Internal micro-grid balance optimization method and system for source network load storage virtual power plant

Abstract

The invention provides an internal micro-grid balance optimization method and system for a source network load storage virtual power plant, and relates to the field of virtual power plants, and the method comprises the steps: obtaining first target power information from a power system, and carrying out the cleaning, standardization, clustering and grading classification processing, and obtaining second target power information; constructing a load model, designing a time sequence network according to the second target power information, and constructing a time sequence model; determining an adjustable variable and an adjustable capability range to obtain a boundary constraint condition; building a double-layer optimization architecture oriented to user requirements and user equipment response scheduling, and constructing a multi-task learning model; and solving the multi-task learning model, and outputting a target solution to adjust the internal micro-grid balance of the virtual power plant. According to the method, the target solution obtained by solving the multi-task learning model can represent the resource allocation instruction required by each device under the optimization target, and then the response strategy is determined to flexibly adjust the user load, so that the multi-target optimization of the source network load storage resources is realized.

Description

Technical Field

[0001] The present application relates to the technical field of virtual power plants, and in particular to an internal microgrid balance optimization method and system for source-grid-load-storage virtual power plants. Background Art

[0002] A virtual power plant is a power coordination management system that uses advanced information and communication technologies and software systems to participate in the power market and power grid operation as a special power plant. The core of the concept of a virtual power plant can be summarized as "communication" and "aggregation". The key technologies of virtual power plants mainly include coordinated control technology, intelligent metering technology, and information and communication technology. Virtual power plants can provide management and auxiliary services for distribution networks and transmission networks. In actual operation, it is crucial to control the balance of energy supply and demand in the microgrid within the virtual power plant. The balance of the microgrid is conducive to maintaining the stability of voltage and frequency, and can also ensure that users' electricity needs can be met at any time.

[0003] At present, virtual power plant technology is in its initial development stage and has some limitations. Among the related technologies, the cost of energy storage systems is relatively high, and energy storage efficiency and life are key factors that restrict performance. It is difficult to dynamically feedback the optimal aggregation of multiple loads and demand response, and it is difficult to accurately predict load changes and optimize scheduling. The degree of technical integration between the source, grid, load and storage links is not high, making it difficult to achieve efficient coordination. Summary of the invention

[0004] In response to the deficiencies in the prior art, the present application provides an internal microgrid balance optimization method and system for a source-grid-load-storage virtual power plant, which solves the problems of insufficient internal microgrid balance and unsatisfactory optimization scheduling faced by current virtual power plant technologies.

[0005] To achieve the above objectives, this application is implemented through the following technical solutions:

[0006] In the first aspect, an embodiment of the present application provides an internal microgrid balance optimization method for a source-grid-load-storage virtual power plant, and the internal microgrid balance optimization method includes: according to the required data type, obtaining first target power information representing historical load data and historical time series data from the power system; cleaning, standardizing, clustering and hierarchical classification of the first target power information to obtain second target power information; according to the second target power information, combined with the random process of fitting Brownian motion, constructing a neural network learning framework, and adjusting hyperparameters for optimization to obtain a load model; designing a timing network according to the second target power information, and constructing a timing model to establish time association; based on the load model and the timing model, determining the adjustable variables and adjustable capacity range corresponding to the optimization target, and obtaining boundary constraints; building a two-layer optimization architecture for user demand and user device response scheduling, and constructing a multi-task learning model based on boundary constraints; solving the multi-task learning model, and outputting the target solution to adjust the internal microgrid balance of the virtual power plant; wherein the target solution is used to characterize the resource allocation instructions required for each device under the optimization target.

[0007] In the second aspect, an embodiment of the present application provides an internal microgrid balancing optimization system for a source-grid-load-storage virtual power plant, and the internal microgrid balancing optimization system includes: an acquisition module, a processing module, a first building module, a second building module, a constraint determination module, a third building module and a solution module.

[0008] Specifically, the acquisition module is used to acquire the first target power information representing the historical load data and the historical time series data from the power system according to the required data type; the processing module is used to clean, standardize, cluster and grade the first target power information to obtain the second target power information; the first construction module is used to construct a neural network learning framework based on the second target power information and the random process of fitting Brownian motion, and adjust the hyperparameters for optimization to obtain the load model; the second construction module is used to design the time series network according to the second target power information, and construct the time series model to establish time association; the constraint determination module is used to determine the adjustable variables and adjustable capacity ranges corresponding to the optimization target based on the load model and the time series model, and obtain the boundary constraints; the third construction module is used to build a two-layer optimization architecture for user demand and user device response scheduling, and construct a multi-task learning model based on the boundary constraints; the solution module is used to solve the multi-task learning model and output the target solution to adjust the internal microgrid balance of the virtual power plant; wherein the target solution is used to characterize the resource allocation instructions required for each device under the optimization target.

