Multi-microgrid intelligent energy scheduling method and system based on deep learning

Abstract

The invention relates to the field of power grid energy scheduling, in particular to a multi-microgrid intelligent energy scheduling method and system based on deep learning. The method comprises the following steps: acquiring a plurality of micro-grid multi-dimensional state parameters; performing multi-time-window change analysis on the multi-dimensional state parameters of the plurality of microgrids, and performing deep convolution learning to generate state convolution diagrams of the plurality of microgrids; identifying a plurality of microgrid nodes; performing node relative position analysis on the plurality of micro-grid nodes one by one, and performing node topological association mining to generate micro-grid node topological association representation; performing dynamic state attribute mapping based on the micro-grid node topology association representation and the state convolution graphs of the plurality of micro-grids, and constructing a micro-grid topology twinborn model; and performing power grid state trend evolution on the state convolution graphs of the plurality of micro-grids one by one, and performing energy demand prediction to generate a plurality of micro-grid energy demand prediction values. According to the invention, high-efficiency and high-utilization-rate energy scheduling is realized.

Description

Technical Field

[0001] The present invention relates to the field of power grid energy dispatching, and in particular to a multi-microgrid intelligent energy dispatching method and system based on deep learning. Background Art

[0002] With the rapid development of distributed energy systems, the interconnection between multiple microgrid systems is becoming an important direction for the construction of future smart grids. Microgrid systems are composed of a variety of distributed power generation equipment, energy storage devices and user loads, and there are complex topological structures and energy flow relationships within them. In the context of the coordinated operation of multiple microgrids, how to achieve intelligent energy scheduling for each microgrid system is a technical problem that needs to be solved urgently.

[0003] Traditional energy scheduling methods mainly rely on manual experience and local optimization algorithms, which are difficult to fully reflect the complex dynamic characteristics of multi-microgrid systems. There are often problems of low energy scheduling efficiency and low utilization. With the continuous advancement of artificial intelligence technology, modeling and optimization methods based on deep learning are becoming an emerging trend in solving complex energy system problems.

[0004] Deep learning can automatically extract the complex correlation between system status and energy supply and demand through massive historical data, providing strong support for intelligent scheduling decisions. At the same time, based on the modeling method of deep learning, a virtual model that fully reflects the topological structure and dynamic state of the microgrid is constructed, providing a simulation evaluation platform for dynamic optimization scheduling. Therefore, designing a multi-microgrid intelligent energy scheduling method based on deep learning has important practical significance. Summary of the invention

[0005] In order to solve the above-mentioned technical problems, the present invention proposes a multi-microgrid intelligent energy scheduling method and system based on deep learning to solve at least one of the above-mentioned technical problems.

[0006] To achieve the above object, the present invention provides a multi-microgrid intelligent energy scheduling method based on deep learning, comprising the following steps:

[0007] Step S1: obtaining multi-dimensional state parameters of multiple microgrids; performing multi-time window change analysis on the multi-dimensional state parameters of the multiple microgrids, and performing deep convolution learning to generate state convolution graphs of the multiple microgrids;

[0008] Step S2: identifying multiple microgrid nodes; performing relative position analysis on multiple microgrid nodes one by one, and performing node topology association mining to generate a microgrid node topology association representation;

[0009] Step S3: Based on the microgrid node topology association representation and the state convolution graphs of multiple microgrids, dynamic state attribute mapping is performed to construct a microgrid topology twin model;

[0010] Step S4: performing grid state trend evolution on the state convolution graphs of multiple microgrids one by one, and performing energy demand forecasting to generate multiple microgrid energy demand forecast values;

[0011] Step S5: Based on multiple microgrid energy demand forecast values, dynamic energy scheduling calculation is performed on the microgrid topology twin model, and distributed collaborative decision-making is performed to construct a microgrid dynamic scheduling strategy;

[0012] Step S6: Iteratively fine-tune the topology structure of the microgrid topology twin model according to the microgrid dynamic dispatch strategy, and construct a microgrid topology optimization structure model to perform intelligent energy dispatch operations.

[0013] The present invention uses multi-time window change analysis and deep convolution learning to better understand the state change law of each node in the microgrid, improve the grasp of the microgrid state, generate a state convolution graph of the microgrid, convert multi-dimensional state information into a processable data form, and provide useful input for subsequent steps. The relative position analysis of each node and the node topology association mining reveal the association and topological structure between the nodes in the microgrid, providing a basis for subsequent modeling, generating a microgrid node topology association representation to understand the relationship between the internal nodes of the microgrid, and providing support for the establishment of a microgrid topology twin model. The microgrid topology twin model is constructed using the node topology association representation and the state convolution graph, which integrates the structure and state information of the microgrid and provides comprehensive information for subsequent decision-making. The establishment of the microgrid topology twin model comprehensively understands the state and structure of the microgrid and provides support for intelligent scheduling. Based on the supply foundation, the trend evolution of grid status and energy demand forecast are analyzed one by one to predict the future energy demand of the microgrid, improve the accuracy and efficiency of energy scheduling, generate energy demand forecast values ​​to help plan future energy supply, optimize the energy utilization of the microgrid, and perform dynamic energy scheduling calculations and distributed collaborative decision-making based on energy demand forecast values ​​to achieve effective utilization and allocation of energy resources within the microgrid. Building a dynamic scheduling strategy for the microgrid can improve the energy utilization efficiency of the microgrid, reduce energy costs, and achieve more intelligent energy management. Fine-tune the microgrid topology according to the dynamic scheduling strategy, build a microgrid topology optimization structure model, optimize the energy flow path and scheduling logic within the microgrid, and perform intelligent energy scheduling operations to improve the operating efficiency and stability of the microgrid, reduce energy waste, and protect the energy security of enterprise data assets to the greatest extent.

[0014] Preferably, step S1 comprises the following steps:

[0015] Step S11: Acquire multi-dimensional state parameters of multiple microgrids;

[0016] Step S12: extracting the load and power generation of each microgrid based on the multi-dimensional state parameters of the multiple microgrids;

[0017] Step S13: performing electrical parameter characteristic analysis on the load and power generation of each microgrid to obtain electrical parameter characteristics of multiple microgrids;

[0018] Step S14: performing multi-time window change analysis on the electrical parameter characteristics of the multiple microgrids to generate time sequence state change curves of the multiple microgrids;

[0019] Step S15: performing deep convolution learning on the time-series state change curves of the multiple microgrids to generate state convolution graphs of the multiple microgrids.

[0020] The present invention obtains multi-dimensional state parameters to fully understand the current state of the microgrid, including important parameters such as power load, power generation, battery status, etc. The multi-dimensional state parameters provide key information for the operation of the microgrid, provide basic data for subsequent analysis and scheduling, extract load and power generation to quantify the energy consumption and output of the microgrid, and provide key data for subsequent analysis. Load and power generation are important indicators of microgrid operation, which help evaluate the operating efficiency and stability of the microgrid. Electrical parameter feature analysis can deeply understand the power characteristics of the microgrid, including the changing laws of parameters such as power, current, and voltage, and obtain electrical parameter characteristics. Reveal the power load and power generation of the microgrid, and provide more detailed information for subsequent analysis. Multi-time window change analysis observes the changing trends of the electrical parameter characteristics of the microgrid in different time windows, helps to discover periodic and trend changes, and generates time series state change curves to fully understand the dynamic changes of the microgrid power characteristics, providing time series information for intelligent energy scheduling. Deep convolution learning learns the state change patterns and laws of the microgrid from the time series state change curves, improves the accuracy of energy scheduling, and generates state convolution graphs to have a deeper understanding and analysis of the microgrid state, providing more effective models and methods for realizing intelligent energy scheduling.

