Power transmission quantity data analysis method and system based on virtual power plant, electronic equipment and storage medium

Abstract

The invention discloses a power transmission quantity data analysis method and system based on a virtual power plant, electronic equipment and a storage medium, and belongs to the technical field of data analysis. Through an AI knowledge mining branch in a joint data analysis algorithm, initial distributed sensing measurement information is deeply mined, and a power transmission state change knowledge vector containing rich power grid power transmission state information is obtained. Therefore, the monitoring precision of the power transmission state of the power grid is improved, and comprehensive and real-time data support is provided for the scheduling decision of a power grid company. Meanwhile, by combining a distributed lightweight algorithm and an edge decision tree branch in the virtual power plant resource coordination system, accurate optimization and scheduling of the power transmission state of the power grid are realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of data analysis, and specifically relates to a power transmission data analysis technology based on a virtual power plant, which is specifically used for analyzing distributed data of a virtual power plant. Background Art

[0002] Virtual Power Plant (VPP) is a power coordination management system that uses advanced information and communication technologies and software systems to achieve the aggregation and coordinated optimization of distributed energy resources (DER) such as distributed power sources (such as solar energy, wind energy, etc.), energy storage systems, controllable loads, electric vehicles, etc., so as to participate in the power market and power grid operation as a special power plant. This concept was first proposed by Dr. ShimoY Awerbuch in 1997, and its core lies in "communication" and "aggregation".

[0003] In terms of the relationship with real power plants, virtual power plants are not physical power plants in the traditional sense, such as thermal power plants or hydropower plants. Virtual power plants do not have the physical form of physical power plants themselves. However, virtual power plants play a similar or even superior role to real power plants in terms of function. Virtual power plants use information technology to integrate small distributed energy resources scattered in different spaces to form a controllable virtual power source, participate in power market transactions, assist power system operations, optimize resource utilization, and maintain stable and safe power consumption.

[0004] In the current power grid management, real-time monitoring and optimized dispatching of transmission status are the key to ensuring the stable operation of the power grid. However, traditional power grid monitoring methods often rely on single sensor data and simple numerical analysis, which makes it difficult to fully and real-time reflect the transmission status of the power grid. At the same time, with the rise of virtual power plants, the distributed characteristics of the power grid have become increasingly prominent, and traditional centralized data analysis methods can no longer meet the needs of efficient processing of large amounts of distributed sensor data. In addition, the formulation of power grid dispatching decisions often lacks sufficient data support and in-depth analysis, resulting in low dispatching efficiency and failure to meet the development needs of smart grids. Summary of the invention

[0005] The present invention provides a method and system for analyzing power transmission data based on a virtual power plant, which can solve or partially solve the technical problems involved in the above-mentioned background technology.

[0006] The embodiment of the present invention provides a power transmission data analysis method based on a virtual power plant, which is applied to a power transmission data analysis middle-end system. The method comprises: obtaining X initial distributed sensor metering information, performing state change knowledge mining on the X initial distributed sensor metering information through an AI knowledge mining branch in a joint data analysis algorithm, and obtaining power transmission state change knowledge vectors corresponding to the X initial distributed sensor metering information respectively; X is a positive integer; the joint data analysis algorithm further comprises a deep residual branch and a multivariate decision tree branch respectively set in X virtual power plant resource coordination systems; the X initial distributed sensor metering information is obtained through the AI ​​knowledge mining branch in the joint data analysis algorithm. The X virtual power plant resource coordination systems respectively extract the associated intelligent sensor collection information through a pre-matched distributed lightweight algorithm; the distributed lightweight algorithm pre-matched by each virtual power plant resource coordination system is obtained after optimizing the branch network weight of the deep residual branch; one of the X transmission state change knowledge vectors is sent to the X virtual power plant resource coordination systems; the transmission state change knowledge vector received by the target virtual power plant resource coordination system corresponds to the initial distributed sensor metering information extracted by the target virtual power plant resource coordination system; the target virtual power plant resource coordination system for any one of the X virtual power plant resource coordination systems; when the distributed lightweight algorithms and edge decision tree branches respectively set by the X virtual power plant resource coordination systems all meet the edge side debugging compliance requirements, receiving the target algorithm debugging errors respectively reported by the X virtual power plant resource coordination systems; the target algorithm debugging errors reported by the target virtual power plant resource coordination system are generated by the edge decision tree branches pre-matched by the target virtual power plant resource coordination system and the received transmission state change knowledge vector, and each edge decision tree branch pre-matched by the virtual power plant resource coordination system is a multivariate decision tree branch. The cross decision tree network is obtained by optimizing the branch network weight of the AI ​​knowledge mining branch; the branch network weight of the AI ​​knowledge mining branch is optimized through X target algorithm debugging errors to obtain a cross knowledge mining module, and based on the cross knowledge mining module and the distributed lightweight algorithms and edge decision tree branches in the X virtual power plant resource coordination systems, a cross decision tree network corresponding to each virtual power plant resource coordination system is generated; wherein the cross decision tree network linked and implanted in the transmission data analysis middle platform system and the target virtual power plant resource coordination system is used to perform power load optimization scheduling analysis on the intelligent sensor collection information in the target virtual power plant resource coordination system

[0007] An embodiment of the present invention provides a power transmission data analysis middle-station system, which communicates with a virtual power plant resource coordination system, and is used to: obtain X initial distributed sensor metering information, perform state change knowledge mining on the X initial distributed sensor metering information through an AI knowledge mining branch in a joint data analysis algorithm, and obtain power transmission state change knowledge vectors corresponding to the X initial distributed sensor metering information; X is a positive integer; the joint data analysis algorithm also includes a deep residual branch and a multivariate decision tree branch respectively set in the X virtual power plant resource coordination systems; the X initial distributed sensor metering information is used to obtain a state change knowledge vector ... The distributed sensor metering information is obtained by extracting the associated intelligent sensor collection information through the pre-matched distributed lightweight algorithms of the X virtual power plant resource coordination systems; the distributed lightweight algorithms pre-matched by each virtual power plant resource coordination system are obtained after optimizing the branch network weights of the deep residual branches; one of the X transmission state change knowledge vectors is sent to the X virtual power plant resource coordination systems; the transmission state change knowledge vector received by the target virtual power plant resource coordination system corresponds to the initial distributed sensor metering information extracted by the target virtual power plant resource coordination system; the target virtual power plant resource coordination system is configured to obtain the distributed sensor metering information of the target virtual power plant resource coordination system; the distributed sensor metering information of the target virtual power plant resource coordination system is obtained by extracting the associated intelligent sensor collection information through the pre-matched distributed lightweight algorithms of the X virtual power plant resource coordination systems; the distributed lightweight algorithms pre-matched by each virtual power plant resource coordination system are obtained after optimizing the branch network weights of the deep residual branches; the transmission state change knowledge vector of the X transmission state change knowledge vectors is sent to the X virtual power plant resource coordination systems respectively; the transmission state change knowledge vector received by the target virtual power plant resource coordination system corresponds to the initial distributed sensor metering information extracted by the target virtual power plant resource coordination system; the target virtual power plant resource coordination system is configured to obtain the distributed sensor metering information of the target virtual power plant resource coordination system; the distributed sensor metering information of the target virtual power plant resource coordination system is obtained by extracting the associated intelligent sensor collection information through the pre-matched distributed lightweight algorithms of the target virtual power plant resource coordination system; the distributed lightweight algorithms pre-matched by each virtual power plant resource coordination system are obtained by optimizing the branch network weights of the deep residual branches ... The plant resource coordination system is any one of the X virtual power plant resource coordination systems; when the distributed lightweight algorithms and edge decision tree branches respectively set by the X virtual power plant resource coordination systems meet the edge side debugging compliance requirements, the target algorithm debugging errors respectively reported by the X virtual power plant resource coordination systems are received; the target algorithm debugging errors reported by the target virtual power plant resource coordination system are generated by the edge decision tree branches pre-matched by the target virtual power plant resource coordination system and the received transmission state change knowledge vector, and each edge decision tree branch pre-matched by the virtual power plant resource coordination system is a target algorithm debugging error reported by the target virtual power plant resource coordination system. The cross decision tree network is obtained by optimizing the branch network weight of the multivariate decision tree branch; the branch network weight of the AI ​​knowledge mining branch is optimized through X target algorithm debugging errors to obtain a cross knowledge mining module, and based on the cross knowledge mining module and the distributed lightweight algorithms and edge decision tree branches in the X virtual power plant resource coordination systems, a cross decision tree network corresponding to each virtual power plant resource coordination system is generated; wherein, the cross decision tree network linked and implanted in the power transmission data analysis middle platform system and the target virtual power plant resource coordination system is used to perform power load optimization scheduling analysis on the intelligent sensor collection information in the target virtual power plant resource coordination system.

[0008] The present invention also provides an electronic device, comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the above method.

[0009] Preferably, the memory may be a separate device independent of the processor, or may be integrated into the processor.

[0010] Preferably, the electronic device may further include a transceiver, and the processor may control the transceiver to interact with other devices, specifically, may send information or data to other devices, or receive information or data sent by other devices.

[0011] The present invention further provides a readable storage medium, on which a program or instruction is stored, and the program or instruction implements the steps of the above method when executed by a processor.

[0012] In the embodiment of the present invention, the initial distributed sensor metering information is deeply mined by combining the AI ​​knowledge mining branch in the data analysis algorithm to obtain the transmission state change knowledge vector containing rich power grid transmission state information. In this way, not only the monitoring accuracy of the power grid transmission state is improved, but also comprehensive and real-time data support is provided for the dispatching decision of the power grid company. At the same time, by combining the distributed lightweight algorithm and the edge decision tree branch in the virtual power plant resource coordination system, the precise optimization and dispatching of the power grid transmission state is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A flow chart of a method for analyzing power transmission data based on a virtual power plant provided in an embodiment of the present invention.

[0014] Figure 2 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0015] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are 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.

[0016] Figure 1 A power transmission data analysis method based on a virtual power plant is shown, which is applied to a power transmission data analysis middle-end system. The method includes the following steps 202-208.

[0017] In one application scenario, a power grid company manages X virtual power plant resource coordination systems (referred to as virtual power plants in the embodiment of the present invention) distributed in various places. These virtual power plants are connected to the power transmission data analysis platform system through intelligent sensor devices. In order to grasp the power transmission status of the power grid in real time and optimize the dispatch, the power grid company decided to use the power transmission data analysis platform system to conduct in-depth analysis of the sensor data of the virtual power plants.

[0018] Step 202: Obtain X pieces of initial distributed sensor metering information, perform state change knowledge mining on the X pieces of initial distributed sensor metering information through the AI ​​knowledge mining branch in the joint data analysis algorithm, and obtain power transmission state change knowledge vectors corresponding to the X pieces of initial distributed sensor metering information.

[0019] Wherein, X is a positive integer; the joint data analysis algorithm also includes a deep residual branch and a multivariate decision tree branch respectively set in X virtual power plant resource coordination systems; the X initial distributed sensor metering information is obtained by extracting the associated intelligent sensor acquisition information through the pre-matched distributed lightweight algorithm of the X virtual power plant resource coordination system; the distributed lightweight algorithm pre-matched by each virtual power plant resource coordination system is obtained after optimizing the branch network weight of the deep residual branch.