[0009] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the internal microgrid balancing optimization method for the source-grid-load-storage virtual power plant in the aforementioned first aspect is implemented.

[0010] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium storing a program or instruction, which, when executed by a processor, implements the internal microgrid balancing optimization method for a source-grid-load-storage virtual power plant in the aforementioned first aspect.

[0011] This application provides an internal microgrid balance optimization method and system for source-grid-load-storage virtual power plants. Compared with the prior art, it has the following beneficial effects:

[0012] Based on the first target power information that characterizes historical load data and historical time series data, this application constructs a load model and a time series model respectively, analyzes the load data and time series data of the power system, and can determine the adjustable variables and adjustable capacity ranges for the set optimization goals, and determine the adjustable potential. In order to effectively dispatch and regulate the source, grid, load and storage resources, and optimize the microgrid balance corresponding to user demand and user equipment, this application constructs a two-layer optimization architecture and establishes a multi-task learning model. Based on the determined adjustable variables and adjustable capacity ranges, the boundary constraints of the multi-task learning model can be determined; the target solution obtained by solving the multi-task learning model can characterize the resource allocation instructions required for each device under the optimization goal, and then determine the response strategy to flexibly adjust the user load, interact between the supply and demand sides, and realize the multi-objective optimization of the source, grid, load and storage resources, which is beneficial to the internal microgrid balance. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0014] Figure 1 It is a flow chart of an internal microgrid balance optimization method for a source-grid-load-storage virtual power plant provided by an embodiment of the present application;

[0015] Figure 2 yes Figure 1 An exemplary flow chart of S120;

[0016] Figure 3 It is a structural schematic diagram of an internal microgrid balancing optimization system for a source-grid-load-storage virtual power plant provided by an embodiment of the present application;

[0017] Figure 4 It is a structural schematic diagram of another internal microgrid balancing optimization system for a source-grid-load-storage virtual power plant provided in an embodiment of the present application;

[0018] Figure 5 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0021] The embodiments of the present application solve the problems of insufficient internal microgrid balance and unsatisfactory optimization scheduling faced by current virtual power plant technologies by providing an internal microgrid balance optimization method and system for a source-grid-load-storage virtual power plant.

[0022] The technical solution in the embodiment of the present application is to solve the above technical problems, and the overall idea is as follows:

[0023] A virtual power plant is a power coordination management system that uses advanced information and communication technologies and software systems to participate in the power market and power grid operation as a special power plant. The core of the concept of a virtual power plant can be summarized as "communication" and "aggregation". The key technologies of virtual power plants mainly include coordinated control technology, intelligent metering technology, and information and communication technology. Virtual power plants can provide management and auxiliary services for distribution networks and transmission networks. In actual operation, it is crucial to control the balance of energy supply and demand in the microgrid within the virtual power plant. The balance of the microgrid is conducive to maintaining the stability of voltage and frequency, and can also ensure that users' electricity needs can be met at any time.

[0024] At present, virtual power plant technology is in its initial development stage and has some limitations. Among the related technologies, the cost of energy storage systems is relatively high, and energy storage efficiency and life are key factors that restrict performance. It is difficult to dynamically feedback the optimal aggregation of multiple loads and demand response, and it is difficult to accurately predict load changes and optimize scheduling. The degree of technical integration between the source, grid, load and storage links is not high, making it difficult to achieve efficient coordination.

[0025] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0026] The following first introduces an internal microgrid balance optimization method for a source-grid-load-storage virtual power plant provided in an embodiment of the present application.

[0027] The present application embodiment provides a flow chart of an internal microgrid balance optimization method for a source-grid-load-storage virtual power plant, such as Figure 1 As shown, the internal microgrid balance optimization method for the source-grid-load-storage virtual power plant may include the following steps S110-S170.

[0028] S110. Acquire first target power information representing historical load data and historical time series data from the power system according to the required data type.

[0029] S120: Clean, standardize, cluster and classify the first target power information to obtain second target power information.

[0030] S130. According to the second target power information, combined with the random process of fitting Brownian motion, a neural network learning framework is constructed, and hyperparameters are adjusted for optimization to obtain a load model.

[0031] S140. Design a timing network according to the second target power information, and construct a timing model to establish time association.

[0032] S150: Based on the load model and the time series model, determine the adjustable variables and adjustable capacity range corresponding to the optimization target to obtain boundary constraint conditions.