[0021] Preferably, the specific steps of step S14 are:

[0022] Define the time window length;

[0023] Based on the time window length, multiple sliding window slice analyses are performed on the electrical parameter characteristics of multiple microgrids to generate a timing parameter characteristic of each microgrid;

[0024] The parameter variation of adjacent time windows is analyzed for the timing parameter characteristics of each microgrid to obtain the multi-segment timing parameter variation characteristics;

[0025] Extract timestamps of the characteristics of changes in multiple time series parameters;

[0026] Performing sequential serialization processing on multiple sections of timing parameter change characteristics according to the timestamp to generate a microgrid timing parameter change characteristic sequence;

[0027] The state change curve is fitted for the characteristic sequence of microgrid timing parameter changes to generate multiple microgrid timing state change curves.

[0028] The present invention controls the time span of analysis by defining the time window length, making the analysis more accurate and effective. The selection of the time window length affects the timing and granularity of subsequent analysis. The appropriate time window length captures the state change of the microgrid. The sliding window slice analysis is used to divide the time series data into multiple windows, so as to capture the state characteristics of the microgrid in different time periods. The time series parameter characteristics of each microgrid are generated to describe the state change of the microgrid in more detail, providing richer information for subsequent analysis. The adjacent time window parameter change analysis finds the change law of the microgrid parameters in different time periods, helps to identify sudden changes or trend changes, obtains multiple sections of time series parameter change characteristics, and understands the evolution process of the microgrid state in more detail, providing a basis for the prediction and processing of state changes. The timestamp is extracted to record the occurrence time of each time series parameter change feature, helping to establish time series correlation. The timestamp provides the time information of the time series data, associates the time clues of the change of the microgrid state, fits the time series parameter change feature sequence, integrates the scattered features into a continuous state change curve, better presents the overall change trend of the microgrid state, generates a time series state change curve, intuitively displays the change of the microgrid state, and provides a visual reference for intelligent energy scheduling.

[0029] Preferably, the specific steps of step S2 are:

[0030] Step S21: identifying multiple microgrid nodes;

[0031] Step S22: performing spatial position calculation on a plurality of microgrid nodes to generate spatial position coordinates of the microgrid nodes;

[0032] Step S23: performing relative position analysis on multiple microgrid nodes one by one to generate relative position relationships between the nodes;

[0033] Step S24: performing adjacent interactive collaborative analysis on the relative position relationship between the nodes to generate adjacent node interactive collaborative data;

[0034] Step S25: performing node topology association mining on the interactive collaborative data between adjacent nodes based on the spatial position coordinates of the microgrid nodes to generate a microgrid node topology association representation.

[0035] The present invention clarifies the identity and characteristics of each important component in the microgrid by identifying the microgrid nodes. Node identification provides basic data for subsequent spatial position calculation and relationship analysis. Spatial position calculation determines the specific position of the microgrid node in space, understands the layout and structure of the microgrid, and generates the spatial position coordinates of the microgrid node, which provides basic information for subsequent node relationship analysis and topological association mining. Relative position analysis reveals the spatial relationship between microgrid nodes, including information such as the distance and direction between nodes. Understanding the relative position relationship between nodes provides a basis for subsequent node interactive collaborative analysis. Interactive collaborative analysis reveals the interaction mode and collaborative relationship between nodes, understands the connection and influence between microgrid nodes, and node topological association mining reveals the topological structure and association relationship between microgrid nodes, helps understand the connection mode and transmission path within the microgrid, generates microgrid node topological association representation, establishes an association model between microgrid nodes, and provides more comprehensive data support for intelligent energy scheduling.

[0036] Preferably, step S3 specifically comprises the following steps:

[0037] Step S31: performing spatial distribution analysis on multiple microgrid nodes to obtain spatial distribution data of microgrid nodes;

[0038] Step S32: performing node space topology structure analysis on the microgrid node space distribution data based on the microgrid node topology association representation to obtain node space topology structure data;

[0039] Step S33: Perform global topological connection based on the node space topological structure data to construct a microgrid topological node space;

[0040] Step S34: Based on the state convolution graphs of multiple microgrids, dynamic state attribute mapping is performed on the microgrid topology node space to construct a microgrid topology twin model.

[0041] The present invention understands the distribution law and characteristics of microgrid nodes in space through spatial distribution analysis, provides basic data for subsequent topological structure analysis, and obtains spatial distribution data of microgrid nodes to provide spatial background information for the association and connection between nodes. Node spatial topological structure analysis can reveal the topological relationship and structural characteristics between microgrid nodes, including the connection mode and path information between nodes. The node spatial topological structure data obtained provides a basis for subsequent global topological connection and helps to build the overall network structure of the microgrid. The global topological connection integrates the topological structure relationship between microgrid nodes into an overall network, reveals the global connection mode of nodes inside the microgrid, constructs a microgrid topological node space to form an overall topological view of the microgrid, and provides a basis for subsequent modeling and analysis. Dynamic state attribute mapping combines microgrid state information and topological structure to reveal the propagation and influence path of states between microgrid nodes, and constructs a microgrid topological twin model to integrate the structure and state information of the microgrid, providing a more comprehensive reference and decision-making basis for intelligent energy scheduling.

[0042] Preferably, the specific steps of step S4 are:

[0043] Step S41: performing grid state trend evolution on the state convolution graphs of multiple microgrids one by one to generate state trend evolution data of each microgrid;

[0044] Step S42: performing energy demand situation analysis on the state trend evolution data of each microgrid to obtain microgrid energy demand situation data;

[0045] Step S43: performing energy demand forecasting for continuous time periods on the microgrid energy demand situation data to obtain energy demand forecast values ​​for multiple time periods;

[0046] Step S44: performing iterative error optimization calculation on energy demand forecast values ​​for multiple time periods to generate multiple microgrid energy demand forecast values.

[0047] The present invention can reveal the state change law and trend of each microgrid through the evolution of power grid state trend, provide basic data for subsequent energy demand analysis, generate state trend evolution data of each microgrid to understand the state change situation inside the microgrid, and provide a basis for energy demand situation analysis. Energy demand situation analysis helps to understand the energy demand situation and change trend of the microgrid, and provides a basis for subsequent energy demand forecasting. The microgrid energy demand situation data is obtained to grasp the demand characteristics of the microgrid, provide support for energy demand forecasting and planning, and perform energy demand forecasting for continuous time periods to estimate and plan the future energy demand situation of the microgrid, reasonably arrange energy supply, and obtain energy demand forecast values ​​for multiple time periods to provide important forecast data basis for energy scheduling and optimization of the microgrid. Iterative error optimization calculation improves the accuracy and stability of energy demand forecasting, and continuously optimizes the forecasting model to improve the accuracy of the forecasting results. Generate multiple microgrid energy demand forecast values ​​to provide a more reliable and accurate data basis for the energy supply and scheduling of the microgrid.

[0048] Preferably, the specific steps of step S5 are:

[0049] Step S51: simulating the evolution of the operating state of the microgrid topology twin model to generate microgrid operation simulation data;

[0050] Step S52: Calculate the energy supply of each node on the microgrid operation simulation data to obtain energy supply parameters of each node;

[0051] Step S53: Based on multiple microgrid energy demand forecast values, energy supply parameters of each node are identified as energy demand shortage, and demand deviation microgrid nodes are marked;

[0052] Step S54: performing dynamic energy scheduling calculation on the demand deviation microgrid node to generate dynamic energy scheduling data;

[0053] Step S55: Perform distributed collaborative decision-making on dynamic energy scheduling data to build a microgrid dynamic scheduling strategy.