[0020] First, the power transmission data analysis middle-end system obtains X initial distributed sensor metering information from X virtual power plant resource coordination systems. The initial distributed sensor metering information is obtained by the virtual power plant through the pre-matched distributed lightweight algorithm to extract the intelligent sensor collection information. Each virtual power plant resource coordination system has a corresponding distributed lightweight algorithm, which is obtained by optimizing the branch network weights of the deep residual branch in the joint data analysis algorithm to ensure the accuracy and efficiency of the algorithm.

[0021] After obtaining the initial distributed sensor metering information, the power transmission data analysis center system uses the AI ​​knowledge mining branch in the joint data analysis algorithm to perform state change knowledge mining on the initial distributed sensor metering information. The AI ​​knowledge mining branch can analyze the implicit patterns in the initial distributed sensor metering information, thereby obtaining the power transmission state change knowledge vector corresponding to each initial distributed sensor metering information. The power transmission state change knowledge vector contains important information about the power transmission state change, such as transmission load, voltage stability, current fluctuation, etc.

[0022] Step 204: Send one power transmission state change knowledge vector among the X power transmission state change knowledge vectors to each of the X virtual power plant resource coordination systems.

[0023] Among them, the transmission state change knowledge vector received by the target virtual power plant resource coordination system corresponds to the initial distributed sensor metering information extracted by the target virtual power plant resource coordination system; the target virtual power plant resource coordination system is any one of the X virtual power plant resource coordination systems.

[0024] After obtaining X transmission state change knowledge vectors, the transmission quantity data analysis middle-end system sends the transmission state change knowledge vectors back to the corresponding X virtual power plant resource coordination systems. The transmission state change knowledge vectors received by each virtual power plant resource coordination system correspond to the initial distributed sensor metering information extracted. In this way, each virtual power plant resource coordination system can obtain real-time feedback on the transmission state of the power grid within its jurisdiction, and perform preliminary power grid dispatch optimization based on these feedbacks.

[0025] Step 206: When the distributed lightweight algorithms and edge decision tree branches respectively set by the X virtual power plant resource coordination systems all meet the edge side debugging compliance requirements, receive the target algorithm debugging errors respectively reported by the X virtual power plant resource coordination systems.

[0026] Among them, the target algorithm debugging error reported by the target virtual power plant resource coordination system is generated by the edge decision tree branches pre-matched by the target virtual power plant resource coordination system and the received transmission state change knowledge vector. Each edge decision tree branch pre-matched by the virtual power plant resource coordination system is obtained after optimizing the branch network weights of the multivariate decision tree branches.

[0027] After the virtual power plant resource coordination system receives the transmission state change knowledge vector, it will use the pre-matched edge decision tree branch to process the transmission state change knowledge vector. Each virtual power plant resource coordination system has a corresponding edge decision tree branch, which is obtained by optimizing the branch network weights of the multivariate decision tree branches in the joint data analysis algorithm. By processing the transmission state change knowledge vector, the edge decision tree branch can generate the target algorithm debugging error, which reflects the gap between the algorithm performance of the current virtual power plant resource coordination system and the ideal performance.

[0028] When the distributed lightweight algorithms and edge decision tree branches in the X virtual power plant resource coordination systems meet the edge-side debugging requirements, they will report the target algorithm debugging errors to the power transmission data analysis middle-end system. The power transmission data analysis middle-end system uses these target algorithm debugging errors to further optimize the branch network weights of the AI ​​knowledge mining branches.

[0029] Step 208: Optimize the branch network weights of the AI ​​knowledge mining branches through X target algorithm debugging errors to obtain a cross knowledge mining module, and generate a cross decision tree network corresponding to each virtual power plant resource coordination system based on the cross knowledge mining module and the distributed lightweight algorithms and edge decision tree branches in the X virtual power plant resource coordination systems.

[0030] Among them, the cross decision tree network linked and implanted in the power transmission data analysis middle-end system and the target virtual power plant resource coordination system is used to perform power load optimization scheduling analysis on the intelligent sensor collection information in the target virtual power plant resource coordination system.

[0031] After optimizing the branch network weights of AI knowledge mining branches through X target algorithm debugging errors, the power transmission data analysis middle platform system obtained a cross knowledge mining module. The cross knowledge mining module integrates the knowledge of power transmission state changes of different virtual power plants and improves the overall analysis and optimization capabilities.

[0032] Based on the cross knowledge mining module and the distributed lightweight algorithms and edge decision tree branches in the X virtual power plant resource coordination systems, the power transmission data analysis platform system generates a cross decision tree network corresponding to each virtual power plant resource coordination system. The cross decision tree network can perform real-time power load optimization and dispatch analysis on the intelligent sensor collection information in the virtual power plant, and perform refined management and optimization of the power grid based on the analysis results.

[0033] For example, the cross decision tree network can calculate the key indicators of the power grid such as transmission load, voltage stability, current fluctuation, etc. in real time based on the information collected by intelligent sensors, and optimize the power grid according to these indicators. For example, when the transmission load in a certain area suddenly increases, the cross decision tree network can quickly adjust the power grid operation parameters in that area to ensure the stability and security of the power grid.

[0034] In addition, the cross decision tree network can also predict the future operation status of the power grid and optimize the grid dispatch in advance based on the prediction results. For example, when it is predicted that the power grid in a certain area will experience a load peak in the next few hours, the cross decision tree network can adjust the power grid operation parameters in the area in advance to avoid overload or collapse of the power grid.

[0035] It can be seen that by applying steps 202 to 208, by integrating advanced data analysis algorithms and intelligent sensing technology, real-time monitoring and optimized dispatching of power grid transmission status are realized. It not only improves the operation efficiency and stability of the power grid, but also provides a powerful decision-making support tool for power grid companies.

[0036] In the embodiment of the present invention, the above steps 202 to 208 involve a joint data analysis algorithm, an AI knowledge mining branch, a deep residual branch, a multivariate decision tree branch, a distributed lightweight algorithm, an edge decision tree branch, a cross knowledge mining module, and a cross decision tree network. These algorithms / branches / modules / networks are introduced and explained below.

[0037] (1) The joint data analysis algorithm is a comprehensive data analysis framework that integrates multiple technologies such as AI knowledge mining branch, deep residual branch and multivariate decision tree branch to comprehensively analyze the sensor data of the virtual power plant. Unlike other single algorithms, the advantage of the joint data analysis algorithm lies in its comprehensiveness and integration. It can use multiple technologies to mine data, extract features and make decisions at the same time, thereby providing more comprehensive and accurate analysis results.

[0038] (2) The AI ​​knowledge mining branch focuses on mining implicit knowledge and patterns from the sensor data of the virtual power plant, such as the changing patterns of transmission load and the factors affecting voltage stability. Unlike the deep residual branch and the multivariate decision tree branch, the AI ​​knowledge mining branch focuses more on knowledge mining and discovery rather than model optimization or decision making. It provides valuable information for the optimal dispatch of the power grid by analyzing the implicit patterns in the data.

[0039] (3) Deep residual branching is a deep learning model that improves training efficiency and accuracy by introducing residual connections. In the virtual power plant application scenario, it is used to extract features from sensor data, such as extracting key features such as voltage and current. Unlike AI knowledge mining branches and multivariate decision tree branches, deep residual branching is a deep learning technology that focuses more on model training and feature extraction. It processes sensor data through deep learning algorithms and extracts feature information that is valuable for power grid optimization and dispatching.

[0040] (4) The multivariate decision tree branch is a decision tree model used to make multi-classification decisions on the transmission status of the virtual power plant, such as determining whether the power grid is in a high load, low voltage, etc. Different from the AI ​​knowledge mining branch and the deep residual branch, the multivariate decision tree branch is a decision tree-based model that focuses more on classification and prediction tasks. It classifies the transmission status by constructing a decision tree and provides decision support for the optimal dispatch of the power grid.

[0041] (5) Distributed lightweight algorithm is a lightweight algorithm designed for distributed systems. It is used to efficiently process the sensor data of virtual power plants in resource-constrained environments. Unlike other algorithms, distributed lightweight algorithm focuses more on the lightweight and distributed characteristics of the algorithm. It optimizes the algorithm structure and computational complexity so that the algorithm can run efficiently on resource-constrained edge devices and realize real-time processing of sensor data.

[0042] (6) The edge decision tree branch is a decision tree module deployed on the edge device of the virtual power plant, which is used to classify and predict sensor data in real time, such as judging the real-time operating status of the power grid. Unlike the multivariate decision tree branch, the edge decision tree branch focuses more on real-time data processing and decision making on the edge device. It optimizes the decision tree structure and computing efficiency so that the decision tree can run in real time on the edge device, providing a fast response for the optimal scheduling of the power grid.

[0043] (7) The cross-knowledge mining module integrates the knowledge of transmission state changes of different virtual power plants for cross-analysis and mining to discover the correlations and patterns between different power plants. Unlike other modules, the cross-knowledge mining module focuses more on cross-analysis and knowledge mining between different data sources. By integrating the knowledge of transmission state changes of multiple virtual power plants, it improves the overall analysis and optimization capabilities and provides more comprehensive support for the optimal dispatch of the power grid.

[0044] (8) The cross decision tree network is a decision tree network generated based on the cross knowledge mining module, distributed lightweight algorithm and edge decision tree branch, which is used to optimize the power load scheduling analysis of the virtual power plant in real time. Unlike other networks, the cross decision tree network pays more attention to real-time and optimized scheduling tasks. It realizes the refined management and optimization of the power grid by integrating multiple technologies and modules such as cross knowledge mining module, distributed lightweight algorithm and edge decision tree branch. At the same time, it is also the only component directly applied to the optimization scheduling of the power grid, which can make rapid decisions and adjustments based on real-time sensor data.

[0045] In an embodiment of the present invention, the transmission state change knowledge vector is one of the core invention points. The transmission state change knowledge vector is an important output obtained by the transmission quantity data analysis middle-end system when processing the sensor data of the virtual power plant. It contains rich information about the transmission state of the power grid.

[0046] First, the knowledge vector of power transmission state change is obtained by deeply analyzing the initial distributed sensor metering information through the AI ​​knowledge mining branch in the joint data analysis algorithm. This vector is not a simple set of values, but contains multiple dimensions, each of which corresponds to a key indicator of the power grid transmission state and is represented by specific numerical features.

[0047] For example, one dimension of the transmission state change knowledge vector may be the transmission load, and its numerical feature may be a real number between 0 and 100, representing the current transmission load level of the power grid. For example, a value of 45 may mean that the power grid is currently transmitting at 45% of its capacity. Another dimension may be voltage stability, and its numerical feature may be a real number between -1 and 1, which is used to measure the voltage fluctuation. Values ​​close to 0 indicate stable voltage, while values ​​close to -1 or 1 indicate large voltage fluctuations.

[0048] In addition to transmission load and voltage stability, the transmission state change knowledge vector may also contain other key indicators, such as current fluctuation, transmission efficiency, etc. Each indicator is reflected in the vector with its unique numerical characteristics, and together they constitute a comprehensive description of the power grid transmission state.

[0049] Secondly, the numerical characteristics of the knowledge vector of power transmission status change are not static. They change in real time with the operation of the power grid and the continuous updating of sensor data. This real-time feature enables the power grid company to grasp the latest power transmission status of the power grid in a timely manner and perform corresponding dispatch optimization.