[0033] S160. Build a two-layer optimization architecture for user demand and user device response scheduling, and construct a multi-task learning model based on boundary constraints.

[0034] S170. Solve the multi-task learning model and output a target solution to adjust the internal microgrid balance of the virtual power plant; wherein the target solution is used to characterize the resource allocation instructions required for each device under the optimization target.

[0035] The above is a specific implementation method of an internal microgrid balancing optimization method for a source-grid-load-storage virtual power plant provided in an embodiment of the present application. It can be understood that the present application constructs a load model and a time series model based on the first target power information that characterizes historical load data and historical time series data, respectively, and analyzes the load data and time series data of the power system. It can determine the adjustable variables and adjustable capacity range regarding the set optimization target, and determine the adjustable potential.

[0036] Furthermore, in order to effectively dispatch and regulate the source, grid, load and storage resources and optimize the microgrid balance corresponding to user demand and user equipment, the present application constructs a two-layer optimization architecture and establishes a multi-task learning model. The boundary constraints of the multi-task learning model can be determined based on the determined adjustable variables and adjustable capacity range; the target solution obtained by solving the multi-task learning model can characterize the resource allocation instructions required for each device under the optimization target, and then determine the response strategy to flexibly adjust the user load, interact between the supply and demand sides, and realize multi-objective optimization of source, grid, load and storage resources, which is beneficial to the internal microgrid balance.

[0037] In one example, the load model includes a drift network and a diffusion network, wherein the drift network is used to capture the uncertainty in the power grid output data and the load data; the diffusion network is used to capture the uncertainty in the preset load forecasting model; wherein the load forecasting model is used to generate predicted future data.

[0038] In some embodiments, Figure 2 As shown, the first target power information is cleaned, standardized, clustered and graded to obtain the second target power information, that is, the aforementioned S120 may specifically include the following steps:

[0039] S210, cleaning the first target power information by Lagrange interpolation method, eliminating abnormal values ​​and noise of the data, and identifying and correcting errors and missing values ​​to determine the completeness and accuracy of the data, and obtaining first intermediate information;

[0040] S220, calculating a scaling factor based on the mapped target range and the original range, so as to perform data conversion through a preset scaling formula so that the data have the same scale, and obtain second intermediate information;

[0041] S230, using an unsupervised learning method combining contrastive learning and meta-learning, generating positive sample pairs of the same data through SimCLR enhancement, and using different data as negative sample pairs, clustering the second intermediate information to obtain a clustering result;

[0042] S240: According to the clustering result, the load-side adjustable resources are classified and classified to obtain the second target power information.

[0043] In the embodiments of the present application, it can be understood that the first target power information may include load side time series data, user power consumption and power consumption behavior and other related data and potential uncertain interference factor data from the power system survey; the present application uses the Lagrange interpolation method to clean the load curve data, eliminate outliers and noise in the data, and identify and correct errors and missing values ​​in the load curve to ensure the integrity and accuracy of the data.

[0044] In one example, during the data normalization process, the original range of the data can be defined by determining the minimum and maximum values ​​in the data set corresponding to the first target power information; the scaling factor is calculated based on the preset mapping target range and the original range; and the data is converted by calculating the preset scaling formula so that the data has the same scale.

[0045] In some embodiments, the aforementioned design of a timing network according to the second target power information and construction of a timing model to establish a time association, that is, the aforementioned S140 may specifically include the following steps:

[0046] S310, performing time series decomposition, autocorrelation analysis, and feature extraction on the time series data in the second target power information to obtain feature information representing the load data;

[0047] S320, processing the time dependency of feature information through a preset long short-term memory network or a gated recurrent unit, and performing modeling to obtain an initial model;

[0048] S330, selecting a suitable loss function and adjusting network parameters, and training the initial model to obtain a time series model.

[0049] In some embodiments, after designing a timing network according to the second target power information and constructing a timing model to establish a time association, that is, after the aforementioned S140, the internal microgrid balance optimization method for the source-grid-load-storage virtual power plant may also include:

[0050] S141. Define nodes and edges so that each node represents a load unit, and the node position represents the position of the corresponding load unit in the virtual power grid, and the edges represent the relationship between nodes; wherein the weight of the edge is set based on the historical load correlation.

[0051] S142. Use one of the graph convolutional network GCN and the graph attention network GAT as the basic architecture, introduce an incremental learning algorithm, aggregate the information of neighbor nodes through multi-layer graph convolution operations, introduce residual connections and layer normalization to improve the training effect and convergence speed of the model.