[0054] The present invention simulates the operation of the microgrid under different conditions through the operation state simulation evolution, helps to understand the response and performance of the microgrid system, generates microgrid operation simulation data to evaluate the stability and reliability of the microgrid, and provides basic data for subsequent energy supply calculations. The node energy supply calculation determines the energy supply situation of each node, including energy type, supply amount and other information, provides data support for subsequent energy demand analysis, obtains the energy supply parameters of each node to understand the energy supply situation of the nodes inside the microgrid, and provides a basis for energy demand shortage identification. Energy demand shortage identification compares the energy supply and demand, marks the microgrid nodes with demand deviation, and helps to discover potential In the energy demand problem, the demand deviation microgrid node is marked to focus on and solve the problem of insufficient energy supply or excess demand in the microgrid. The dynamic energy scheduling calculation flexibly adjusts the energy distribution according to the situation of the demand deviation node, optimizes the energy utilization and supply efficiency of the microgrid, and generates dynamic energy scheduling data to adjust the energy configuration of the microgrid in real time to ensure that the energy supply of each node meets the demand. Distributed collaborative decision-making can build a dynamic scheduling strategy that adapts to the actual situation of the microgrid through multi-party collaboration, improve the overall energy scheduling efficiency of the microgrid, and build a microgrid dynamic scheduling strategy to realize the collaborative operation between the nodes of the microgrid, thereby improving the overall energy utilization efficiency and system stability.

[0055] Preferably, the specific steps of step S54 are:

[0056] According to the microgrid topology twin model, the deviation node of the demand deviation microgrid node is located to obtain the demand deviation node position data;

[0057] Based on the demand deviation node location data, the neighboring nodes in the region are analyzed to obtain the microgrid nodes in the region;

[0058] According to the energy supply parameters of each node, the dispatchable microgrid nodes in the region are identified as nodes with sufficient energy, and the dispatchable neighboring microgrid nodes are obtained;

[0059] Calculate the energy compensation demand of the demand deviation microgrid node and obtain the compensation demand value of the deviation node;

[0060] The dynamic energy scheduling calculation of the deviation node compensation demand value is performed by calling the neighboring microgrid nodes to generate dynamic energy scheduling data.

[0061] The present invention accurately locates nodes with abnormal demand in the microgrid by locating the demand deviation node position data, provides target nodes for further energy scheduling, obtains the demand deviation node position data and focuses on solving the energy demand problem existing in the microgrid, improves the overall energy supply and demand balance, conducts neighboring node analysis in the region to determine other nodes adjacent to the demand deviation node, establishes the connection and influence relationship between the nodes in the microgrid, obtains the microgrid nodes in the region to determine the node range that affects the energy supply, provides a basis for subsequent energy scheduling, identifies nodes with sufficient energy to determine nodes that can be used to compensate for the demand deviation nodes, optimizes the energy scheduling plan, improves the overall energy utilization rate, and obtains adjustable Use neighboring microgrid nodes to determine the dispatchable nodes, provide available resources for dynamic energy scheduling, quantify the energy compensation required by the demand deviation nodes through compensation demand calculation, provide specific demand values ​​for dynamic energy scheduling, obtain the compensation demand value of the deviation node, clarify the compensation demand of the demand deviation node, and provide data support for subsequent energy scheduling. Dynamic energy scheduling calculation adjusts energy distribution in real time according to the compensation demand value of the demand deviation node and the energy supply of neighboring nodes, optimizes the energy utilization and supply efficiency of the microgrid, generates dynamic energy scheduling data to realize coordinated scheduling between nodes within the microgrid, ensure the overall energy supply and demand balance, and improve the reliability and efficiency of the microgrid.

[0062] Preferably, the specific steps of step S6 are:

[0063] Step S61: Calculate the energy supply and demand efficiency of the microgrid topology twin model to generate the supply and demand efficiency of the current topology model;

[0064] Step S62: dynamically optimizing the microgrid dynamic dispatch strategy based on the supply and demand efficiency of the current topology model to generate a dynamic dispatch efficiency optimization strategy;

[0065] Step S63: Iteratively fine-tune the topology structure of the microgrid topology twin model according to the dynamic scheduling efficiency optimization strategy, and construct a microgrid topology optimization structure model to perform intelligent energy scheduling operations.

[0066] The present invention evaluates the energy utilization efficiency and supply and demand balance of the current microgrid system by calculating the energy supply and demand efficiency of the microgrid topology twin model, generates the supply and demand efficiency data of the current topology model to help identify potential energy supply and demand contradictions, and provides benchmark data for subsequent optimization. The dynamic scheduling strategy is optimized based on the supply and demand efficiency of the current topology model to adjust the energy scheduling plan of the microgrid according to actual conditions, improve the overall energy utilization efficiency, generate a dynamic scheduling efficiency optimization strategy to adjust the energy allocation strategy of the microgrid according to real-time demand and supply conditions, improve the response speed and efficiency of the system, iterate and fine-tune the topology structure of the microgrid topology twin model according to the dynamic scheduling efficiency optimization strategy, adjust the structure of the microgrid according to actual conditions, optimize the energy transmission efficiency, construct a microgrid topology optimization structure model to optimize the connection relationship between the internal nodes of the microgrid, and improve the energy transmission efficiency and system stability.

[0067] In this specification, a multi-microgrid intelligent energy dispatching system based on deep learning is provided, which is used to execute the multi-microgrid intelligent energy dispatching method based on deep learning as described above, including:

[0068] A multi-dimensional state change module is used to obtain multi-dimensional state parameters of multiple microgrids; perform multi-time window change analysis on the multi-dimensional state parameters of multiple microgrids, and perform deep convolution learning to generate state convolution graphs of multiple microgrids;

[0069] A topological association mining module is used to identify multiple microgrid nodes; analyze the relative positions of multiple microgrid nodes one by one, and perform node topological association mining to generate a microgrid node topological association representation;

[0070] The twin model module is used to map dynamic state attributes based on the topological association representation of microgrid nodes and the state convolution graphs of multiple microgrids to construct a microgrid topology twin model;

[0071] The energy demand prediction module is used to perform the state trend evolution of multiple microgrids one by one on the state convolution graphs, and perform energy demand prediction to generate multiple microgrid energy demand prediction values;

[0072] The dynamic energy scheduling module is used to perform dynamic energy scheduling calculations on the microgrid topology twin model based on multiple microgrid energy demand forecast values, and to make distributed collaborative decisions to build a microgrid dynamic scheduling strategy;

[0073] The topology optimization module is used to iteratively fine-tune the topology structure of the microgrid topology twin model according to the microgrid dynamic scheduling strategy, and to build a microgrid topology optimization structure model to perform intelligent energy scheduling operations.

[0074] The present invention obtains multi-dimensional state parameters of multiple microgrids to comprehensively understand the operation status of the microgrid system, provides a data basis for subsequent analysis and optimization, performs multi-time window change analysis and deep convolution learning to capture the time series characteristics of the microgrid state, generates a state convolution graph to reveal the complex relationship between the microgrid states, identifies multiple microgrid nodes and performs node topology association mining to establish the relationship between the nodes in the microgrid, provides key information for subsequent model construction, generates microgrid node topology association representation to help understand the correlation between the internal nodes of the microgrid, provides input features for the construction of the twin model, constructs a microgrid topology twin model based on the microgrid node topology association representation and the state convolution graph to perform dynamic state attribute mapping, realizes the fusion modeling of the microgrid state and structure, constructs the twin model to comprehensively consider the structural characteristics and state information of the microgrid, and provides more information for subsequent energy scheduling. Accurate model support, energy demand forecasting for the state convolution graphs of multiple microgrids to predict future energy demand trends, provide forecasting basis for dynamic energy scheduling, generate energy demand forecast values ​​to help the microgrid system make adjustments in advance, optimize energy allocation strategies, and improve energy utilization efficiency. Dynamic energy scheduling calculations are performed based on energy demand forecast values. Energy allocation is adjusted according to real-time energy demand conditions to improve the flexibility and response speed of energy scheduling. A dynamic scheduling strategy for distributed collaborative decision-making is constructed to coordinate scheduling between internal nodes of the microgrid to achieve overall energy supply and demand balance. According to the dynamic scheduling strategy, the topology structure of the microgrid topology twin model is iteratively fine-tuned to optimize the structure of the microgrid, improve energy transmission efficiency and system stability, and construct a topology optimization structure model to fine-tune the connection relationship between internal nodes of the microgrid to provide more efficient support for intelligent energy scheduling operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 A schematic diagram of the steps of a multi-microgrid intelligent energy scheduling method based on deep learning in the present invention;

[0076] Figure 2 Detailed implementation flow chart of step S1;

[0077] Figure 3 Detailed implementation flow chart of step S2;

[0078] Figure 4 Detailed implementation flow chart of step S3. DETAILED DESCRIPTION

[0079] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0080] The present application example provides a multi-microgrid intelligent energy dispatching method and system based on deep learning. The execution subject of the multi-microgrid intelligent energy dispatching method and system based on deep learning includes but is not limited to: mechanical equipment, data processing platform, cloud server node, network upload device, etc. equipped with the system can be regarded as the general computing node of the present application, and the data processing platform includes but is not limited to: at least one of an audio image management system, an information management system, and a cloud data management system.

