the target fitting sequence output by the encoder module with the actual target sequence. At the same time, the collaborative mechanism model takes into account the fitting accuracy and corrects the internal hyperparameters of the encoder module based on accuracy feedback. The actual target sequence is the actual value of the electric power of the baseload host or the actual value of the electric power of the dual-condition host. The collaborative mechanism model is the working mechanism constraint set of the baseload host or the working mechanism constraint set of the dual-condition host.

2. The power reuse optimization method for a virtual power plant participating in multiple power services according to claim 1, characterized in that: The ice storage system includes: a cold station part, which includes: a cooling tower, a cooling water pump, a chilled water pump, a base load host and a dual-mode host; the underlying decision variables include: the speed ratio of the fan in the cooling tower, the decision variables of the base load host and the decision variables of the dual-mode host; the decision variables of the base load host include: chilled water flow, cooling water flow, instantaneous cooling power of the base load host, cooling water temperature flowing into the base load host, cooling water temperature flowing out of the base load host, chilled water temperature flowing into the base load host and chilled water temperature flowing out of the base load host; the decision variables of the dual-mode host include: ethylene glycol flow, cooling water flow, cooling water temperature flowing into the dual-mode host, cooling water temperature flowing out of the dual-mode host, ethylene glycol temperature flowing into the dual-mode host and ethylene glycol temperature flowing out of the dual-mode host.

3. The power reuse optimization method for a virtual power plant participating in multiple power services according to claim 2, characterized in that: The process of determining the baseload host power consumption prediction model includes: Initialize the MHSA-PINN model; the MHSA-PINN model is a model that combines a multi-head self-attention mechanism with a physical information neural network; Obtaining a first training feature dataset, a first training target dataset, and a working mechanism constraint set of the base host; the first training feature dataset includes: actual values of the training feature decision variables of the base host at multiple first historical moments; the first training target dataset includes: actual values of the power consumption of the base host at each first historical moment; The MHSA-PINN model is trained using the first training feature data set, the first training target data set, and the working mechanism constraint set of the base host to obtain the base host power consumption prediction model.

4. The power reuse optimization method for a virtual power plant participating in multiple power services according to claim 2, characterized in that: The process of determining the dual-operating-mode host power consumption prediction model includes: Initialize the MHSA-PINN model; the MHSA-PINN model is a model that combines a multi-head self-attention mechanism with a physical information neural network; Obtaining a second training feature dataset, a second training target dataset, and a working mechanism constraint set of the dual-mode host; the second training feature dataset includes: actual values of the training feature decision variables of the dual-mode host at multiple second historical moments; the second training target dataset includes: actual values of the power consumption of the dual-mode host at each second historical moment; The MHSA-PINN model is trained using the second training feature data set, the second training target data set, and the working mechanism constraint set of the dual-mode host to obtain the dual-mode host power consumption prediction model.

5. The power reuse optimization method for virtual power plants participating in multiple power services according to claim 2, characterized in that: The power reuse optimization model includes: ； ； ； ； in, is the objective function; is the net benefit of the virtual power plant; is the total revenue of the virtual power plant; is the total cost of the virtual power plant; is the total number of moments in the preset period; for Moment , The electric power of distributed photovoltaics in the virtual power plant used to participate in green electricity trading; for The electricity price of green electricity trading at the moment; is the time interval; for Moment , Regulate electric power for ice storage system to participate in peak load auxiliary service; for The compensation price for peak-shaving ancillary services at the given moment; for Moment , is the electrical energy power from the grid; for The electricity selling price of the power grid at that moment.

6. The power reuse optimization method for a virtual power plant participating in multiple power services according to claim 5, characterized in that: The optimization model constraints include: ； ； ； ； ； ； ； ； ； ； ； ； ； ； in, is the power consumption of the cooling tower; is the speed ratio of the fan in the cooling tower; The maximum power consumption of the cooling tower; is the power consumption of the cooling water pump; is the density of water; is the acceleration due to gravity; is the cooling water flow rate; is the cooling water pump head; The working efficiency of the cooling water pump; is the motor efficiency; is the inverter efficiency; is the power consumption of the chilled water pump; is the chilled water flow rate; is the chilled water pump head; The working efficiency of the chilled water pump; for The ice storage capacity of the ice storage tank at any time; for The ice storage capacity of the ice storage tank at any time; is the heat loss coefficient of the ice storage tank; is ice storage efficiency; for The storage cooling power at all times; for The melting power of time; For ice melting efficiency; is the instantaneous heat exchange power of the plate heat exchanger; is the overall heat transfer coefficient of the plate heat exchanger; is the heat transfer area of the plate heat exchanger; is the heat transfer temperature difference of the fluid on both sides of the plate heat exchanger; It is the total delay of chilled water delivery; Chiller delay for chilled water delivery; Delay for riser pipes delivering chilled water; Plane delay for chilled water delivery; Chiller pipe line for chilled water delivery; The chiller delivery flow rate of chilled water; Riser pipe for chilled water delivery; The riser delivery flow rate of chilled water; A flat pipe system for chilled water transportation; is the plane delivery velocity of chilled water; The temperature rise of the water pump during the chilled water transportation process; is the number of chilled water pumps in series; The ratio of heat generated by the water pump to power consumption; is the specific heat of water; For the room The power consumption of central air conditioner; is the outdoor temperature; For the room The set temperature of the central air conditioner; is the energy efficiency ratio of central air conditioning; is the equivalent thermal resistance of the central air conditioner; is the power generation capacity of distributed photovoltaics; The electric power delivered to the ice storage system by distributed photovoltaics; is the power consumption of the ice storage system; is the power load caused by the terminal temperature control load; is the maximum power generation of distributed photovoltaics; is the power consumption of the base host; The power consumption of the dual-operating host; The cooling capacity provided for the base load host; The cooling capacity provided for direct cooling of dual-mode host; is the cooling loss caused by pipeline transmission; The cooling capacity required for the terminal temperature control load.

7. The power reuse optimization method for a virtual power plant participating in multiple power services according to claim 6, characterized in that: The optimization process of the underlying decision variables at each moment in the preset period under any current number of cycles includes: Determine whether the current loop is the first time; If the current loop is the first time, then: Determine the initial predicted value of the objective function based on the baseload host power consumption prediction model, the dual-mode host power consumption prediction model, the power reuse optimization model, the optimization model constraints, and the underlying decision variables at each moment in the initialized preset time period; Determine the initial predicted value of the objective function as the initial optimal value of the objective function; If the current loop is not the first time, then: Obtain the underlying decision variables at each moment in the preset time period under the current number of cycles, and determine the predicted value of the objective function under the current number of cycles based on the baseload host power consumption prediction model, the dual-mode host power consumption prediction model, the power reuse optimization model, the optimization model constraints, and the underlying decision variables at each moment in the preset time period under the current number of cycles; Determine whether a stopping condition is met; the stopping condition is that the current number of cycles reaches the maximum number of cycles or the function difference under the current number of cycles is less than a preset value; the function difference under the current number of cycles is the difference between the predicted value of the objective function under the current number of cycles and the optimal value of the objective function under the previous number of cycles; If the stopping condition is met, the underlying decision variables at each moment in the preset time period under the current number of cycles are determined as the target underlying decision variables at each moment in the preset time period.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the power reuse optimization method for a virtual power plant participating in multiple power services as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the power reuse optimization method for a virtual power plant participating in multiple power services as described in any one of claims 1-7.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, it implements the power reuse optimization method for a virtual power plant participating in multiple power services as described in any one of claims 1-7.

Images