[0052] S143. Dynamically adjust the edge weights according to changes in load data, and use the stream processing framework to process and update graph data in real time so that the model can reflect the latest load changes.

[0053] In some embodiments, the aforementioned two-layer optimization architecture for user demand and user device response scheduling is constructed, and a multi-task learning model is constructed based on boundary constraints, that is, the aforementioned S160 may specifically include the following steps:

[0054] S410, building a two-layer optimization architecture based on user demand and user device response scheduling sub-problems, and designing load types for different optimization sub-problems;

[0055] S420, designing a double-layer encoding method for continuous-discrete mixed variables, where the continuous-discrete mixed variables include: load regulation and energy storage regulation continuous variables at the load side user end, and discrete variables for scheduling the next process of each user;

[0056] S430, determining at least one of load-side adjustable capacity, user satisfaction, load-side energy loss, minimization of response cost, and energy storage cost as an optimization target;

[0057] S440. Based on the optimization objective, the variables are restricted by boundary constraints and model constraints are performed to construct a multi-task learning model; the multi-task learning model includes an upper-level user demand sub-model and a lower-level load-side response scheduling sub-model.

[0058] In some embodiments, based on the optimization goal, the variables are restricted by boundary constraints and model constraints are performed to construct a multi-task learning model, that is, the aforementioned S440 may specifically include the following steps:

[0059] S510. Based on the two-layer optimization architecture, the upper-layer continuous decision variables are used as the input of the lower-layer optimization problem, reconstructed into discrete scheduling variables in the form of multi-layer encoding, and regarded as independent scenarios for simultaneous optimization.

[0060] S520. In the decision space, based on the problem characteristics and population distribution, determine whether there are similarities between tasks, and set a similarity measurement index as a benchmark for multi-task learning.

[0061] In some embodiments, after the aforementioned multi-task learning model is constructed based on the optimization objective, by limiting the variables through boundary constraints and performing model constraints, that is, after the aforementioned S440, the internal microgrid balance optimization method for the source-grid-load-storage virtual power plant may also include:

[0062] S441. Based on the weighted and shared network structure of the inter-task loss function, shared transfer learning jointly optimizes all tasks to perform multi-task learning model training;

[0063] S442. Define the shareable knowledge types through the main tasks and auxiliary tasks in the multi-task scenario, and decide whether to transfer local knowledge or overall knowledge to establish a knowledge transfer mechanism;

[0064] S443. Use the model input results after knowledge transfer as the benchmark for environment update and solution selection, select the top-K individuals from the new input individuals and historical individuals respectively as the initial path scheduling selection plan for the next generation, evaluate and feedback the results, judge the effectiveness of the knowledge to realize the dynamic update mechanism of the transferred knowledge.

[0065] In some embodiments, the internal microgrid balancing optimization method for source-grid-load-storage virtual power plants provided in the present application can meet the needs of different usage scenarios; the internal microgrid balancing optimization method for source-grid-load-storage virtual power plants can also include: dynamic environmental change trend estimation, current and historical environmental interactivity relationship detection, and feedback mechanism design.

[0066] Specifically, in the process of estimating the trend of dynamic environmental changes, the present application designs an environmental detection mechanism at fixed time nodes to detect dynamic changes in the target space; the current environment is at time t+1, and time t is used as the measurement benchmark to evaluate the interactive relationship between the environmental uncertainty coupling parameters before and after the moments, and sense the trend of environmental changes.

[0067] In the process of detecting the interactive relationship between the current and historical environments, the present application re-evaluates the differences in regional information contained in multiple uncertain coupling parameters before and after the environment based on the characteristics of uncertain coupling parameters; extracts different environmental features and detects them. If it is detected that there is no intersection between the environments and no significant distribution differences, the current environment is represented as a static optimization problem; if it is detected that there is an intersection between the environments and there are significant differences, the historical environmental features are used as knowledge to migrate the current environment and guide the algorithm solution under the current environment.

[0068] In the process of feedback mechanism design, this application designs a cross-domain migration strategy with a feedback correction mechanism to evaluate the positive and reverse information of the optimization results before and after knowledge migration, and feed back to the source domain task that issues the knowledge; in addition, in order to ensure fairness and impartiality in decision optimization, the model is adjusted and optimized during the training process, the migrated knowledge is corrected and updated, and migrated to other tasks to adapt to tasks of different similarities, effectively improving the quality of problem solving and shortening the task solving time.