[0081] See also Figures 1 to 4 The present invention provides a multi-microgrid intelligent energy scheduling method based on deep learning, comprising the following steps:

[0082] Step S1: obtaining multi-dimensional state parameters of multiple microgrids; performing multi-time window change analysis on the multi-dimensional state parameters of the multiple microgrids, and performing deep convolution learning to generate state convolution graphs of the multiple microgrids;

[0083] Step S2: identifying multiple microgrid nodes; performing relative position analysis on multiple microgrid nodes one by one, and performing node topology association mining to generate a microgrid node topology association representation;

[0084] Step S3: Based on the microgrid node topology association representation and the state convolution graphs of multiple microgrids, dynamic state attribute mapping is performed to construct a microgrid topology twin model;

[0085] Step S4: performing grid state trend evolution on the state convolution graphs of multiple microgrids one by one, and performing energy demand forecasting to generate multiple microgrid energy demand forecast values;

[0086] Step S5: Based on multiple microgrid energy demand forecast values, dynamic energy scheduling calculation is performed on the microgrid topology twin model, and distributed collaborative decision-making is performed to construct a microgrid dynamic scheduling strategy;

[0087] Step S6: Iteratively fine-tune the topology structure of the microgrid topology twin model according to the microgrid dynamic dispatch strategy, and construct a microgrid topology optimization structure model to perform intelligent energy dispatch operations.

[0088] The present invention uses multi-time window change analysis and deep convolution learning to better understand the state change law of each node in the microgrid, improve the grasp of the microgrid state, generate a state convolution graph of the microgrid, convert multi-dimensional state information into a processable data form, and provide useful input for subsequent steps. The relative position analysis of each node and the node topology association mining reveal the association and topological structure between the nodes in the microgrid, providing a basis for subsequent modeling, generating a microgrid node topology association representation to understand the relationship between the internal nodes of the microgrid, and providing support for the establishment of a microgrid topological twin model. The microgrid topological twin model is constructed using the node topology association representation and the state convolution graph, which integrates the structure and state information of the microgrid and provides comprehensive information for subsequent decision-making. The establishment of the microgrid topological twin model fully understands the microgrid. The state and structure of the microgrid are used to provide a basis for intelligent scheduling. The state trend evolution and energy demand forecast of each grid are used to predict the future energy demand of the microgrid, improve the accuracy and efficiency of energy scheduling, generate energy demand forecast values ​​to help plan future energy supply, optimize the energy utilization of the microgrid, and perform dynamic energy scheduling calculations and distributed collaborative decision-making based on the energy demand forecast values ​​to achieve effective utilization and allocation of energy resources within the microgrid. Building a dynamic scheduling strategy for the microgrid can improve the energy utilization efficiency of the microgrid, reduce energy costs, and achieve more intelligent energy management. According to the dynamic scheduling strategy, the microgrid topology structure is fine-tuned, a microgrid topology optimization structure model is constructed, the energy flow path and scheduling logic within the microgrid are optimized, and intelligent energy scheduling operations are performed to improve the operation efficiency and stability of the microgrid.

[0089] In the embodiment of the present invention, refer to Figure 1 , is a schematic diagram of the steps of a multi-microgrid intelligent energy scheduling method based on deep learning of the present invention. In this example, the steps of the multi-microgrid intelligent energy scheduling method based on deep learning include:

[0090] Step S1: obtaining multi-dimensional state parameters of multiple microgrids; performing multi-time window change analysis on the multi-dimensional state parameters of the multiple microgrids, and performing deep convolution learning to generate state convolution graphs of the multiple microgrids;

[0091] In this embodiment, operation data is collected from various systems of the microgrid (such as power generation, energy storage, load, etc.), including multidimensional parameters such as power, voltage, current, temperature, humidity, etc. According to the topological structure of the microgrid, the multidimensional state parameters of different nodes are integrated into a multidimensional state data set, and necessary preprocessing is performed on the data, such as outlier removal, data normalization, etc. The collected multidimensional state data is organized according to the time series to form a time series data matrix, and a sliding window analysis is performed on the time series data matrix to extract the change trend of the multidimensional parameters in different time windows, and the parameter change trend in different time windows is integrated into a multi-time window change analysis result, and the multi-time window change analysis result is input into a deep convolutional neural network model. The convolutional neural network extracts spatiotemporal features and learns the intrinsic correlation between the multidimensional state parameters. The trained deep convolutional network model can be used to generate a state convolution graph of each microgrid node, and the state convolution graph intuitively reflects the coupling relationship between the multidimensional states of each node in the microgrid.

[0092] Step S2: identifying multiple microgrid nodes; performing relative position analysis on multiple microgrid nodes one by one, and performing node topology association mining to generate a microgrid node topology association representation;

[0093] In this embodiment, each node in the microgrid is determined according to the physical topology information of the microgrid, where the node is a generator, an energy storage device, a load device, etc. A unique node ID is assigned to each node to identify and manage different nodes, and basic information of each node, such as node type, capacity, location coordinates, etc., is collected. The relative position relationship between any two nodes is calculated using the geographic location coordinate information of the node. The relative position relationship is represented by metrics such as Euclidean distance and Manhattan distance. The relative position relationships between all nodes are organized into a node relative position matrix, the node relative position matrix is ​​analyzed, the direct physical connection relationship between adjacent nodes is identified, the energy / information transmission path between the nodes is further analyzed, the logical topology association relationship between the nodes is excavated, and the identified node physical connection relationship and logical topology association relationship are integrated into a node topology association table.

[0094] Step S3: Based on the microgrid node topology association representation and the state convolution graphs of multiple microgrids, dynamic state attribute mapping is performed to construct a microgrid topology twin model;

[0095] In this embodiment, by analyzing the position characteristics of the nodes in the topological structure and the dynamic change characteristics reflected in the state convolution diagram, corresponding dynamic state attributes are assigned to each node. These dynamic state attributes include electrical indicators such as node voltage, power, frequency, and environmental parameters such as temperature and humidity. These node dynamic state attributes are associated with the topological structure to form a comprehensive model that describes the time-varying behavior of the entire microgrid. Based on the established node dynamic state-topology association model, a topological twin model that reflects the overall structure and operating status of the microgrid is constructed. The topological twin model not only includes the physical topological structure of the microgrid, but also integrates the dynamic state attribute information of each node. Through the simulation operation of the topological twin model, the overall operating state changes of the microgrid system are simulated and predicted in real time. The constructed topological twin model is bidirectionally interacted with the actual operating system of the microgrid to achieve intelligent monitoring and optimization of the real system.