[0050] Then, it is important to emphasize that the numerical characteristics of the knowledge vector of transmission state change have actual physical meanings. They are not just a bunch of numbers, but are closely related to the actual operation of the power grid. By analyzing and interpreting these numerical characteristics, power grid companies can gain a deeper understanding of the transmission state of the power grid and make more scientific and reasonable dispatching decisions.

[0051] Finally, the application of transmission state change knowledge vectors is extensive. They can not only be used for real-time monitoring and optimal dispatching of power grids, but also serve as an important basis for decision support tools of power grid companies. By analyzing and mining historical transmission state change knowledge vectors, power grid companies can discover the laws and trends of power grid operation and provide strong data support for future power grid planning and construction.

[0052] For example, the following is an example of an exemplary feature vector of a power transmission state change knowledge vector, which is used to further illustrate the composition and application of the power transmission state change knowledge vector:

[0053] (1) Example 1 of knowledge vector of power transmission state change

[0054] Dimension 1: transmission load; numerical feature: 65; description: indicates that the power grid is currently transmitting power at 65% of its capacity.

[0055] Dimension 2: Voltage stability; Numerical characteristics: 0.1; Description: Indicates that the voltage fluctuation is small and the power grid operation is stable.

[0056] Dimension 3: current fluctuation; numerical feature: 0.05; description: indicates that the current fluctuation is small and the power grid transmission is stable.

[0057] Dimension 4: Transmission efficiency; Numerical feature: 92; Description: Indicates that the power grid has high transmission efficiency and low energy loss.

[0058] Dimension 5: Equipment health status; Numerical characteristic: 85; Description: Indicates that the overall health of the power grid equipment is good, but some maintenance may be required.

[0059] (2) Example 2 of knowledge vector of power transmission state change (at peak load of power grid)

[0060] Dimension 1: transmission load; numerical feature: 90; description: indicates that the power grid is currently transmitting electricity at a capacity close to full load.

[0061] Dimension 2: Voltage stability; Numerical characteristics: -0.5; Description: Indicates that the voltage fluctuates to a certain extent, and the grid parameters may need to be adjusted to maintain stability.

[0062] Dimension 3: Current fluctuation; Numerical characteristic: 0.2; Explanation: Indicates that the current fluctuation is large and the power grid transmission is affected to a certain extent.

[0063] Dimension 4: Transmission efficiency; Numerical characteristics: 85; Description: Indicates that the power grid transmission efficiency has decreased and energy loss has increased.

[0064] Dimension 5: Equipment health status; Numerical feature: 75; Description: Indicates that the health status of power grid equipment has declined after long-term high-load operation.

[0065] The above eigenvector examples demonstrate the multidimensionality and real-time nature of the knowledge vector of power transmission status change. By analyzing these vectors, power grid companies can fully understand the power grid's power transmission status and perform scheduling optimization in a timely manner. For example, when the power grid load is at its peak, the power grid company can adjust power grid parameters, optimize power transmission efficiency, arrange equipment maintenance, etc. based on the prompts of the knowledge vector of power transmission status change to ensure the stable operation of the power grid.

[0066] It can be seen that the transmission state change knowledge vector is an important output obtained by the transmission data analysis platform system after processing the virtual power plant sensor data. It comprehensively describes the transmission state of the power grid in the form of a specific numerical feature vector, and has the characteristics of real-time, physical meaning and wide applicability. The power grid company can use the transmission state change knowledge vector to grasp the operation status of the power grid in real time and make scientific and reasonable dispatching decisions to improve the operation efficiency and stability of the power grid.

[0067] In summary, the embodiment of the present invention conducts in-depth analysis of the initial distributed sensor metering information by combining the AI ​​knowledge mining branch in the data analysis algorithm, and innovatively obtains the knowledge vector of power transmission state change, which comprehensively and in real time reflects the power transmission state of the power grid. By feeding back the knowledge vector of power transmission state change to the corresponding virtual power plant resource coordination system, and combining the system's built-in distributed lightweight algorithm and edge decision tree branch, accurate monitoring and optimization of the power transmission state of the power grid is achieved.

[0068] In addition, the embodiment of the present invention also collects the algorithm debugging errors of each virtual power plant resource coordination system, further optimizes the weight of the AI ​​knowledge mining branch, generates a cross knowledge mining module and a cross decision tree network, and thus realizes the optimization and dispatch analysis of the power load of the intelligent sensor collection information. In this way, it not only improves the monitoring accuracy and dispatch efficiency of the power transmission status of the power grid, but also provides strong data support for the decision-making of the power grid company, and promotes the further development of the smart grid.

[0069] In some other possible embodiments, the method further includes: obtaining energy relationship heat maps corresponding to the X virtual power plant resource coordination systems, obtaining power cluster operation information having relationship heat commonality with the energy relationship heat map, and learning and debugging the long short-term memory network according to the power cluster operation information to obtain a joint data analysis algorithm that is in a stable state in multiple feature dimensions; sending the deep residual branch and the multivariate decision tree branch of the joint data analysis algorithm to the X virtual power plant resource coordination systems, so that each virtual power plant resource coordination system optimizes the branch network weight of the pre-matched deep residual branch based on the associated intelligent sensor collection information and the AI ​​knowledge mining branch implanted in the power transmission data analysis middle-end system to obtain a distributed lightweight algorithm, and optimizes the branch network weight of the pre-matched multivariate decision tree branch to obtain an edge decision tree branch.

[0070] In this embodiment, the embodiment of the present invention is further expanded to improve the management and optimization capabilities of the virtual power plant resource coordination system through more refined data processing and analysis methods. Specifically, this embodiment first involves obtaining energy relationship heat maps corresponding to X virtual power plant resource coordination systems. The energy relationship heat map is a visualization tool that intuitively displays the distribution, flow and association status of energy, which can help understand the energy interaction patterns and intensity between different virtual power plants. In this step, advanced data acquisition and processing technology is used to extract key energy interaction data from each virtual power plant resource coordination system, and these data are converted into intuitive heat map form through specific algorithms.

[0071] Next, the next step of the embodiment is to obtain the power cluster operation information that has relational thermal commonalities with the energy relationship thermodynamic diagram. Relational thermal commonalities refer to certain similar energy interaction patterns or characteristics that are commonly exhibited in the energy relationship thermodynamic diagrams of different virtual power plants. These commonalities may reflect some general laws or trends in the operation of the power cluster. In order to identify these commonalities, it is necessary to conduct in-depth data mining and analysis on a large number of energy relationship thermodynamic diagrams, extract the key features therein, and match and verify them with the actual operation information of the power cluster.

[0072] After obtaining the power cluster operation information with relational thermal commonalities, the embodiment further learns and debugs the long short-term memory network (LSTM) based on this information. The long short-term memory network is a special type of recurrent neural network (RNN), which can predict future data trends by learning long-term dependencies in sequence data. In this step, the operation information of the power cluster is used as the input data of the LSTM network, and through continuous training and debugging, the network can learn the inherent laws and characteristics of the operation of the power cluster. This process not only involves the optimization of the LSTM network structure, but also includes the fine adjustment of hyperparameters (such as learning rate, number of hidden layer units, etc.) to ensure that the network can reach a stable state in multiple feature dimensions.

[0073] Once the LSTM network has completed learning and reached a stable state, it can be combined with the joint data analysis algorithm. The joint data analysis algorithm is a core component in the embodiment of the present invention. It integrates a variety of advanced data analysis technologies and methods for deep mining and fusion of data from different sources. In this embodiment, the LSTM network that has been learned and debugged is used as an important part of the joint data analysis algorithm, so that it can fully utilize the advantages of the LSTM network in time series data analysis. At the same time, other parts of the joint data analysis algorithm are also optimized and adjusted accordingly to ensure the stability and efficiency of the algorithm in overall performance.

[0074] Next, the embodiment sends the optimized deep residual branch and multivariate decision tree branch of the joint data analysis algorithm to X virtual power plant resource coordination systems. The deep residual branch is a neural network structure used to process complex nonlinear problems. It can effectively extract feature information from the data by learning the residual relationship between input data and output data. The multivariate decision tree branch is an integrated learning method based on decision trees. It can improve the overall prediction performance and robustness by building multiple decision tree models and integrating their prediction results. In this step, these two branches are used as important components of the joint data analysis algorithm and are implanted into each virtual power plant resource coordination system.

[0075] In each virtual power plant resource coordination system, these two branches will be combined with the associated intelligent sensor collection information and the AI ​​knowledge mining branch implanted in the power transmission data analysis middle platform system. Intelligent sensor collection information is real-time data during the operation of the virtual power plant, which reflects the current operating status and performance of the power plant. The AI ​​knowledge mining branch is an intelligent algorithm for mining valuable knowledge and information from large amounts of data. By combining these two branches with intelligent sensor collection information and AI knowledge mining branches, each virtual power plant resource coordination system can be managed and optimized in a more refined manner.

[0076] For example, each virtual power plant resource coordination system will optimize the branch network weights of the pre-matched deep residual branches based on the associated intelligent sensor acquisition information and the output results of the AI ​​knowledge mining branch. This process involves adjusting and optimizing the network parameters in the deep residual branches so that they can better adapt to the current operating status and data characteristics of the virtual power plant. Through this optimization process, a distributed lightweight algorithm can be obtained, which can reduce computational complexity and storage requirements while maintaining high prediction performance, making it more suitable for deployment and application in virtual power plant resource coordination systems with limited resources.

[0077] At the same time, each virtual power plant resource coordination system will also optimize the branch network weights of the pre-matched multivariate decision tree branches. This process is similar to the optimization process of the deep residual branch, and both involve the adjustment and optimization of network parameters. However, due to the differences in structure and principle between the multivariate decision tree branch and the deep residual branch, different optimization strategies and methods are required to optimize them. Through this optimization process, an edge decision tree branch can be obtained, which can reduce the demand for computing resources and storage resources while maintaining high prediction performance, making it more suitable for deployment and application in an edge computing environment.

[0078] In summary, this embodiment achieves more refined management and optimization of the virtual power plant resource coordination system through a series of steps such as obtaining energy relationship heat maps, extracting relationship heat commonalities, learning and debugging LSTM networks, optimizing joint data analysis algorithms, and issuing and optimizing deep residual branches and multivariate decision tree branches. This embodiment can not only improve the operating efficiency and performance of virtual power plants, but also reduce the demand for computing resources and storage resources, thereby providing strong technical support for the sustainable development of smart grids.

[0079] In addition, in practical applications, the technical solutions can be adjusted and optimized accordingly according to specific business needs and scenario characteristics. For example, the appropriate structure and parameters of the deep residual branch and multivariate decision tree branch can be selected according to the scale and complexity of the virtual power plant; the LSTM network can also be studied and debugged more deeply according to the operating characteristics and data characteristics of the power cluster to improve its prediction performance and generalization ability. Through these flexible adjustments and optimizations, the technical solutions can be better adapted to different application scenarios and business needs, thereby providing strong support for the personalized customization and continuous optimization of smart grids.

[0080] At the same time, under the current development trend of the smart grid field, the management and optimization of the virtual power plant resource coordination system has become an important research direction. However, due to the complexity and diversity of virtual power plants, as well as the uncertainty and dynamics in the operation of power clusters, this task faces many challenges and difficulties. The embodiments of the present invention provide new ideas and methods for solving this problem by introducing advanced machine learning algorithms and data processing technologies.