[0069] In one example, for cold storage type loads: based on the internal microgrid balancing optimization method for the source-grid-load-storage virtual power plant provided in this application, the refrigeration equipment is operated through energy storage technology during the off-peak electricity price period, and the cold energy is stored in the target temperature range; during the peak electricity price period, the stored cold energy is used to maintain the cold storage temperature and reduce the operating time of the refrigeration equipment; combined with the real-time electricity price signal, the optimal operating strategy of the refrigeration equipment is generated through a dynamic optimization model, and the load distribution is flexibly adjusted to achieve peak shifting and valley filling of electricity consumption and minimize costs.

[0070] In one example, for steam type loads, based on the internal microgrid balancing optimization method for the source-grid-load-storage virtual power plant provided in this application, a multi-energy coordination optimization strategy is determined to achieve flexible adjustment of the steam load and balance between supply and demand; during periods of low electricity prices, excess steam is stored through high-efficiency heat storage devices to reduce peak operating pressure; during periods of peak electricity prices, stored steam is released to meet production or living needs, reducing boiler operating time and smoothing load peaks.

[0071] In one example, for ventilation type loads, based on the internal microgrid balancing optimization method for the source-grid-load-storage virtual power plant provided in this application, an intelligent ventilation control strategy is determined, and efficient energy use and operating cost optimization are achieved by adjusting the operating mode of the ventilation equipment in different time periods; the operating power of the ventilation equipment is increased during low electricity price periods to store cold air or expel hot air; the operating power of the ventilation equipment is reduced during peak electricity price periods to save electricity costs; natural ventilation is carried out using regional climate conditions (such as low temperature or wind speed at night), and the wind speed and operating time of the ventilation system are dynamically adjusted according to indoor temperature and humidity requirements, load forecast results and changes in electricity prices.

[0072] In one example, for lighting type loads, based on the internal microgrid balancing optimization method for source-grid-load-storage virtual power plants provided in this application, the intelligent lighting and natural light utilization strategies are determined, and the lighting power load and energy consumption cost are reduced by dynamically adjusting the brightness and time period of lighting equipment; the lighting load is divided into core areas and auxiliary areas, and the lighting equipment in different areas are independently adjusted through an intelligent control system to give priority to the lighting needs of the core areas; combined with the outdoor light intensity, the indoor lighting brightness is dynamically adjusted through sensors to reduce the power consumption of lighting equipment during the day or in clear weather; comprehensive lighting of non-essential areas is carried out during low electricity price periods to meet centralized production or maintenance needs, and only necessary lighting loads are maintained during peak electricity price periods.

[0073] In some embodiments, the present application provides an internal microgrid balance optimization system 600 for a source-grid-load-storage virtual power plant, such as Figure 3 As shown, the internal microgrid balance optimization system 600 may include the following modules:

[0074] An acquisition module 610 is used to acquire first target power information representing historical load data and historical time series data from the power system according to a required data type;

[0075] The processing module 620 is used to clean, standardize, cluster and classify the first target power information to obtain the second target power information;

[0076] The first construction module 630 is used to construct a neural network learning framework according to the second target power information and the random process of fitting Brownian motion, and adjust the hyperparameters for optimization to obtain a load model;

[0077] A second construction module 640 is used to design a timing network according to the second target power information and construct a timing model to establish a time association;

[0078] The constraint determination module 650 is used to determine the adjustable variables and adjustable capacity ranges corresponding to the optimization target based on the load model and the time series model to obtain boundary constraint conditions;

[0079] The third construction module 660 is used to build a two-layer optimization architecture for user demand and user device response scheduling, and build a multi-task learning model based on boundary constraints;

[0080] The solution module 670 is used to solve the multi-task learning model and output the target solution to adjust the internal microgrid balance of the virtual power plant; wherein the target solution is used to characterize the resource allocation instructions required for each device under the optimization target.

[0081] According to an embodiment of the present application, any multiple modules among the acquisition module 610, the processing module 620, the first construction module 630, the second construction module 640, the constraint determination module 650, the third construction module 660 and the solution module 670 can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module.

[0082] In some embodiments, the processing module 620 may be specifically configured to:

[0083] Cleaning the first target power information by Lagrange interpolation method, eliminating outliers and noise in the data, and identifying and correcting errors and missing values ​​to determine the completeness and accuracy of the data, and obtaining first intermediate information;

[0084] Calculating a scaling factor based on the mapped target range and the original range to convert the data using a preset scaling formula so that the data have the same scale, thereby obtaining second intermediate information;

[0085] An unsupervised learning method combining contrastive learning and meta-learning is used to generate positive sample pairs of the same data through SimCLR enhancement, and different data are used as negative sample pairs to cluster the second intermediate information and obtain clustering results.