[0096] Step S4: performing grid state trend evolution on the state convolution graphs of multiple microgrids one by one, and performing energy demand forecasting to generate multiple microgrid energy demand forecast values;

[0097] In this embodiment, according to the state convolution graph of each microgrid, the time evolution trend of the multidimensional state parameters of each node is analyzed, and for each node in the state convolution graph, the historical changes of its key state parameters such as power, voltage, and current are extracted. Based on time series analysis methods, such as ARIMA model, neural network, etc., the change trend of these state parameters in a certain period of time in the future is predicted, and the prediction results of the state parameters of each node are summarized to obtain the state evolution trend of the entire microgrid. Combined with the type (power generation, energy storage, load, etc.) and capacity information of the microgrid node, the prediction results of the node state parameters are converted into energy demand predictions. For example, for load nodes, the future electricity demand is calculated based on the power prediction results; for power generation nodes, the future power generation demand is calculated based on the output prediction results, and the energy demand prediction results of various nodes are integrated to obtain the overall energy demand prediction of the entire microgrid in a certain period of time in the future. The state trend evolution and energy demand prediction are performed on multiple microgrids one by one to obtain the energy demand prediction value of each microgrid in a certain period of time in the future, and the energy demand prediction values ​​of multiple microgrids are integrated into a comprehensive prediction result to provide a basis for the overall planning and scheduling of the microgrid group.

[0098] Step S5: Based on multiple microgrid energy demand forecast values, dynamic energy scheduling calculation is performed on the microgrid topology twin model, and distributed collaborative decision-making is performed to construct a microgrid dynamic scheduling strategy;

[0099] In this embodiment, a topological twin model is used to perform optimized scheduling calculations based on the energy demand forecast of each microgrid. The goal of the scheduling calculation is to minimize costs, maximize renewable energy utilization, etc. During the optimization calculation process, the topological association constraints and dynamic state coupling relationships between microgrid nodes are considered to obtain the optimal scheduling plan that meets the energy demand forecast, including the output of each power generation node, the charging and discharging plan of the energy storage node, etc. The above-mentioned optimal scheduling plan for a single microgrid is further extended to a microgrid group, and a distributed collaborative decision-making mechanism is adopted to coordinate the energy scheduling between multiple microgrids. Multi-agent game, alliance mechanism and other methods are used to promote information sharing and collaborative optimization between microgrids, and a microgrid dynamic scheduling strategy based on topological twin model and energy demand forecast is constructed to achieve collaborative operation of microgrid groups.

[0100] Step S6: Iteratively fine-tune the topology structure of the microgrid topology twin model according to the microgrid dynamic dispatch strategy, and construct a microgrid topology optimization structure model to perform intelligent energy dispatch operations.

[0101] In this embodiment, according to the dynamic dispatching strategy of the microgrid, the dispatching strategy is simulated and implemented in the topological twin model, and the impact of the dispatching strategy on the microgrid system indicators (cost, renewable utilization rate, reliability, etc.) under the premise of meeting the energy demand forecast is evaluated. The problems encountered in the execution of the dispatching strategy, such as overload of some nodes and imbalance of supply and demand in some areas, are analyzed, and the microgrid topology structure is fine-tuned. The fine-tuning methods include adding / deleting lines, adjusting node capacity / type, etc., setting optimization goals, such as further reducing costs, improving the utilization rate of renewable energy, etc., and guiding the iterative optimization of the topological structure. For the microgrid system in actual operation, the optimized structure model is used to make dynamic dispatching decisions, and the twin model is used for simulation evaluation to ensure that the dispatching decision can not only meet the energy supply and demand balance of the system, but also optimize the overall topological structure of the microgrid. The optimized dispatching plan is fed back to the actual microgrid system to guide its intelligent operation, and this optimized topological structure model is used as the basis for intelligent energy dispatching of the microgrid.

[0102] In this embodiment, refer to Figure 2 , is a flowchart of detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:

[0103] Step S11: Acquire multi-dimensional state parameters of multiple microgrids;

[0104] Step S12: extracting the load and power generation of each microgrid based on the multi-dimensional state parameters of the multiple microgrids;

[0105] Step S13: performing electrical parameter characteristic analysis on the load and power generation of each microgrid to obtain electrical parameter characteristics of multiple microgrids;

[0106] Step S14: performing multi-time window change analysis on the electrical parameter characteristics of the multiple microgrids to generate time sequence state change curves of the multiple microgrids;

[0107] Step S15: performing deep convolution learning on the time-series state change curves of the multiple microgrids to generate state convolution graphs of the multiple microgrids.

[0108] In this embodiment, the Internet of Things devices deployed at each microgrid node collect the operating status parameters of multiple subsystems such as generator sets, energy storage devices, and loads in real time. These status parameters include multi-dimensional indicators such as power, voltage, current, frequency, temperature, and humidity. The real-time load power and capacity of each microgrid are calculated using parameters such as voltage and current. The real-time power generation of each microgrid is calculated based on the power generation of distributed power sources (such as photovoltaics, wind power, etc.) within the microgrid. The calculated load and power generation of each microgrid are used as basic data for subsequent analysis. Based on the load and power generation data, electrical parameter analysis is further performed, including feature extraction of key electrical quantities such as voltage, current, power factor, and harmonic content. Fourier transform, wavelet analysis, and other methods are used to deeply mine each The electrical parameter characteristics of the microgrid are analyzed according to the time series. Different time windows are set, such as 1 hour, 6 hours, 12 hours, etc., to observe the changing trends of various parameters on different time scales. Based on the time series changing trends of these parameters, the state change curves of multiple microgrids are drawn. These time series state change curves reflect the dynamic characteristics of the operating status of each microgrid. According to the state change curves of each microgrid, they are input into a deep convolutional neural network for learning. The convolutional network extracts the complex spatiotemporal characteristics contained in the curve and generates a high-dimensional feature representation of the microgrid state. These high-dimensional feature representations are visualized as a state convolution map similar to an image. This state convolution map comprehensively reflects the operating status of multiple microgrids in time and space dimensions.

[0109] In this embodiment, the specific steps of step S14 are:

[0110] Define the time window length;

[0111] Based on the time window length, multiple sliding window slice analyses are performed on the electrical parameter characteristics of multiple microgrids to generate a timing parameter characteristic of each microgrid;

[0112] The parameter variation of adjacent time windows is analyzed for the timing parameter characteristics of each microgrid to obtain the multi-segment timing parameter variation characteristics;

[0113] Extract timestamps of the characteristics of changes in multiple time series parameters;

[0114] Performing sequential serialization processing on multiple sections of timing parameter change characteristics according to the timestamp to generate a microgrid timing parameter change characteristic sequence;

[0115] The state change curve is fitted for the characteristic sequence of microgrid timing parameter changes to generate multiple microgrid timing state change curves.

[0116] In this embodiment, according to the actual application scenario and data characteristics, a suitable time window length is selected. The selection of the time window length needs to balance the analysis accuracy and the calculation complexity. Generally, time window lengths of different scales such as 1 hour, 6 hours, and 12 hours are selected to cover multiple time scales of microgrid state changes. For each microgrid, its electrical parameter characteristic data is slidingly sliced ​​with the set time window length. For each time window, the corresponding voltage, current, power and other electrical parameter characteristic values ​​are extracted, and the electrical parameter characteristic values ​​of each time window are arranged in time series to generate the time series parameter characteristic data of each microgrid. The electrical parameter characteristic values ​​in adjacent time windows are compared, and the change amplitude and change trend of the parameter value are calculated. According to the significance of the parameter change, the time series parameter characteristics are divided into multiple change segments. For each change segment, The changing characteristics of parameter values, such as the amplitude and rate of change, are extracted, and the start and end time points of each parameter change segment are recorded as the timestamp of the change segment. The timestamps of all change segments are sorted into a time series to form a time axis of the microgrid state change. According to the order of the timestamps, the parameter change characteristics of each change segment are connected in series into a complete time series sequence. For some missing time points in the middle, the corresponding parameter values ​​are filled in using interpolation and other methods to finally form a parameter feature sequence that reflects the full time series change of the microgrid state. Mathematical modeling methods such as curve fitting are used to model the microgrid time series parameter change feature sequence. According to the fitting results, the state change curve of each microgrid is drawn to show the evolution trajectory of its operating state over time. The state change curves of different microgrids are compared and analyzed to find the differences and commonalities between them.