[0081] It is worth mentioning that when discussing the optimization and management of smart grid and virtual power plant resource coordination system, the energy relationship heat map is an innovative and profound concept involved in the embodiments of the present invention. The energy relationship heat map not only provides a new perspective to examine and understand the flow and interaction of energy between virtual power plants, but also reveals the valuable information and rules hidden behind these complex data.

[0082] First of all, the core value of the energy relationship heat map lies in its visualization characteristics. In traditional energy management and power system analysis, large amounts of data and complex models are often relied on to try to understand the flow and distribution of energy. However, these data can be abstract and difficult to understand intuitively, which makes it very challenging for decision makers to formulate strategies and optimize systems. The emergence of energy relationship heat maps can transform abstract energy data into intuitive and easy-to-understand graphical representations, allowing decision makers to see at a glance the energy interaction patterns and intensities between different virtual power plants. This visualization feature not only improves the readability of the data, but also greatly enhances the cognition and understanding of energy flow and distribution.

[0083] Secondly, energy relationship heat maps play an important role in revealing energy interaction patterns and laws. Through in-depth data mining and analysis, key interaction patterns and features can be extracted from a large amount of energy data, and this information can be presented in the form of heat maps. These patterns and features may reflect the energy supply and demand relationship between virtual power plants, the path and intensity of energy flow, and the stability and dynamics of energy interaction. By observing and analyzing these heat maps, we can have a deeper understanding of the energy interaction behavior between virtual power plants, reveal the laws and trends therein, and provide strong support for optimizing energy management and improving the efficiency of power systems.

[0084] Then, the energy relationship heat map also has strong flexibility and scalability. In practical applications, the heat map can be customized and optimized according to specific business needs and scenario characteristics. For example, according to the scale and complexity of the virtual power plant, appropriate color mapping and graphical representation methods can be selected to more intuitively display the patterns and characteristics of energy interactions. Other advanced data analysis technologies, such as machine learning algorithms and time series data analysis methods, can also be combined to further mine and analyze the energy relationship heat map to reveal more valuable information and laws. This flexibility and scalability make the energy relationship heat map widely applicable and practical in practical applications.

[0085] In addition, the generation process of the energy relationship heat map is also a process full of innovation and challenges. In order to generate an accurate and reliable energy relationship heat map, it is necessary to first collect and process a large amount of energy data, including the power generation, power consumption, energy flow path and intensity of the virtual power plant. These data can come from different data sources and acquisition systems, with different formats and qualities. Therefore, the data needs to be cleaned, integrated, and converted to ensure the accuracy and consistency of the data. Next, it is necessary to use advanced data mining and analysis techniques to extract key energy interaction patterns and features from these data and convert them into the form of heat maps. This process requires an in-depth understanding and mastery of data mining and analysis techniques, as well as rich practical experience and innovative thinking.

[0086] In the application process of energy relationship heat map, other advanced technologies and methods can be combined for further optimization and management. For example, machine learning algorithms can be used to automatically identify and classify energy relationship heat maps to quickly locate and analyze energy interaction patterns with specific characteristics. It is also possible to combine virtual reality and augmented reality technologies to present energy relationship heat maps to users in a more intuitive and vivid way to improve users' cognition and understanding of energy interaction behaviors. The application of these advanced technologies and methods will further expand the application areas and value of energy relationship heat maps.

[0087] It can be seen that the energy relationship heat map, as an innovative visualization tool, plays an important role in the optimization and management of smart grid and virtual power plant resource coordination systems. It not only improves the cognition and understanding of energy flow and distribution, but also reveals the valuable information and laws hidden behind these complex data. Through in-depth research and application of energy relationship heat maps, it is possible to more effectively manage and optimize the virtual power plant resource coordination system, improve the efficiency and stability of the power system, and provide strong support for the sustainable development of smart grids.

[0088] In some other preferred embodiments, the obtaining of X initial distributed sensor metering information includes: obtaining an edge monitoring cycle, selecting X virtual power plant resource coordination systems from Y virtual power plant resource coordination systems according to the edge monitoring cycle, and sending a learning and debugging task to the X virtual power plant resource coordination systems, so that each virtual power plant resource coordination system performs sensor metering identification on the intelligent sensor collection information through a distributed lightweight algorithm to obtain initial distributed sensor metering information; Y is a positive integer greater than X; and obtaining the initial distributed sensor metering information reported by the X virtual power plant resource coordination systems respectively.

[0089] It can be understood that this embodiment takes into account the actual operating environment and needs in more depth, especially in the process of obtaining distributed sensor metering information. First of all, it is necessary to clarify that the goal is to obtain X initial distributed sensor metering information, which is a key step to provide basic data for subsequent data analysis and system optimization. In order to achieve this goal, the solution starts with the edge monitoring cycle, which sets the rhythm of data collection and processing, ensuring the timeliness and effectiveness of the data.

[0090] According to the preset edge monitoring cycle, Y virtual power plant resource coordination systems are screened. Since Y is a positive integer greater than X, this means that there are multiple candidate systems, and X need to be selected as target systems. The selection process may be based on a variety of factors, such as system stability, data quality, historical performance, etc., to ensure that the final selected system can provide accurate and reliable sensor measurement information.

[0091] Once X virtual power plant resource coordination systems are selected, the next step is to send learning and debugging tasks to these systems. The core of this task is to enable each virtual power plant resource coordination system to perform sensor metering identification on the intelligent sensor collection information through its built-in distributed lightweight algorithm. The design of the distributed lightweight algorithm takes into account computing efficiency and storage requirements, enabling it to operate effectively in resource-limited edge environments. Through this algorithm, the system can process sensor data in real time and extract key metering information, providing a basis for subsequent data analysis and decision-making.

[0092] In the process of learning and debugging tasks, each virtual power plant resource coordination system will perform independent calculations and analyses based on the intelligent sensor information it receives. These intelligent sensor information may include various power parameters such as current, voltage, and power, which reflect the operating status of the power plant in real time. Through the processing of distributed lightweight algorithms, this information is converted into initial distributed sensor metering information, which is a more structured and easy-to-analyze data form.

[0093] After completing the learning and debugging tasks, the X virtual power plant resource coordination systems will report their initial distributed sensor metering information. This process may involve data encryption, compression, and transmission to ensure data security and transmission efficiency. The reported data is then aggregated and analyzed to extract valuable insights and trends to guide subsequent power resource coordination and optimization strategies.

[0094] In this embodiment, through the designed steps and algorithms, the effective acquisition of X initial distributed sensor metering information is achieved. This process not only takes into account the accuracy and timeliness of the data, but also fully considers the complexity and resource constraints of the actual operating environment. First, by setting the edge monitoring cycle, it is ensured that the rhythm of data collection matches the actual application needs. Secondly, by selecting X target systems from Y candidate systems, the quality and reliability of the data are further improved. Then, by sending learning and debugging tasks and using distributed lightweight algorithms for sensor metering identification, the effective processing and analysis of intelligent sensor acquisition information is achieved. Finally, by summarizing and analyzing the reported initial distributed sensor metering information, strong data support is provided for subsequent power resource coordination and optimization strategies.

[0095] In addition, in actual applications, parameters such as edge monitoring cycle, number of selected systems, and specific content of learning and debugging tasks can be flexibly adjusted according to specific business needs and environmental changes to ensure that the technical solution can always meet the needs of actual applications. At the same time, with the continuous development of technology and the emergence of new algorithms, distributed lightweight algorithms can also be continuously optimized and upgraded to improve their processing efficiency and analysis accuracy, further enhancing the performance and value of the entire technical solution.

[0096] Therefore, through the designed steps and algorithms, the effective acquisition and processing of X initial distributed sensor metering information is achieved. It not only takes into account the accuracy and timeliness of the data, but also fully considers the complexity and resource limitations of the actual operating environment. Through the application of this embodiment, the operating status and performance of the virtual power plant resource coordination system can be more accurately understood and mastered, providing strong data support for subsequent power resource coordination and optimization strategies.

[0097] In some other possible embodiments, the performing state change knowledge mining on the X initial distributed sensor metering information by the AI ​​knowledge mining branch in the joint data analysis algorithm to obtain the power transmission state change knowledge vectors corresponding to the X initial distributed sensor metering information respectively includes: performing cross-feature downsampling processing on the X initial distributed sensor metering information to obtain cross-downsampling features corresponding to the X initial distributed sensor metering information respectively; performing state mapping operation on the X cross-downsampling features to obtain state mapping intermediate features corresponding to the X cross-downsampling features respectively, and generating the power transmission state change knowledge vectors corresponding to the X initial distributed sensor metering information respectively based on the X state mapping intermediate features.

[0098] This embodiment focuses on deeply mining the knowledge of power transmission state changes contained in X initial distributed sensor metering information by combining the AI ​​knowledge mining branch in the data analysis algorithm. This process aims to extract valuable information from massive sensor data and provide a solid foundation for subsequent system optimization and decision support.

[0099] First, the X initial distributed sensor metering information needs to be cross-feature down-sampling. This step is a key link in data processing. It aims to reduce the dimension of the data by performing specific sampling and feature extraction on the original data, while retaining the most critical information for subsequent analysis. The core of the cross-feature down-sampling process lies in the word "cross", which means that in the sampling process, not only the independent features of a single sensor metering information are focused on, but also the cross-features between different sensor metering information are emphasized. For example, in the power system, current and voltage are two basic sensor metering information, each of which has independent characteristics. However, in the cross-feature down-sampling process, more attention is paid to the cross-features between current and voltage, such as their phase relationship, the synchronization of change trends, etc. These cross-features often contain important information about the operating status of the power system, which is of great significance for the subsequent knowledge mining of transmission state changes.

[0100] After cross-feature downsampling, the cross-downsampled features corresponding to the X initial distributed sensor metering information are obtained. These features are refined representations of the original sensor metering information, which retain the most critical information for describing the changes in the transmission state while removing redundancy and noise.

[0101] Next, it is necessary to perform a state mapping operation on these X cross-downsampled features. The purpose of this step is to transform the cross-downsampled features into a more abstract form that is easier to analyze and understand, namely, the state mapping intermediate features. The state mapping operation can be regarded as a feature conversion process, which transforms the original sensor metering data into a feature representation that can reflect the operating state of the transmission system. For example, in the power system, the cross-downsampled features of current and voltage can be transformed into a feature representing the load state of the power system through the state mapping operation. This feature can more intuitively reflect the operating state of the power system and facilitate subsequent knowledge mining.

[0102] Based on X state mapping intermediate features, the transmission state change knowledge vectors corresponding to X initial distributed sensor metering information can be further generated. This step is the core link of knowledge mining. It extracts the key knowledge of transmission state change by deeply analyzing and mining the intermediate features of state mapping. The transmission state change knowledge vector is a highly abstract knowledge representation form, which contains important information such as the changing laws, trends and possible abnormal conditions of the operating state of the transmission system. For example, in the power system, the transmission state change knowledge vector can represent the stability of the power system, the load change trend and possible fault warning information. This information has important guiding significance for the operation and maintenance of the power system.