[0086] According to the clustering results, the load-side adjustable resources are graded and classified to obtain the second target power information.

[0087] In some embodiments, the second building module 640 may be specifically used to:

[0088] Performing time series decomposition, autocorrelation analysis, and feature extraction on the time series data in the second target power information to obtain feature information representing the load data;

[0089] The initial model is obtained by processing the time dependency of feature information through a preset long short-term memory network or gated recurrent unit;

[0090] Select a suitable loss function and adjust the network parameters, and train the initial model to obtain the time series model.

[0091] In some embodiments, Figure 4 As shown, the internal microgrid balance optimization system 600 may further include an incremental learning module 680, which may be specifically used for:

[0092] Nodes and edges are defined so that each node represents a load unit, the node position represents the position of the corresponding load unit in the virtual power grid, and the edge represents the relationship between nodes; wherein the weight of the edge is set based on the historical load correlation;

[0093] Use one of the graph convolutional network (GCN) and the graph attention network (GAT) as the basic architecture, introduce an incremental learning algorithm, aggregate the information of neighboring nodes through multi-layer graph convolution operations, introduce residual connections and layer normalization to improve the training effect and convergence speed of the model;

[0094] According to the changes in load data, the weights of the edges are dynamically adjusted, and the stream processing framework is used to process and update the graph data in real time so that the model can reflect the latest load changes.

[0095] In some embodiments, the third building block 660 may specifically include the following units:

[0096] The architecture building unit 661 is used to build a two-layer optimization architecture based on user demand and user device response scheduling sub-problems, and design load types for different optimization sub-problems;

[0097] A double-layer coding unit 662 is used to design a double-layer coding method for continuous-discrete mixed variables, where the continuous-discrete mixed variables include: load regulation and energy storage regulation continuous variables at the load side user end, and discrete variables for scheduling the next process of each user;

[0098] A target determination unit 663, configured to determine at least one of load-side adjustable capacity, user satisfaction, load-side energy loss, minimization of response cost, and energy storage cost as an optimization target;

[0099] The learning model building unit 664 is used to build a multi-task learning model based on the optimization goal, by limiting variables through boundary constraints and performing model constraints; the multi-task learning model includes an upper-level user demand sub-model and a lower-level load-side response scheduling sub-model.

[0100] In some embodiments, the learning model building unit 664 may be specifically used to:

[0101] Based on the two-layer optimization architecture, the upper-layer continuous decision variables are used as the input of the lower-layer optimization problem, reconstructed into discrete scheduling variables in the form of multi-layer encoding, and regarded as independent scenarios for simultaneous optimization;

[0102] In the decision space, based on the problem characteristics and population distribution, we judge whether there are similarities between tasks, and set a similarity measurement indicator as a benchmark for multi-task learning.

[0103] In some embodiments, Figure 4 As shown, the internal microgrid balance optimization system 600 may further include a feedback optimization module 690, and the feedback optimization module 690 may be specifically used for:

[0104] Based on the weighted loss function between tasks and the shared network structure, shared transfer learning jointly optimizes all tasks to train multi-task learning models;

[0105] Define the types of knowledge that can be shared through the main tasks and auxiliary tasks in the multi-task scenario, and decide whether to transfer local knowledge or overall knowledge to establish a knowledge transfer mechanism;

[0106] The model input results after knowledge transfer are used as the benchmark for environment update and solution selection. The top-K individuals are selected from the new input individuals and historical individuals as the initial path scheduling selection plan for the next generation. The results are evaluated and fed back to judge the effectiveness of the knowledge in order to realize the dynamic update mechanism of the transferred knowledge.

[0107] Figure 4 Each module in the system shown has the function of implementing each step in the aforementioned internal microgrid balance optimization method for the source-grid-load-storage virtual power plant, and can achieve its corresponding technical effects. For the sake of concise description, they will not be repeated here.

[0108] In some embodiments, the present application provides an electronic device, the structural diagram of the electronic device is as follows Figure 5 shown.

[0109] The electronic device may include a processor 710 and a memory 720 storing computer program instructions.

[0110] Specifically, the processor 710 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0111] The memory 720 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 720 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 720 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 720 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 720 is a non-volatile solid-state memory.

[0112] The memory 720 may include a read-only memory (ROM), a random access memory (RAM), a disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Therefore, generally, the memory 720 includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it can perform the operations described in any one of the internal microgrid balancing optimization methods for source-grid-load-storage virtual power plants in the above-mentioned embodiments.