[0117] In this embodiment, refer to Figure 3 , is a flowchart of detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:

[0118] Step S21: identifying multiple microgrid nodes;

[0119] Step S22: performing spatial position calculation on a plurality of microgrid nodes to generate spatial position coordinates of the microgrid nodes;

[0120] Step S23: performing relative position analysis on multiple microgrid nodes one by one to generate relative position relationships between the nodes;

[0121] Step S24: performing adjacent interactive collaborative analysis on the relative position relationship between the nodes to generate adjacent node interactive collaborative data;

[0122] Step S25: performing node topology association mining on the interactive collaborative data between adjacent nodes based on the spatial position coordinates of the microgrid nodes to generate a microgrid node topology association representation.

[0123] In this embodiment, by identifying and classifying various types of equipment and facilities in the microgrid system, the various nodes constituting the microgrid are determined. The nodes include various types of energy units such as power generation equipment, energy storage devices, and load equipment. For each node, the microgrid to which it belongs and the node type are determined. The actual spatial position of each node is calculated using the microgrid topology map and geographic information system data. The coordinate information of the node in two-dimensional or three-dimensional space is obtained by using methods such as triangulation and three-dimensional laser scanning. The spatial position coordinates of these nodes are integrated into a global microgrid node spatial distribution map. Based on the obtained node spatial position coordinates, the relative position relationship between any two nodes is calculated. The relative position relationship includes indicators such as the distance, angle, and height difference between the nodes. By calculating the relative position relationship between all nodes, a relative position matrix between nodes is formed. According to the relative position relationship between nodes, combined with the node type and actual function, the interactive coordination characteristics between adjacent nodes are analyzed. The interactive coordination characteristics include electrical parameter coordination indicators such as power exchange, frequency coordination, and phase matching. By analyzing the actual interaction between adjacent nodes, a data set describing the collaborative relationship between nodes is formed. Combined with the node spatial position coordinates and the interactive coordination data between adjacent nodes, the topological association characteristics between nodes are analyzed, and the topological characteristic parameters such as the connection relationship and connection strength between nodes are extracted. The node topological association model of the microgrid is constructed, and the node topological association differences between different microgrids are compared and analyzed to discover their unique network structure characteristics.

[0124] In this embodiment, reference Figure 4 The above is a schematic flow chart of the detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include:

[0125] Step S31: performing spatial distribution analysis on multiple microgrid nodes to obtain spatial distribution data of microgrid nodes;

[0126] Step S32: performing node space topology structure analysis on the microgrid node space distribution data based on the microgrid node topology association representation to obtain node space topology structure data;

[0127] Step S33: Perform global topological connection based on the node space topological structure data to construct a microgrid topological node space;

[0128] Step S34: Based on the state convolution graphs of multiple microgrids, dynamic state attribute mapping is performed on the microgrid topology node space to construct a microgrid topology twin model.

[0129] In this embodiment, according to the node spatial position coordinate data, the distribution characteristics of each node in the three-dimensional space are analyzed, and cluster analysis, spatial density estimation and other methods are used to identify the clustered areas and sparse areas of the nodes in the space, and the distribution characteristic data of the nodes in the space are organized into a microgrid node spatial distribution data set. Combined with the node topology association representation obtained above, the interconnection relationship of the nodes in the space is analyzed, and the topological structure characteristics such as clusters and subnetworks formed by the nodes in the space are identified. The key parameters of the node spatial topological structure, such as the node clustering coefficient and the network diameter, are extracted to form the node spatial topological structure data. The spatial position coordinates of each node and the topological connection relationship between the nodes are integrated into a topological node space model of the microgrid as a whole. The model not only includes the distribution position of the nodes in the space, but also describes the interconnection relationship between the nodes, intuitively displays the overall spatial topological structure of the microgrid, and provides a basis for subsequent analysis and application. The real-time status data of multiple microgrids during operation, such as voltage, power, temperature, etc., are collected, and these status data are mapped to the topological node space model using methods such as convolutional neural networks to form dynamic state attributes. The topological twin model reflects the operating status of the microgrid system in real time and provides support for intelligent decision-making.

[0130] In this embodiment, step S4 includes the following steps:

[0131] Step S41: performing grid state trend evolution on the state convolution graphs of multiple microgrids one by one to generate state trend evolution data of each microgrid;

[0132] Step S42: performing energy demand situation analysis on the state trend evolution data of each microgrid to obtain microgrid energy demand situation data;

[0133] Step S43: performing energy demand forecasting for continuous time periods on the microgrid energy demand situation data to obtain energy demand forecast values ​​for multiple time periods;

[0134] Step S44: performing iterative error optimization calculation on energy demand forecast values ​​for multiple time periods to generate multiple microgrid energy demand forecast values.

[0135] In this embodiment, based on the constructed microgrid topology twin model, the state change data of each microgrid in the time series is extracted, and the state convolution graph of each microgrid is trend predicted and evolved by using time series analysis, data mining and other methods. The state trend evolution data of each microgrid is organized into an independent data set. Combined with the load characteristics and energy supply of the microgrid, the energy demand situation of each microgrid is analyzed, and key indicators reflecting energy demand, such as peak-to-valley load ratio and load fluctuation amplitude, are extracted from the state trend evolution data. The energy demand situation characteristics of each microgrid are organized into a microgrid energy demand situation database. Time series prediction, machine learning and other methods are used to perform long-term prediction of the energy demand situation data of each microgrid. The prediction results cover multiple time periods, such as energy demand prediction values ​​of different granularities such as annual, quarterly, and monthly. A prediction database for future energy demand of each microgrid is established. Iterative optimization is performed according to the energy demand prediction value to reduce the prediction error. The prediction model parameters are dynamically updated by rolling prediction, adaptive adjustment and other methods to improve the prediction accuracy. Finally, energy demand prediction data of multiple microgrids in different time periods in the future are formed to provide support for subsequent optimization and scheduling.

[0136] In this embodiment, the specific steps of step S5 are:

[0137] Step S51: simulating the evolution of the operating state of the microgrid topology twin model to generate microgrid operation simulation data;

[0138] Step S52: Calculate the energy supply of each node on the microgrid operation simulation data to obtain energy supply parameters of each node;

[0139] Step S53: Based on multiple microgrid energy demand forecast values, energy supply parameters of each node are identified as energy demand shortage, and demand deviation microgrid nodes are marked;

[0140] Step S54: performing dynamic energy scheduling calculation on the demand deviation microgrid node to generate dynamic energy scheduling data;

[0141] Step S55: Perform distributed collaborative decision-making on dynamic energy scheduling data to build a microgrid dynamic scheduling strategy.

[0142] In this embodiment, the constructed microgrid topology twin model is input into a simulation engine, and the attributes and connection relationships of each node in the twin model are simulated and evolved to generate a data set reflecting the changes in the overall operating status of the microgrid. These operating simulation data include the process of changes in key electrical indicators such as node power, voltage, and frequency over time. The actual energy supply situation of each node is calculated, including the output capacity of the power generation equipment, the charging and discharging status of the energy storage device, etc., reflecting the supply capacity of each node. The energy supply parameters of these nodes are integrated into a data set describing the supply and demand status of the entire microgrid, and the energy demand forecast value is compared and analyzed with the node energy supply data to identify those energy Nodes whose supply cannot meet the predicted demand are marked as demand deviation microgrid nodes. A portrait of the demand deviation microgrid nodes is established to provide a basis for subsequent dynamic scheduling. For microgrid nodes with demand deviation, dynamic energy scheduling is performed using optimization algorithms and other methods. The scheduling strategy includes adjusting local power generation, coordinating energy exchange between nodes and other measures. Through dynamic scheduling calculations, energy scheduling plans for each node in the future time period are generated. The dynamic scheduling plans of each node are coordinated and integrated to form an overall dynamic scheduling strategy for the microgrid. Distributed optimization, multi-intelligence collaboration and other methods are used to fully consider the demands and constraints of each node, and finally a dynamic scheduling strategy for the microgrid is generated to provide guidance for energy management in actual operations.