[0103] To illustrate the process more specifically, a practical example can be given. In a virtual power plant resource coordination system, two initial distributed sensor metering information, current and voltage, are obtained. Through cross-feature downsampling processing, cross-downsampling features such as the phase relationship and synchronization of the change trend between them are extracted. Then, these cross-downsampling features are converted into a feature representing the load state of the power system through state mapping operations. Finally, based on this load state feature, a transmission state change knowledge vector is generated, which contains information such as the stability of the power system, load change trends, and possible fault warnings. This information can help to understand the operating status of the power system more accurately and provide strong support for subsequent decision-making and optimization.

[0104] In this way, the knowledge of power transmission state changes contained in the X initial distributed sensor metering information is deeply mined through the AI ​​knowledge mining branch in the joint data analysis algorithm. This process includes key steps such as cross-feature downsampling of the original data, state mapping of the cross-downsampling features, and generation of power transmission state change knowledge vectors based on the intermediate features of the state mapping. Through the implementation of these steps, valuable information can be extracted from massive sensor data, providing a solid foundation for subsequent system optimization and decision support.

[0105] It is worth mentioning that cross-sampling features are a technical means to process multi-dimensional feature vectors. In the power system, when faced with multiple linear numerical feature vectors (such as current, voltage, power, etc.), cross-sampling feature extraction aims to combine these features in some way to discover the potential relationship or pattern between them. This combination is not a simple linear superposition, but a representative cross-feature is extracted from multiple feature vectors through a specific downsampling strategy (such as maximum pooling, average pooling, etc.). These cross-features can more comprehensively reflect the operating status of the power system and provide richer information for subsequent analysis.

[0106] The state mapping intermediate features are the results obtained by further processing based on the cross-downsampling features. After obtaining the cross-downsampling features, these features need to be converted into a new space or representation through a certain mapping relationship for better understanding and analysis. This mapping relationship can be linear or nonlinear, depending on the complexity of the problem and the analysis requirements. In the power system, the state mapping intermediate features may represent the key parameters or indicators of the power system in different states, such as load rate, stability index, etc. These intermediate features provide important input for the generation of the knowledge vector of transmission state change.

[0107] The knowledge vector of transmission state change is one of the ultimate goals of data analysis. It generates a knowledge representation that can fully reflect the change of transmission state by further processing and integrating the intermediate features of state mapping. This knowledge vector may contain multiple dimensions, each of which corresponds to a specific aspect or attribute of the transmission state. For example, in the power system, the knowledge vector of transmission state change may include multiple dimensions such as the stability state of the power system, load change trend, fault warning signal, etc. These dimensions together constitute a comprehensive description of the change of transmission state, providing important decision support for the operation and maintenance of the power system.

[0108] Therefore, cross-sampling features, state mapping intermediate features and transmission state change knowledge vectors are three key stages in the data processing and analysis process. They use linear numerical feature vectors as a starting point, gradually dig deep into the potential information and laws in the data, and finally generate a comprehensive knowledge representation of transmission state changes.

[0109] In some other possible embodiments, the optimizing the branch network weights of the AI ​​knowledge mining branches through X target algorithm debugging errors to obtain a cross knowledge mining module includes: generating local algorithm debugging errors corresponding to the X target algorithm debugging errors respectively through X target algorithm debugging errors and the branch network weights in the AI ​​knowledge mining branches, generating loss change partial derivative variables based on the X local algorithm debugging errors and preset debugging hyperparameters, optimizing the branch network weights of the AI ​​knowledge mining branches according to the loss change partial derivative variables, and obtaining a cross knowledge mining module.

[0110] It can be understood that the focus of this embodiment is on how to optimize the branch network weights of the AI ​​knowledge mining branch through X target algorithm debugging errors, thereby obtaining a refined and efficient cross-knowledge mining module. This process not only involves complex algorithm debugging, but also covers deep-level strategies for weight optimization, aiming to improve the effect and accuracy of AI knowledge mining from multiple dimensions.

[0111] Through X target algorithm debugging errors and the branch network weights in the AI ​​knowledge mining branch, the local algorithm debugging errors corresponding to the X target algorithm debugging errors are generated. This step is the key preparation stage for weight optimization. The local algorithm debugging error is a further refinement of the original debugging error, which takes into account the impact of the branch network weight on the debugging error. In other words, try to find out the part of the error caused by the branch network weight in each target algorithm debugging error. In this way, the weight can be optimized more specifically, thereby reducing the debugging error.

[0112] After generating X local algorithm debugging errors, a loss change partial derivative variable is generated based on these local errors and the preset debugging hyperparameters. The loss change partial derivative variable is a key intermediate quantity that reflects the trend of the debugging error as the branch network weight changes. By calculating the loss change partial derivative variable, we can know how to adjust the branch network weights to reduce the debugging error. The preset debugging hyperparameters play an important guiding role in this process, and they determine the direction and step size of weight optimization.

[0113] With the loss change partial derivative variable, the branch network weight of the AI ​​knowledge mining branch can be optimized based on this variable. The optimization process is an iterative process, which continuously adjusts the branch network weights and then recalculates the debugging error and loss change partial derivative variable until a preset optimization goal or upper limit of the number of iterations is reached. Through this series of optimization steps, the debugging error can be gradually reduced and the performance of the AI ​​knowledge mining branch can be improved.

[0114] Finally, after optimizing the weights of the branch networks, a refined and efficient cross-knowledge mining module was obtained. This module not only integrates the knowledge of multiple target algorithms, but also improves its performance on specific tasks through weight optimization. In practical applications, the cross-knowledge mining module can exert powerful data analysis capabilities and mine valuable knowledge of power transmission state changes from massive sensor data.

[0115] For example, in a smart grid system of an electric power company, an AI knowledge mining branch was deployed to monitor and analyze the operation status of the power grid. Three target algorithms were selected: Algorithm A, Algorithm B, and Algorithm C, which are responsible for monitoring the voltage stability, current fluctuation, and load balance of the power grid, respectively. During the training process, it was found that all three algorithms had certain debugging errors. In order to optimize the performance of the AI ​​knowledge mining branch, the local algorithm debugging errors corresponding to the debugging errors of the three target algorithms were first generated. Then, based on these local errors and preset debugging hyperparameters (such as learning rate, number of iterations, etc.), the loss change partial derivative variables were calculated. Next, the branch network weights of the AI ​​knowledge mining branch were optimized based on this variable. After a series of iterations and optimization steps, a refined and efficient cross-knowledge mining module was obtained. This module can more accurately monitor and analyze the operation status of the power grid, providing strong support for the operation and maintenance of the power company.

[0116] In this process, cross-sampling features, state mapping intermediate features, and transmission state change knowledge vectors all play an important role. First, cross-sampling features are semantic representations obtained after specific processing of the original sensor data. These features not only retain the key information of the original data, but also incorporate the correlation information between multiple sensor data through cross-combination. In the example of a smart grid system, cross-sampling features may include cross-relationships between voltage, current, and load, such as their phase difference, synchronization of change trends, etc. These features provide rich input information for subsequent AI knowledge mining.

[0117] Secondly, the state mapping intermediate features are obtained by further processing based on the cross-downsampling features. They transform the cross-downsampling features into a more abstract form that is easier to analyze and understand through a certain mapping relationship. In the example of a smart grid system, the state mapping intermediate features may represent key parameters or indicators of the power grid in different states, such as voltage stability index, current fluctuation index, and load balance index. These intermediate features provide important input for the generation of the knowledge vector of transmission state changes.

[0118] Finally, the knowledge vector of transmission state change is one of the ultimate goals of data analysis. It generates a knowledge representation that can fully reflect the change of transmission state by further processing and integrating the intermediate features of state mapping. In the example of smart grid system, the knowledge vector of transmission state change may include multiple dimensions such as the stability state of the power grid, load change trend, fault warning signal, etc. These dimensions together constitute a comprehensive description of the change of power grid operation state, providing an important basis for the decision-making and support of power companies.

[0119] It can be seen that the process of optimizing the branch network weights of AI knowledge mining branches through X target algorithm debugging errors to obtain a refined and efficient cross knowledge mining module is a complex and meticulous process. It involves multiple steps such as debugging multiple algorithm models, generating local algorithm debugging errors, calculating loss change partial derivative variables, and optimizing branch network weights. In this process, cross-sampling features, state mapping intermediate features, and transmission state change knowledge vectors all play an important role. Together, they constitute the core framework of data analysis and provide strong support for the monitoring and analysis of smart grid systems.

[0120] When we delve into how to generate loss change partial derivative variables based on X local algorithm debugging errors and preset debugging hyperparameters, we first need to understand the core purpose of this process: to minimize the overall loss function by accurately calculating the direction and size of the weight adjustment, thereby improving the performance of the AI ​​knowledge mining branch.

[0121] First, the overall loss function is set to L, which represents the comprehensive debugging error of the AI ​​knowledge mining branch on all target algorithms. The goal is to find a way to minimize L by adjusting the branch network weights. To achieve this goal, it is necessary to calculate the partial derivative of the loss function L with respect to each weight, that is, the loss change partial derivative variable.

[0122] Secondly, considering that there are X local algorithm debugging errors, the overall loss function L can be expressed as the weighted sum of these local debugging errors. Each local debugging error corresponds to a specific target algorithm, and its weight reflects the importance of the algorithm in the overall performance evaluation. Therefore, the overall loss function L can be expressed as: L = w1*E1+w2*E2+...+wX*EX.

[0123] Among them, wx represents the weight of the x-th local debugging error Ex, and the sum of these weights is 1.

[0124] Then, in order to calculate the partial derivative of the loss change, it is necessary to calculate the partial derivative of the overall loss function L with respect to each branch network weight. This calculation process involves the chain rule and the calculation of partial derivatives of multivariate functions. For example, for each weight wk, it is necessary to calculate the partial derivative of L with respect to wk, that is:

[0125] pd_L / pd_wk=pd_L / pd_E1*pd_E1 / pd_wk+pd_L / pd_E2*pd_E2 / pd_wk+...+pd_L / pd_EX*pd_EX / pd_wk.

[0126] Among them, pd_L / pd_Ex represents the partial derivative of the overall loss function L with respect to the x-th local debugging error Ex, and pd_Ex / pd_wk represents the partial derivative of the x-th local debugging error Ex with respect to the weight wk.

[0127] Finally, the results of the above partial derivative calculations are combined with the preset debugging hyperparameters to generate the final loss change partial derivative variables. The preset debugging hyperparameters may include learning rate, momentum term, etc., which are used to control the amplitude and speed of weight adjustment. By multiplying these hyperparameters with partial derivatives or performing other mathematical operations, the direction and size of each weight adjustment can be obtained, thereby optimizing the weights of the branch network.

[0128] Therefore, the process of generating loss change partial derivative variables based on X local algorithm debugging errors and preset debugging hyperparameters is a complex process involving the calculation of multivariate function partial derivatives and the formulation of weight adjustment strategies. By accurately calculating the partial derivatives of the loss function with respect to each weight and combining it with the preset debugging hyperparameters, the performance of the AI ​​knowledge mining branch can be effectively optimized, thereby improving its accuracy and efficiency in specific tasks.