[0113] The processor 710 reads and executes computer program instructions stored in the memory 720 to implement any one of the internal microgrid balancing optimization methods for the source-grid-load-storage virtual power plant in the above embodiments.

[0114] In one example, the electronic device may further include a communication interface 730 and a bus 700. Figure 5 As shown, the processor 710, the memory 720, and the communication interface 730 are connected via a bus 700 and communicate with each other.

[0115] The communication interface 730 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0116] Bus 700 includes hardware, software or both, and the parts of online data flow billing equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front-end bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 700 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the present application considers any suitable bus or interconnection.

[0117] In addition, in combination with the internal microgrid balancing optimization method for source-grid-load-storage virtual power plants in the above embodiments, the present application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when the computer program instructions are executed by the processor, any one of the internal microgrid balancing optimization methods for source-grid-load-storage virtual power plants in the above embodiments is implemented.

[0118] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0119] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0120] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0121] Aspects of the present disclosure are described above with reference to the flowchart and / or block diagram of the method, device (system) and computer program product according to the embodiment of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the function / action specified in one or more boxes of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It can also be understood that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can also be implemented by dedicated hardware that performs a specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0122] In summary, compared with the prior art, this application has the following beneficial effects:

[0123] 1. This application constructs a two-layer optimization architecture and establishes a multi-task learning model. Based on the determined adjustable variables and adjustable capacity range, the boundary constraints of the multi-task learning model can be determined; the target solution obtained by solving the multi-task learning model can characterize the resource allocation instructions required for each device under the optimization goal, and then determine the response strategy to flexibly adjust the user load, interact on both the supply and demand sides, and realize multi-objective optimization of source, grid, load and storage resources, which is beneficial to the internal microgrid balance.

[0124] 2. This application predicts future load demand and power generation capacity by analyzing historical operating data and real-time monitoring data, and adopts machine learning and deep learning algorithms to improve load forecasting accuracy and optimize scheduling efficiency. The scheduling strategy can be adjusted according to real-time data to cope with load changes and fluctuations in renewable energy; and multi-objective optimization of source, grid, load and storage resources can be achieved, including maximizing economic benefits, minimizing carbon emissions and balancing supply and demand.

[0125] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An internal microgrid balance optimization method for source-grid-load-storage virtual power plant, characterized in that: include: According to the required data type, first target power information representing historical load data and historical time series data is obtained from the power system; Cleaning, standardizing, clustering and grading the first target power information to obtain second target power information; According to the second target power information, combined with the random process of fitting Brownian motion, a neural network learning framework is constructed, and hyperparameters are adjusted for optimization to obtain a load model; Designing a timing network according to the second target power information and constructing a timing model to establish a time association; Based on the load model and the timing model, determining the adjustable variables and adjustable capacity range corresponding to the optimization target, and obtaining boundary constraint conditions; Build a two-layer optimization architecture for user demand and user device response scheduling, and build a multi-task learning model based on the boundary constraints; The multi-task learning model is solved and a target solution is output to adjust the internal microgrid balance of the virtual power plant; wherein the target solution is used to characterize the resource allocation instructions required for each device under the optimization target.

2. The internal microgrid balance optimization method for source-grid-load-storage virtual power plant according to claim 1, characterized in that: The cleaning, standardization, clustering and hierarchical classification of the first target power information to obtain the second target power information includes: Cleaning the first target power information by Lagrange interpolation method, eliminating outliers and noise in the data, and identifying and correcting errors and missing values ​​to determine the completeness and accuracy of the data, and obtaining first intermediate information; Calculating a scaling factor based on the mapped target range and the original range to convert the data using a preset scaling formula so that the data have the same scale, thereby obtaining second intermediate information; An unsupervised learning method combining contrastive learning and meta-learning is adopted to generate positive sample pairs of the same data through SimCLR enhancement, and different data are used as negative sample pairs to cluster the second intermediate information to obtain a clustering result; According to the clustering result, the load-side adjustable resources are classified and graded to obtain the second target power information.

3. The internal microgrid balance optimization method for source-grid-load-storage virtual power plant according to claim 1, characterized in that: The step of designing a timing network according to the second target power information and constructing a timing model to establish a time association includes: Performing time series decomposition, autocorrelation analysis, and feature extraction on the time series data in the second target power information to obtain feature information characterizing the load data; The time dependency of the feature information is processed by a preset long short-term memory network or a gated recurrent unit to perform modeling to obtain an initial model; A suitable loss function is selected and network parameters are adjusted, and the initial model is trained to obtain a time series model.