[0143] In this embodiment, the specific steps of step S54 are:

[0144] According to the microgrid topology twin model, the deviation node of the demand deviation microgrid node is located to obtain the demand deviation node position data;

[0145] Based on the demand deviation node location data, the adjacent nodes in the region are analyzed to obtain the microgrid nodes in the region;

[0146] According to the energy supply parameters of each node, the dispatchable microgrid nodes in the region are identified as nodes with sufficient energy, and the dispatchable temporary microgrid nodes are obtained;

[0147] Calculate the energy compensation demand of the demand deviation microgrid node and obtain the compensation demand value of the deviation node;

[0148] The dynamic energy scheduling calculation is performed on the compensation demand value of the deviation node by calling the adjacent microgrid node to generate dynamic energy scheduling data.

[0149] In this embodiment, combined with the previously generated microgrid operation simulation data, topology analysis and other methods are used to identify nodes whose energy supply cannot meet the predicted demand, and the location information of these demand deviation nodes is organized into demand deviation node location data, which provides a basis for subsequent regional analysis. For each demand deviation node, its adjacent neighboring nodes are analyzed to divide the regional range where the demand deviation node is located, and all microgrid node information in the area is organized into microgrid node data in the area to provide a reference for energy compensation. Combined with the previously established energy supply database, the energy supply capacity of each node in the area is evaluated, and those nodes with sufficient energy supply are screened out to provide regulation. The nodes supported by the degree are used as callable temporary microgrid nodes. The information of these callable nodes is organized into callable temporary microgrid node data. The energy demand forecast value of the demand deviation node is compared with the actual energy supply, and the gap value, that is, the compensation demand, is calculated. The compensation demand value of each demand deviation node is organized into deviation node compensation demand data. The callable temporary microgrid nodes are coordinated by using optimization algorithms and other methods to meet the compensation demand of the deviation nodes. The scheduling strategy includes measures such as adjusting local power generation and coordinating energy exchange between nodes. Through dynamic scheduling calculation, the energy scheduling plan for each node in the future time period is generated to form dynamic energy scheduling data.

[0150] In this embodiment, the specific steps of step S6 are:

[0151] Step S61: Calculate the energy supply and demand efficiency of the microgrid topology twin model to generate the supply and demand efficiency of the current topology model;

[0152] Step S62: dynamically optimizing the microgrid dynamic dispatch strategy based on the supply and demand efficiency of the current topology model to generate a dynamic dispatch efficiency optimization strategy;

[0153] Step S63: Iteratively fine-tune the topology structure of the microgrid topology twin model according to the dynamic scheduling efficiency optimization strategy, and construct a microgrid topology optimization structure model to perform intelligent energy scheduling operations.

[0154] In this embodiment, according to the microgrid topology twin model, the energy supply and demand data of each node are input, and the efficiency indicators between the energy supply and demand of the entire microgrid system, such as energy utilization rate, energy loss rate, etc., are calculated by using methods such as energy flow analysis, and a dynamic scheduling optimization model is designed. The goal is to improve the energy supply and demand efficiency of the entire microgrid system. The optimization model includes control strategies such as energy scheduling between nodes, energy storage scheduling, and distributed power generation output adjustment. The optimization model is solved by an optimization algorithm (such as genetic algorithm, particle swarm optimization, etc.) to obtain a dynamic scheduling optimization strategy that can improve the overall supply and demand efficiency. The dynamic scheduling efficiency optimization strategy is fed back to the microgrid topology twin model. According to the optimization strategy, the node connection relationship, energy transmission lines, etc. in the model are fine-tuned to further improve the overall supply and demand efficiency. After iterative optimization, a topology-optimized microgrid structure model is obtained, that is, a microgrid topology optimization structure model. The optimized structure model is applied to actual intelligent energy scheduling to guide the real-time operation and optimization of the microgrid.

[0155] In this embodiment, a multi-microgrid intelligent energy dispatching system based on deep learning is provided, which is used to execute the multi-microgrid intelligent energy dispatching method based on deep learning as described above, including:

[0156] A multi-dimensional state change module is used to obtain multi-dimensional state parameters of multiple microgrids; perform multi-time window change analysis on the multi-dimensional state parameters of multiple microgrids, and perform deep convolution learning to generate state convolution graphs of multiple microgrids;

[0157] A topological association mining module is used to identify multiple microgrid nodes; analyze the relative positions of multiple microgrid nodes one by one, and perform node topological association mining to generate a microgrid node topological association representation;

[0158] The twin model module is used to map dynamic state attributes based on the topological association representation of microgrid nodes and the state convolution graphs of multiple microgrids to construct a microgrid topology twin model;

[0159] The energy demand prediction module is used to perform the state trend evolution of multiple microgrids one by one on the state convolution graphs, and perform energy demand prediction to generate multiple microgrid energy demand prediction values;

[0160] The dynamic energy scheduling module is used to perform dynamic energy scheduling calculations on the microgrid topology twin model based on multiple microgrid energy demand forecast values, and to make distributed collaborative decisions to build a microgrid dynamic scheduling strategy;

[0161] The topology optimization module is used to iteratively fine-tune the topology structure of the microgrid topology twin model according to the microgrid dynamic scheduling strategy, and to build a microgrid topology optimization structure model to perform intelligent energy scheduling operations.

[0162] The present invention obtains multi-dimensional state parameters of multiple microgrids to comprehensively understand the operation status of the microgrid system, provides a data basis for subsequent analysis and optimization, performs multi-time window change analysis and deep convolution learning to capture the time series characteristics of the microgrid state, generates a state convolution graph to reveal the complex relationship between the microgrid states, identifies multiple microgrid nodes and performs node topology association mining to establish the relationship between the nodes in the microgrid, provides key information for subsequent model construction, generates microgrid node topology association representation to help understand the correlation between the internal nodes of the microgrid, provides input features for the construction of the twin model, constructs a microgrid topology twin model based on the microgrid node topology association representation and the state convolution graph to perform dynamic state attribute mapping, realizes the fusion modeling of the microgrid state and structure, constructs the twin model to comprehensively consider the structural characteristics and state information of the microgrid, and provides more information for subsequent energy scheduling. Accurate model support, energy demand forecasting for the state convolution graphs of multiple microgrids to predict future energy demand trends, provide forecasting basis for dynamic energy scheduling, generate energy demand forecast values ​​to help the microgrid system make adjustments in advance, optimize energy allocation strategies, and improve energy utilization efficiency. Dynamic energy scheduling calculations are performed based on energy demand forecast values. Energy allocation is adjusted according to real-time energy demand conditions to improve the flexibility and response speed of energy scheduling. A dynamic scheduling strategy for distributed collaborative decision-making is constructed to coordinate scheduling between internal nodes of the microgrid to achieve overall energy supply and demand balance. According to the dynamic scheduling strategy, the topology structure of the microgrid topology twin model is iteratively fine-tuned to optimize the structure of the microgrid, improve energy transmission efficiency and system stability, and construct a topology optimization structure model to fine-tune the connection relationship between internal nodes of the microgrid to provide more efficient support for intelligent energy scheduling operations.