[0129] In an optional technical idea, the cross-knowledge mining module and the distributed lightweight algorithms and edge decision tree branches in the X virtual power plant resource coordination systems are used to generate a cross-decision tree network corresponding to each virtual power plant resource coordination system, including: performing a weighted operation on the branch network weights of the distributed lightweight algorithms and the branch network weights of the edge decision tree branches in the X virtual power plant resource coordination systems to obtain an algorithm optimization indication feature, and sending the algorithm optimization indication feature to the X virtual power plant resource coordination systems, so that each virtual power plant resource coordination system selects a pre-matched distributed lightweight algorithm according to the algorithm optimization indication feature. The neuron weights of the quantity algorithm and the edge decision tree branches are optimized to obtain a lightweight crossover algorithm and a crossover decision branch; when the crossover knowledge mining module meets the debugging indicators of the middle-office system, the crossover knowledge mining module is combined with the lightweight crossover algorithm and the crossover decision branch pre-matched by each virtual power plant resource coordination system to obtain a crossover decision tree network corresponding to each virtual power plant resource coordination system; wherein, the crossover knowledge mining module is combined with the lightweight crossover algorithm and the crossover decision branch pre-matched by the target virtual power plant resource coordination system to obtain a crossover decision tree network corresponding to the target virtual power plant resource coordination system.

[0130] When it comes to how to generate a cross decision tree network corresponding to each virtual power plant resource coordination system based on the cross knowledge mining module and the distributed lightweight algorithm and edge decision tree branches in X virtual power plant resource coordination systems, it is necessary to deeply understand the core purpose and implementation mechanism of this technical idea. This process aims to improve the overall performance and decision efficiency of the virtual power plant resource coordination system by optimizing the weights of the algorithm and decision tree branches. The following is a detailed explanation of the technical solution and its background in the debugging indicators of the middle platform system.

[0131] First, the branch network weights of the distributed lightweight algorithms and the branch network weights of the edge decision tree branches in the X virtual power plant resource coordination systems are weighted. This step is the basis of algorithm optimization. By comprehensively considering the importance and contribution of different algorithms and decision tree branches, a comprehensive algorithm optimization indicator feature can be obtained. This feature reflects the current status and optimization direction of the distributed lightweight algorithms and edge decision tree branches in each virtual power plant resource coordination system.

[0132] Next, this algorithm optimization indication feature is sent to X virtual power plant resource coordination systems. Each system will optimize the neuron weights of the pre-matched distributed lightweight algorithm and edge decision tree branches based on this feature. This process is dynamic and continuous, which ensures that each virtual power plant resource coordination system can adjust the weights of its algorithm and decision tree branches in real time according to the latest optimization instructions to achieve the best coordination effect.

[0133] Through this weight optimization, a lightweight crossover algorithm and crossover decision branches are obtained. These optimized algorithms and decision branches are more suitable for the actual operating environment of virtual power plants, and can effectively reduce computational complexity and resource consumption while ensuring decision accuracy.

[0134] However, this is not the whole technical solution. The debugging indicators of the middle-office system also need to be considered. When the middle-office system meets specific debugging indicators, it means that the cross-knowledge mining module has reached a stable and optimized state. At this time, the cross-knowledge mining module is combined with the lightweight cross-algorithm and cross-decision branch pre-matched by each virtual power plant resource coordination system to obtain the cross-decision tree network corresponding to each virtual power plant resource coordination system.

[0135] The debugging indicators of the middle office system are important criteria for measuring the performance and stability of the entire system. For example, these indicators may include the system's response time, decision accuracy, resource utilization, etc. Only when the middle office system meets the preset standards on these indicators can it be considered stable and optimized, and then combined with the optimization algorithm and decision branch of the virtual power plant resource coordination system.

[0136] Through this combination, a cross decision tree network corresponding to each virtual power plant resource coordination system is obtained. This network combines the deep mining capabilities of the cross knowledge mining module with the actual operation requirements of the virtual power plant resource coordination system, and can achieve more efficient and intelligent resource coordination while ensuring decision accuracy.

[0137] Taking a specific virtual power plant resource coordination system as an example, we can have a deeper understanding of the practical application of this technical solution. The virtual power plant needs to coordinate a variety of distributed energy sources, including wind energy, solar energy and energy storage equipment. In the optimization process, the weights of the distributed lightweight algorithm and the edge decision tree branch are first weighted to obtain the algorithm optimization indication feature. Then, this feature is sent to the virtual power plant resource coordination system to optimize the weights of the pre-matched algorithm and decision branch.

[0138] After a period of optimization, a lightweight crossover algorithm and crossover decision branches were obtained. These optimized algorithms and decision branches are more suitable for the actual operating environment of virtual power plants, and can effectively reduce computational complexity and resource consumption while ensuring the accuracy of energy coordination.

[0139] At this time, if the middle-office system meets the preset debugging indicators, such as shortening the response time to milliseconds, improving decision accuracy to more than 99%, and increasing resource utilization to 80%, then the cross-knowledge mining module can be combined with the lightweight cross-algorithm and cross-decision branch of this virtual power plant resource coordination system.

[0140] Through this combination, the cross decision tree network corresponding to the virtual power plant resource coordination system is obtained. This network not only has the deep mining capability of the cross knowledge mining module, but also fully considers the actual operation requirements and constraints of the virtual power plant. In practical applications, it can coordinate and make energy decisions more accurately and efficiently, providing strong support for the stable operation and sustainable development of virtual power plants.

[0141] Therefore, it is a complex and delicate process to generate a cross decision tree network corresponding to each virtual power plant resource coordination system based on the cross knowledge mining module and the distributed lightweight algorithm and edge decision tree branches in the X virtual power plant resource coordination systems. It involves multiple links such as algorithm optimization, weight adjustment, and middle-end system debugging, and it is necessary to fully consider the actual operating environment and needs of the virtual power plant. Through the application of this technical solution, the resource coordination efficiency and decision-making accuracy of the virtual power plant can be effectively improved, laying a solid foundation for its stable operation and sustainable development.

[0142] In some other exemplary embodiments, the algorithm optimization indication feature includes knowledge mining optimization variables and decision processing optimization variables; the branch network weights of the distributed lightweight algorithms and the branch network weights of the edge decision tree branches in the X virtual power plant resource coordination systems are weighted to obtain the algorithm optimization indication feature, including: obtaining the branch network weights of the distributed lightweight algorithms in the X virtual power plant resource coordination systems, performing weighted operations on the branch network weights of the X distributed lightweight algorithms, and obtaining the knowledge mining optimization variables corresponding to the X virtual power plant resource coordination systems respectively; the knowledge mining optimization variables are used to be sent to the corresponding virtual A power plant resource coordination system is provided, and is used to optimize the neuron weight of the distributed lightweight algorithm corresponding to the corresponding virtual power plant resource coordination system to obtain a lightweight crossover algorithm; the branch network weights of the edge decision tree branches in the X virtual power plant resource coordination systems are obtained, and the branch network weights of the X edge decision tree branches are weighted to obtain the decision processing optimization variables corresponding to the X virtual power plant resource coordination systems respectively; the decision processing optimization variables are used to be sent down to the corresponding virtual power plant resource coordination system, and are used to optimize the neuron weight of the edge decision tree branches corresponding to the corresponding virtual power plant resource coordination system to obtain a crossover decision branch.

[0143] This embodiment involves how to generate a cross decision tree network corresponding to each virtual power plant resource coordination system based on a cross knowledge mining module and a distributed lightweight algorithm and edge decision tree branches in X virtual power plant resource coordination systems. Among them, the algorithm optimization indication feature plays a vital role, which is specifically refined into knowledge mining optimization variables and decision processing optimization variables, which together guide the optimization direction of the virtual power plant resource coordination system.

[0144] First, we focus on the generation of knowledge mining optimization variables. In the X virtual power plant resource coordination systems, each system runs a distributed lightweight algorithm, which reflects its unique knowledge mining capabilities through branch network weights. In order to obtain knowledge mining optimization variables, it is necessary to first obtain the branch network weights of these distributed lightweight algorithms. Subsequently, these weights are weighted. The purpose of this step is to comprehensively consider the contribution and importance of different algorithms in knowledge mining, so as to obtain a comprehensive evaluation index. This index obtained after the weighted operation is the so-called knowledge mining optimization variable, which reflects the optimization potential and direction of the virtual power plant resource coordination system in knowledge mining.

[0145] After obtaining the knowledge mining optimization variable, it will be sent to the corresponding virtual power plant resource coordination system. This variable plays the role of a "baton" here, guiding the system to optimize the neuron weights of its internal distributed lightweight algorithm. Through this optimization, lightweight crossover algorithms can be obtained. These algorithms further reduce the computational complexity and improve the operating efficiency while retaining the original knowledge mining capabilities.

[0146] Next, we turn to the generation of decision processing optimization variables. In the virtual power plant resource coordination system, the edge decision tree branch is responsible for processing the actual decision-making task, and its branch network weight determines the quality and speed of the decision. In order to obtain the decision processing optimization variables, it is necessary to first obtain the branch network weights of these edge decision tree branches. Similar to the generation of knowledge mining optimization variables, these weights also need to be weighted to comprehensively consider the contribution and importance of different decision tree branches in decision processing. After the weighted operation, the decision processing optimization variables are obtained, which reflects the optimization potential and direction of the virtual power plant resource coordination system in decision processing.

[0147] Similarly, the decision processing optimization variable will be sent to the corresponding virtual power plant resource coordination system. This variable plays the role of a "navigator" here, guiding the system to optimize the neuron weights of its internal edge decision tree branches. Through this optimization, cross-decision branches can be obtained, which further improve the accuracy and speed of decision-making while retaining the original decision processing capabilities.

[0148] After completing the generation and distribution of the above-mentioned algorithm optimization indication features, it is also necessary to pay attention to the debugging indicators of the middle-office system. These indicators are important criteria for measuring the performance and stability of the middle-office system, and are also the key basis for judging whether the system meets the training standards. For example, the debugging indicators of the middle-office system may include the response time of the system, the accuracy of the decision, the utilization rate of resources, and the stability of the system. Only when the middle-office system meets the preset standards on these indicators can it be considered stable and optimized, and then combined with the optimization algorithm and decision-making branch of the virtual power plant resource coordination system.

[0149] For example, when the middle-office system meets the debugging indicators, the cross-knowledge mining module will be combined with the lightweight cross-algorithm and cross-decision branches pre-matched by each virtual power plant resource coordination system. This combination process is based on the deep integration of the global optimization capability of the middle-office system and the local optimization capability of the virtual power plant resource coordination system. Through this combination, a cross-decision tree network corresponding to each virtual power plant resource coordination system can be obtained. This network not only has the deep mining capability of the cross-knowledge mining module, but also fully considers the actual operation requirements and constraints of the virtual power plant. In practical applications, it can coordinate resources and make decisions more accurately and efficiently, providing strong support for the stable operation and sustainable development of virtual power plants.

[0150] Take a power company as an example. The power company has multiple virtual power plant resource coordination systems, which are responsible for energy dispatching and management in different areas. In order to improve the performance and efficiency of these systems, the company decided to use the above technical solutions for optimization. First, the company acquired and weighted the branch network weights of the distributed lightweight algorithm and edge decision tree branches in each virtual power plant resource coordination system, and obtained the corresponding knowledge mining optimization variables and decision processing optimization variables. Then, the company sent these two optimization variables to the corresponding virtual power plant resource coordination system to guide it to optimize the neuron weights. After a period of optimization, the company obtained lightweight crossover algorithms and crossover decision branches, and combined them with the crossover knowledge mining module of the middle platform system. Finally, the company obtained the crossover decision tree networks corresponding to each virtual power plant resource coordination system. These networks have shown excellent performance and efficiency in practical applications, bringing significant benefits to the company's energy dispatching and management.