4. The internal microgrid balance optimization method for source-grid-load-storage virtual power plant according to claim 1, characterized in that: After designing a timing network according to the second target power information and constructing a timing model to establish a time association, the method further includes: Nodes and edges are defined so that each node represents a load unit, and the node position represents the position of the corresponding load unit in the virtual power grid, and the edge represents the relationship between nodes; wherein the weight of the edge is set based on the historical load correlation; Use one of the graph convolutional network (GCN) and the graph attention network (GAT) as the basic architecture, introduce an incremental learning algorithm, aggregate the information of neighboring nodes through multi-layer graph convolution operations, introduce residual connections and layer normalization to improve the training effect and convergence speed of the model; According to the changes in load data, the weights of the edges are dynamically adjusted, and the stream processing framework is used to process and update the graph data in real time so that the model can reflect the latest load changes.

5. The internal microgrid balance optimization method for source-grid-load-storage virtual power plant according to any one of claims 1 to 4, characterized in that: The two-layer optimization architecture for user demand and user device response scheduling is constructed, and a multi-task learning model is constructed based on the boundary constraints, including: A two-layer optimization architecture is built based on user demand and user device response scheduling sub-problems, and load types are designed for different optimization sub-problems; Design a double-layer encoding method for continuous-discrete mixed variables, wherein the continuous-discrete mixed variables include: load regulation and energy storage regulation continuous variables at the load side user end, and discrete variables for scheduling the next process of each user; Determining at least one of load-side adjustable capacity, user satisfaction, load-side energy loss, minimizing response cost, and energy storage cost as the optimization target; Based on the optimization objective, variables are restricted by the boundary constraints and model constraints are performed to construct a multi-task learning model; the multi-task learning model includes an upper-layer user demand sub-model and a lower-layer load-side response scheduling sub-model.

6. The internal microgrid balance optimization method for source-grid-load-storage virtual power plant according to claim 5, characterized in that: Based on the optimization objective, the variables are restricted by the boundary constraints and the model constraints are performed to construct a multi-task learning model, including: Based on the two-layer optimization architecture, the upper-layer continuous decision variables are used as inputs of the lower-layer optimization problem, reconstructed into discrete scheduling variables in the form of multi-layer encoding, and are considered as independent scenarios and optimized simultaneously; In the decision space, based on the problem characteristics and population distribution, we judge whether there are similarities between tasks, and set a similarity measurement indicator as a benchmark for multi-task learning.

7. The internal microgrid balance optimization method for source-grid-load-storage virtual power plant according to claim 5, characterized in that: After constructing the multi-task learning model based on the optimization objective, limiting the variables by the boundary constraint conditions and performing model constraints, the method further includes: Based on the weighted loss function between tasks and the shared network structure, shared transfer learning jointly optimizes all tasks to train multi-task learning models; Define the types of knowledge that can be shared through the main tasks and auxiliary tasks in the multi-task scenario, and decide whether to transfer local knowledge or overall knowledge to establish a knowledge transfer mechanism; The model input results after knowledge transfer are used as the benchmark for environment update and solution selection. The top-K individuals are selected from the new input individuals and historical individuals as the initial path scheduling selection plan for the next generation. The results are evaluated and fed back to judge the effectiveness of the knowledge in order to realize the dynamic update mechanism of the transferred knowledge.

8. An internal microgrid balance optimization system for source-grid-load-storage virtual power plants, characterized in that: include: An acquisition module, used for acquiring first target power information representing historical load data and historical time series data from the power system according to the required data type; a processing module, configured to clean, standardize, cluster and classify the first target power information to obtain second target power information; A first building module is used to build a neural network learning framework according to the second target power information and a random process of fitting Brownian motion, and adjust hyperparameters for optimization to obtain a load model; A second construction module is used to design a timing network according to the second target power information and construct a timing model to establish a time association; A constraint determination module, used to determine the adjustable variables and adjustable capacity range corresponding to the optimization target based on the load model and the timing model, and obtain boundary constraint conditions; The third building module is used to build a two-layer optimization architecture for user demand and user device response scheduling, and build a multi-task learning model based on the boundary constraints; A solution module is used to solve the multi-task learning model and output a target solution to adjust the internal microgrid balance of the virtual power plant; wherein the target solution is used to characterize the resource allocation instructions required for each device under the optimization target.

9. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the internal microgrid balancing optimization method for a source-grid-load-storage virtual power plant as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores programs or instructions, and when the programs or instructions are executed by the processor, the internal microgrid balance optimization method for the source-grid-load-storage virtual power plant as described in any one of claims 1 to 7 is implemented.

Images