[0163] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0164] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A multi-microgrid intelligent energy scheduling method based on deep learning, characterized in that: The following steps are involved: Step S1: obtaining multi-dimensional state parameters of multiple microgrids; performing multi-time window change analysis on the multi-dimensional state parameters of the multiple microgrids, and performing deep convolution learning to generate state convolution graphs of the multiple microgrids; Step S2: identifying multiple microgrid nodes; performing relative position analysis on multiple microgrid nodes one by one, and performing node topology association mining to generate a microgrid node topology association representation; Step S3: Based on the microgrid node topology association representation and the state convolution graphs of multiple microgrids, dynamic state attribute mapping is performed to construct a microgrid topology twin model; Step S4: performing grid state trend evolution on the state convolution graphs of multiple microgrids one by one, and performing energy demand forecasting to generate multiple microgrid energy demand forecast values; Step S5: Based on multiple microgrid energy demand forecast values, dynamic energy scheduling calculation is performed on the microgrid topology twin model, and distributed collaborative decision-making is performed to construct a microgrid dynamic scheduling strategy; Step S6: Iteratively fine-tune the topology structure of the microgrid topology twin model according to the microgrid dynamic dispatch strategy, and construct a microgrid topology optimization structure model to perform intelligent energy dispatch operations.

2. The multi-microgrid intelligent energy scheduling method based on deep learning according to claim 1 is characterized in that: The specific steps of step S1 are: Step S11: Acquire multi-dimensional state parameters of multiple microgrids; Step S12: extracting the load and power generation of each microgrid based on the multi-dimensional state parameters of the multiple microgrids; Step S13: performing electrical parameter characteristic analysis on the load and power generation of each microgrid to obtain electrical parameter characteristics of multiple microgrids; Step S14: performing multi-time window change analysis on the electrical parameter characteristics of the multiple microgrids to generate time sequence state change curves of the multiple microgrids; Step S15: performing deep convolution learning on the time-series state change curves of the multiple microgrids to generate state convolution graphs of the multiple microgrids.

3. The multi-microgrid intelligent energy scheduling method based on deep learning according to claim 1 is characterized in that: The specific steps of step S14 are: Define the time window length; Based on the time window length, multiple sliding window slice analyses are performed on the electrical parameter characteristics of multiple microgrids to generate a timing parameter characteristic of each microgrid; The parameter variation of adjacent time windows is analyzed for the timing parameter characteristics of each microgrid to obtain the multi-segment timing parameter variation characteristics; Extract timestamps of changing features of multiple time series parameters; Sequentially serializing the multiple-segment timing parameter change characteristics according to the timestamp to generate a microgrid timing parameter change characteristic sequence; The state change curve is fitted for the characteristic sequence of microgrid timing parameter changes to generate multiple microgrid timing state change curves.

4. The multi-microgrid intelligent energy dispatching method based on deep learning according to claim 1 is characterized in that: The specific steps of step S2 are: Step S21: identifying multiple microgrid nodes; Step S22: performing spatial position calculation on a plurality of microgrid nodes to generate spatial position coordinates of the microgrid nodes; Step S23: performing relative position analysis on multiple microgrid nodes one by one to generate relative position relationships between the nodes; Step S24: performing adjacent interactive collaborative analysis on the relative position relationship between the nodes to generate adjacent node interactive collaborative data; Step S25: performing node topology association mining on the interactive collaborative data between adjacent nodes based on the spatial position coordinates of the microgrid nodes to generate a microgrid node topology association representation.

5. The multi-microgrid intelligent energy dispatching method based on deep learning according to claim 1 is characterized in that: The specific steps of step S3 are: Step S31: performing spatial distribution analysis on multiple microgrid nodes to obtain spatial distribution data of microgrid nodes; Step S32: performing node space topology structure analysis on the microgrid node space distribution data based on the microgrid node topology association representation to obtain node space topology structure data; Step S33: Perform global topological connection based on the node space topological structure data to construct a microgrid topological node space; Step S34: Based on the state convolution graphs of multiple microgrids, dynamic state attribute mapping is performed on the microgrid topology node space to construct a microgrid topology twin model.

6. The multi-microgrid intelligent energy dispatching method based on deep learning according to claim 1 is characterized in that: The specific steps of step S4 are: Step S41: performing grid state trend evolution on the state convolution graphs of multiple microgrids one by one to generate state trend evolution data of each microgrid; Step S42: performing energy demand situation analysis on the state trend evolution data of each microgrid to obtain microgrid energy demand situation data; Step S43: performing energy demand forecasting for continuous time periods on the microgrid energy demand situation data to obtain energy demand forecast values ​​for multiple time periods; Step S44: performing iterative error optimization calculation on energy demand forecast values ​​for multiple time periods to generate multiple microgrid energy demand forecast values.

7. The multi-microgrid intelligent energy dispatching method based on deep learning according to claim 1 is characterized in that: The specific steps of step S5 are: Step S51: simulating the evolution of the operating state of the microgrid topology twin model to generate microgrid operation simulation data; Step S52: Calculate the energy supply of each node on the microgrid operation simulation data to obtain energy supply parameters of each node; Step S53: Based on multiple microgrid energy demand forecast values, energy supply parameters of each node are identified as energy demand shortage, and demand deviation microgrid nodes are marked; Step S54: performing dynamic energy scheduling calculation on the demand deviation microgrid node to generate dynamic energy scheduling data; Step S55: Perform distributed collaborative decision-making on dynamic energy scheduling data to build a microgrid dynamic scheduling strategy.

8. The multi-microgrid intelligent energy dispatching method based on deep learning according to claim 7 is characterized in that: The specific steps of step S54 are: According to the microgrid topology twin model, the deviation node of the demand deviation microgrid node is located to obtain the demand deviation node position data; Based on the demand deviation node location data, the neighboring nodes in the region are analyzed to obtain the microgrid nodes in the region; According to the energy supply parameters of each node, the dispatchable microgrid nodes in the region are identified as nodes with sufficient energy, and the dispatchable neighboring microgrid nodes are obtained; Calculate the energy compensation demand of the demand deviation microgrid node and obtain the compensation demand value of the deviation node; The dynamic energy scheduling calculation of the deviation node compensation demand value is performed by calling the neighboring microgrid nodes to generate dynamic energy scheduling data.

9. The multi-microgrid intelligent energy dispatching method based on deep learning according to claim 1 is characterized in that: The specific steps of step S6 are: Step S61: Calculate the energy supply and demand efficiency of the microgrid topology twin model to generate the supply and demand efficiency of the current topology model; Step S62: dynamically optimizing the microgrid dynamic dispatch strategy based on the supply and demand efficiency of the current topology model to generate a dynamic dispatch efficiency optimization strategy; Step S63: Iteratively fine-tune the topology structure of the microgrid topology twin model according to the dynamic scheduling efficiency optimization strategy, and construct a microgrid topology optimization structure model to perform intelligent energy scheduling operations.

10. A multi-microgrid intelligent energy dispatching system based on deep learning, characterized in that: The method for executing the multi-microgrid intelligent energy dispatching method based on deep learning as claimed in claim 1 comprises: A multi-dimensional state change module is used to obtain multi-dimensional state parameters of multiple microgrids; perform multi-time window change analysis on the multi-dimensional state parameters of multiple microgrids, and perform deep convolution learning to generate state convolution graphs of multiple microgrids; A topological association mining module is used to identify multiple microgrid nodes; analyze the relative positions of multiple microgrid nodes one by one, and perform node topological association mining to generate a microgrid node topological association representation; The twin model module is used to map dynamic state attributes based on the topological association representation of microgrid nodes and the state convolution graphs of multiple microgrids to construct a microgrid topology twin model; The energy demand prediction module is used to perform the state trend evolution of multiple microgrids one by one on the state convolution graphs, and perform energy demand prediction to generate multiple microgrid energy demand prediction values; The dynamic energy scheduling module is used to perform dynamic energy scheduling calculations on the microgrid topology twin model based on multiple microgrid energy demand forecast values, and to make distributed collaborative decisions to build a microgrid dynamic scheduling strategy; The topology optimization module is used to iteratively fine-tune the topology structure of the microgrid topology twin model according to the microgrid dynamic scheduling strategy, and to construct a microgrid topology optimization structure model to perform intelligent energy scheduling operations.

Images