[0151] This design makes full use of the guiding role of knowledge mining optimization variables and decision processing optimization variables, and obtains lightweight crossover algorithms and crossover decision branches through the optimization of neuron weights. At the same time, attention is also paid to the debugging indicators and training compliance conditions of the middle platform system to ensure the stability and reliability of the entire optimization process. Finally, a crossover decision tree network corresponding to each virtual power plant resource coordination system is obtained, which provides strong support for the stable operation and sustainable development of virtual power plants.

[0152] In addition, based on the above content, the method also includes: optimizing the scheduling of power loads in the target power area based on time series according to the power load optimization scheduling viewpoint output by the cross decision tree network that is linked and implanted in the power transmission data analysis middle-office system and the target virtual power plant resource coordination system.

[0153] It can be understood that the above content aims to achieve efficient management and optimized scheduling of power loads through advanced data analysis and decision support systems, thereby improving the stability and operating efficiency of the power system.

[0154] First of all, it should be clear that the transmission data analysis middle platform system, as the basic component of this technical solution, plays a vital role. The system is responsible for collecting, integrating and processing massive power data from different sources, including but not limited to historical power consumption records, real-time power load monitoring data, weather forecast information, and user behavior patterns. These data are the key basis for subsequent power load forecasting and optimal scheduling. Through in-depth analysis of these data, the middle platform system can reveal the law of power load changes and provide data support for the formulation of accurate scheduling strategies. At the same time, the target virtual power plant resource coordination system is another core element for achieving optimal scheduling of power load. As an innovative power management model, the virtual power plant integrates decentralized power resources (such as distributed energy, energy storage devices, controllable loads, etc.) through information technology to form a "virtual" power plant that can be uniformly scheduled and managed. The main task of the resource coordination system is to reasonably configure and schedule these virtual power plant resources based on the current power supply and demand situation, market price signals, and environmental constraints to achieve the goals of supply and demand balance and cost optimization.

[0155] The introduction of the cross decision tree network into this system is a major technical innovation of the embodiment of the present invention. As a powerful machine learning model, the decision tree network can accurately predict and classify unknown situations by learning complex patterns and correlations in historical data. In the scenario of power load optimization scheduling, the cross decision tree network is used to comprehensively analyze the multi-dimensional information from the transmission data analysis center and the virtual power plant resource coordination system to generate strategic recommendations on power load optimization scheduling. These recommendations not only take into account the current power supply and demand situation, but also proactively predict the possible load change trend in the future, thereby providing a scientific basis for scheduling decisions.

[0156] With the support of the above technical solutions, we can further design how to optimize the scheduling of time-series power loads in the target power area based on the outputs of these cross-decision tree networks.

[0157] 1. Data integration and analysis: First, the power transmission data analysis platform system will collect and integrate power load data from various monitoring points in real time, as well as relevant data on external influencing factors such as meteorology, economy, and social activities. These data are then sent to the cross decision tree network for in-depth analysis to identify the laws and trends of power load changes.

[0158] 2. Load forecasting and scenario simulation: Based on historical data and current conditions, the cross decision tree network will generate forecast results for power load in the future. At the same time, it will also simulate possible results under different dispatching strategies and evaluate the impact of various strategies on power supply and demand balance, system operating costs, and user satisfaction.

[0159] 3. Generation of optimized dispatching strategies: After comprehensively considering the forecast results, simulation scenarios, and the availability and cost-effectiveness of virtual power plant resources, the cross decision tree network will output one or more sets of optimized dispatching strategies. These strategies specify in detail how to adjust the output of various resources in the virtual power plant in different time periods, how to deploy energy storage devices, and how to guide user behavior to achieve effective management of power load.

[0160] 4. Strategy execution and feedback adjustment: Finally, the generated optimized dispatch strategy will be passed to the target virtual power plant resource coordination system, which will be responsible for its specific implementation. During the implementation process, the system will continuously collect real-time feedback data, monitor the execution effect of the strategy, and make necessary adjustments based on actual conditions. This closed-loop feedback mechanism ensures that the dispatch strategy can adapt to the ever-changing power market environment and maintain its effectiveness and advancement.

[0161] Through the above steps, not only the accuracy of power load forecasting and the scientific nature of scheduling decisions are improved, but also the overall operation efficiency and reliability of the power system are effectively improved through the flexible configuration of virtual power plant resources.

[0162] Under some independent design ideas, the power load optimization scheduling viewpoint output by the cross decision tree network linked and implanted in the power transmission data analysis middle-end system and the target virtual power plant resource coordination system performs optimal scheduling of the target power area based on the time series power load, including: inputting the power load optimization scheduling viewpoint into the semantic recognition thread in the scheduling decision cloud server, and obtaining the first suggested viewpoint semantics and the second suggested viewpoint semantics of the power load optimization scheduling viewpoint output by the semantic recognition thread, wherein the semantic recognition thread includes a plurality of residual cores connected to each other, the first suggested viewpoint semantics is the suggested viewpoint semantics output by the residual cores except the end of the plurality of residual cores connected to each other, and the second suggested viewpoint The semantics is the suggested view semantics output by the last residual core among the multiple residual cores connected to each other; the second suggested view semantics is input into the hierarchical scheduling thread in the scheduling decision cloud server to obtain the hierarchical scheduling task strategy output by the hierarchical scheduling thread; the first suggested view semantics, the second suggested view semantics and the third suggested view semantics and the hierarchical scheduling task strategy are input into the scheduling conflict response thread in the scheduling decision cloud server to obtain the timing scheduling execution strategy output by the scheduling conflict response thread, wherein the third suggested view semantics is the suggested view semantics output by the residual core in the hierarchical scheduling thread according to the augmented semantics, and the augmented semantics is the semantics obtained by augmenting the second suggested view semantics.

[0163] In the above embodiment, the generation of the viewpoints of power load optimization scheduling is the starting point of this technical solution. These viewpoints are derived from the cross decision tree network, which mines the laws and trends of power load changes by analyzing a large amount of data in the transmission data analysis center and the target virtual power plant resource coordination system, and thus proposes suggestions for optimized scheduling. These suggestions are based on in-depth learning of historical data and comprehensive consideration of multiple factors such as the current power market, meteorological conditions, and user behavior.

[0164] Next, these power load optimization scheduling ideas are input into the semantic recognition thread in the scheduling decision cloud server. The task of this thread is to parse and understand these ideas and convert them into executable scheduling strategies. To achieve this goal, the semantic recognition thread uses multiple interconnected residual cores. The residual core is a deep learning model with powerful feature extraction and representation learning capabilities. Through the cascade of multiple residual cores, the semantic recognition thread can gradually parse the optimization scheduling ideas and extract the key information and semantics.

[0165] In this process, except for the residual core at the end, all other residual cores will output a suggested view semantics, which are called the first suggested view semantics. They represent the understanding and interpretation of the optimization scheduling view at different levels and angles. The residual core at the end outputs the second suggested view semantics, which is the final understanding and refinement of the entire optimization scheduling view.

[0166] Then, the semantics of the second suggestion viewpoint are further input into the hierarchical scheduling thread in the scheduling decision cloud server. The task of this thread is to formulate a specific hierarchical scheduling task strategy based on the semantics of the second suggestion viewpoint. Hierarchical scheduling is an effective scheduling method that can decompose complex scheduling tasks into multiple levels and stages, each of which has clear goals and tasks. In this way, the hierarchical scheduling thread can transform the semantics of the second suggestion viewpoint into a practical scheduling strategy, providing guidance for subsequent scheduling execution.

[0167] At the same time, in order to enhance the diversity and adaptability of scheduling strategies, the residual core in the hierarchical scheduling thread also outputs the third suggestion view semantics based on the augmented semantics. Augmented semantics is the semantics obtained by expanding and supplementing the second suggestion view semantics, which contains more context information and detailed descriptions. By introducing augmented semantics, the third suggestion view semantics can provide richer and more comprehensive scheduling suggestions, providing more choices and references for scheduling conflict response threads.

[0168] Finally, the first suggestion view semantics, the second suggestion view semantics, the third suggestion view semantics and the hierarchical scheduling task strategy are input together into the scheduling conflict response thread in the scheduling decision cloud server. The task of this thread is to resolve conflicts and problems that may arise during the scheduling execution process. Due to the complexity and uncertainty of the power system, the scheduling strategy may encounter various unexpected challenges and difficulties during the execution process. By comprehensively considering multiple suggestion view semantics and hierarchical scheduling task strategies, the scheduling conflict response thread can flexibly respond to these challenges and difficulties and formulate the final timing scheduling execution strategy.

[0169] The time-series scheduling execution strategy is the final output of this technical solution. It is based on the optimization scheduling goal of time-series power load, takes into account the comprehensive impact of multiple proposed viewpoint semantics and hierarchical scheduling task strategies, and formulates a specific, feasible and effective scheduling execution plan. This plan can guide the scheduling decision-making and execution operation of the power system in the actual operation process, and realize the optimal management and efficient utilization of power load.

[0170] In this way, by linking and implanting the cross decision tree network, semantic recognition thread, hierarchical scheduling thread and scheduling conflict response thread, the optimal scheduling of the target power area based on the time series power load is realized. This solution not only makes full use of advanced data analysis and decision-making technology, but also takes into account the complexity and uncertainty of the power system, and formulates a specific, feasible and effective scheduling execution strategy.

[0171] Further, Figure 2 FIG. 2 is a schematic diagram of the structure of an electronic device 200 provided in an embodiment of the present invention. Figure 2 The electronic device 200 shown includes a processor 210. The processor 210 can call and run a computer program from a memory to implement the method in the embodiment of the present invention.

[0172] Alternatively, if Figure 2 As shown, the electronic device 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method in the embodiment of the present invention.

[0173] The memory 230 may be a separate device independent of the processor 210 , or may be integrated into the processor 210 .

[0174] Alternatively, if Figure 2 As shown, the electronic device 200 may further include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices, specifically, to send information or data to other devices, or to receive information or data sent by other devices.

[0175] Optionally, the power transmission data analysis middle-office system can implement the corresponding processes corresponding to the storage engine or components in the storage engine (such as a processing module) or the device deployed with the storage engine in each method of the embodiments of the present invention. For the sake of brevity, they will not be repeated here.

[0176] It should be understood that the processor of the embodiment of the present invention may be an integrated circuit chip with signal processing capability.

[0177] It is understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memory. It should be noted that the memory of the systems and methods described herein is intended to include but is not limited to suitable types of memory.

[0178] Based on the above, a readable storage medium is provided, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the above method are implemented.

[0179] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the purpose of the present invention and the scope of protection of the present invention, all of which are within the protection of the present invention.

Claims

1. A method for analyzing power transmission data based on virtual power plants. It is characterized in that Applied to the power transmission data analysis platform system, the method includes: S202： Acquire X pieces of initial distributed sensor metering information, perform state change knowledge mining on the X pieces of initial distributed sensor metering information through an AI knowledge mining branch in a joint data analysis algorithm, and obtain power transmission state change knowledge vectors corresponding to the X pieces of initial distributed sensor metering information respectively; X is a positive integer; The joint data analysis algorithm also includes a deep residual branch and a multivariate decision tree branch respectively set in X virtual power plant resource coordination systems; the X initial distributed sensor metering information is obtained by extracting the associated intelligent sensor acquisition information through the pre-matched distributed lightweight algorithm of the X virtual power plant resource coordination system; the pre-matched distributed lightweight algorithm of each virtual power plant resource coordination system is obtained after optimizing the branch network weight of the deep residual branch; S204： Sending one of the X transmission state change knowledge vectors to each of the X virtual power plant resource coordination systems; the transmission state change knowledge vector received by the target virtual power plant resource coordination system corresponds to the initial distributed sensor metering information extracted by the target virtual power plant resource coordination system; the target virtual power plant resource coordination system is any one of the X virtual power plant resource coordination systems; S206： When the distributed lightweight algorithms and edge decision tree branches respectively set by the X virtual power plant resource coordination systems all meet the edge side debugging compliance requirements, receiving the target algorithm debugging errors respectively reported by the X virtual power plant resource coordination systems; the target algorithm debugging errors reported by the target virtual power plant resource coordination system are generated by the edge decision tree branches pre-matched by the target virtual power plant resource coordination system and the received power transmission state change knowledge vector, and the edge decision tree branches pre-matched by each virtual power plant resource coordination system are obtained after optimizing the branch network weights of the multivariate decision tree branches; S208： The branch network weights of the AI ​​knowledge mining branches are optimized through X target algorithm debugging errors to obtain a cross knowledge mining module. Based on the cross knowledge mining module and the distributed lightweight algorithms and edge decision tree branches in the X virtual power plant resource coordination systems, a cross decision tree network corresponding to each virtual power plant resource coordination system is generated; wherein the cross decision tree network linked and implanted in the power transmission data analysis middle platform system and the target virtual power plant resource coordination system is used to perform power load optimization scheduling analysis on the intelligent sensor collection information in the target virtual power plant resource coordination system.

2. The method according to claim 1, It is characterized in that Also includes: Obtain energy relationship heat maps corresponding to X virtual power plant resource coordination systems, obtain power cluster operation information that has relationship heat commonality with the energy relationship heat map, and learn and debug the long short-term memory network based on the power cluster operation information to obtain a joint data analysis algorithm that is in a stable state in multiple feature dimensions; The deep residual branch and the multivariate decision tree branch of the joint data analysis algorithm are sent to the X virtual power plant resource coordination systems, so that each virtual power plant resource coordination system optimizes the branch network weights of the pre-matched deep residual branches based on the associated intelligent sensor collection information and the AI ​​knowledge mining branch implanted in the power transmission data analysis middle-end system to obtain a distributed lightweight algorithm, and optimizes the branch network weights of the pre-matched multivariate decision tree branches to obtain edge decision tree branches.

3. The method according to claim 1, It is characterized in that The obtaining of X initial distributed sensor measurement information includes: Obtain an edge monitoring cycle, select X virtual power plant resource coordination systems from Y virtual power plant resource coordination systems according to the edge monitoring cycle, and send a learning and debugging task to the X virtual power plant resource coordination systems, so that each virtual power plant resource coordination system performs sensor metering identification on the intelligent sensor collection information through a distributed lightweight algorithm to obtain initial distributed sensor metering information; Y is a positive integer greater than X; Obtaining initial distributed sensor metering information reported by the X virtual power plant resource coordination systems respectively; The state change knowledge mining is performed on the X initial distributed sensor metering information by the AI ​​knowledge mining branch in the joint data analysis algorithm to obtain the power transmission state change knowledge vectors corresponding to the X initial distributed sensor metering information, including: Performing cross-feature downsampling processing on the X initial distributed sensor measurement information to obtain cross-feature downsampling features corresponding to the X initial distributed sensor measurement information respectively; A state mapping operation is performed on the X cross-downsampled features to obtain state mapping intermediate features corresponding to the X cross-downsampled features respectively, and a power transmission state change knowledge vector corresponding to the X initial distributed sensor metering information respectively is generated based on the X state mapping intermediate features.

4. The method according to claim 1, It is characterized in that The branch network weight optimization of the AI ​​knowledge mining branch is performed through X target algorithm debugging errors to obtain a cross knowledge mining module, including: The local algorithm debugging errors corresponding to the X target algorithm debugging errors are generated by using the X target algorithm debugging errors and the branch network weights in the AI ​​knowledge mining branch. Based on the X local algorithm debugging errors and the preset debugging hyperparameters, the loss change partial derivative variables are generated. The branch network weights of the AI ​​knowledge mining branch are optimized according to the loss change partial derivative variables to obtain a cross knowledge mining module.

5. The method according to claim 1, It is characterized in that The generating of a cross decision tree network corresponding to each virtual power plant resource coordination system based on the cross knowledge mining module and the distributed lightweight algorithm and edge decision tree branches in the X virtual power plant resource coordination systems includes: Performing a weighted operation on the branch network weights of the distributed lightweight algorithms and the branch network weights of the edge decision tree branches in the X virtual power plant resource coordination systems to obtain an algorithm optimization indication feature, and sending the algorithm optimization indication feature to the X virtual power plant resource coordination systems, so that each virtual power plant resource coordination system performs neuron weight optimization on the pre-matched distributed lightweight algorithm and edge decision tree branch according to the algorithm optimization indication feature to obtain a lightweight crossover algorithm and a crossover decision branch; When the cross-knowledge mining module meets the debugging index of the middle-office system, the cross-knowledge mining module is combined with the lightweight cross-algorithm and cross-decision branches pre-matched with each virtual power plant resource coordination system to obtain a cross-decision tree network corresponding to each virtual power plant resource coordination system; wherein the cross-knowledge mining module is combined with the lightweight cross-algorithm and cross-decision branches pre-matched with the target virtual power plant resource coordination system to obtain a cross-decision tree network corresponding to the target virtual power plant resource coordination system; The algorithm optimization indication feature includes knowledge mining optimization variables and decision processing optimization variables; the weighted operation is performed on the branch network weights of the distributed lightweight algorithms in the X virtual power plant resource coordination systems and the branch network weights of the edge decision tree branches to obtain the algorithm optimization indication feature, including: Obtaining the branch network weights of the distributed lightweight algorithms in the X virtual power plant resource coordination systems, performing weighted operations on the branch network weights of the X distributed lightweight algorithms, and obtaining knowledge mining optimization variables corresponding to the X virtual power plant resource coordination systems respectively; the knowledge mining optimization variables are used to be sent to the corresponding virtual power plant resource coordination system, and are used to optimize the neuron weights of the distributed lightweight algorithms corresponding to the corresponding virtual power plant resource coordination system, and obtain a lightweight crossover algorithm; The branch network weights of the edge decision tree branches in the X virtual power plant resource coordination systems are obtained, and the branch network weights of the X edge decision tree branches are weighted to obtain the decision processing optimization variables corresponding to the X virtual power plant resource coordination systems respectively; the decision processing optimization variables are used to be sent to the corresponding virtual power plant resource coordination system, and are used to optimize the neuron weights of the edge decision tree branches corresponding to the corresponding virtual power plant resource coordination system to obtain cross decision branches.

6. A power transmission data analysis platform system, It is characterized in that The power transmission data analysis middle platform system communicates with the virtual power plant resource coordination system, and the power transmission data analysis middle platform system is used to: Obtain X initial distributed sensor metering information, perform state change knowledge mining on the X initial distributed sensor metering information through the AI ​​knowledge mining branch in the joint data analysis algorithm, and obtain the transmission state change knowledge vectors corresponding to the X initial distributed sensor metering information; X is a positive integer; the joint data analysis algorithm also includes a deep residual branch and a multivariate decision tree branch respectively set in the X virtual power plant resource coordination systems; the X initial distributed sensor metering information is obtained by extracting the associated intelligent sensor acquisition information through the pre-matched distributed lightweight algorithm of the X virtual power plant resource coordination system; the pre-matched distributed lightweight algorithm of each virtual power plant resource coordination system is obtained after optimizing the branch network weight of the deep residual branch; Sending one of the X transmission state change knowledge vectors to each of the X virtual power plant resource coordination systems; the transmission state change knowledge vector received by the target virtual power plant resource coordination system corresponds to the initial distributed sensor metering information extracted by the target virtual power plant resource coordination system; the target virtual power plant resource coordination system is any one of the X virtual power plant resource coordination systems; When the distributed lightweight algorithms and edge decision tree branches respectively set by the X virtual power plant resource coordination systems all meet the edge side debugging compliance requirements, receiving the target algorithm debugging errors respectively reported by the X virtual power plant resource coordination systems; the target algorithm debugging errors reported by the target virtual power plant resource coordination system are generated by the edge decision tree branches pre-matched by the target virtual power plant resource coordination system and the received power transmission state change knowledge vector, and the edge decision tree branches pre-matched by each virtual power plant resource coordination system are obtained after optimizing the branch network weights of the multivariate decision tree branches; The branch network weights of the AI ​​knowledge mining branches are optimized through X target algorithm debugging errors to obtain a cross knowledge mining module. Based on the cross knowledge mining module and the distributed lightweight algorithms and edge decision tree branches in the X virtual power plant resource coordination systems, a cross decision tree network corresponding to each virtual power plant resource coordination system is generated; wherein the cross decision tree network linked and implanted in the power transmission data analysis middle platform system and the target virtual power plant resource coordination system is used to perform power load optimization scheduling analysis on the intelligent sensor collection information in the target virtual power plant resource coordination system.

7. The power transmission data analysis platform system according to claim 6, It is characterized in that The power transmission data analysis platform system is also used for: Obtain energy relationship heat maps corresponding to X virtual power plant resource coordination systems, obtain power cluster operation information that has relationship heat commonality with the energy relationship heat map, and learn and debug the long short-term memory network based on the power cluster operation information to obtain a joint data analysis algorithm that is in a stable state in multiple feature dimensions; The deep residual branch and the multivariate decision tree branch of the joint data analysis algorithm are sent to the X virtual power plant resource coordination systems, so that each virtual power plant resource coordination system optimizes the branch network weights of the pre-matched deep residual branches based on the associated intelligent sensor collection information and the AI ​​knowledge mining branch implanted in the power transmission data analysis middle-end system to obtain a distributed lightweight algorithm, and optimizes the branch network weights of the pre-matched multivariate decision tree branches to obtain edge decision tree branches.

8. The power transmission data analysis platform system according to claim 6, It is characterized in that The obtaining of X initial distributed sensor measurement information includes: Obtain an edge monitoring cycle, select X virtual power plant resource coordination systems from Y virtual power plant resource coordination systems according to the edge monitoring cycle, and send a learning and debugging task to the X virtual power plant resource coordination systems, so that each virtual power plant resource coordination system performs sensor metering identification on the intelligent sensor collection information through a distributed lightweight algorithm to obtain initial distributed sensor metering information; Y is a positive integer greater than X; Obtaining initial distributed sensor metering information reported by the X virtual power plant resource coordination systems respectively; The state change knowledge mining is performed on the X initial distributed sensor metering information by the AI ​​knowledge mining branch in the joint data analysis algorithm to obtain the power transmission state change knowledge vectors corresponding to the X initial distributed sensor metering information, including: Performing cross-feature downsampling processing on the X initial distributed sensor measurement information to obtain cross-feature downsampling features corresponding to the X initial distributed sensor measurement information respectively; A state mapping operation is performed on the X cross-downsampled features to obtain state mapping intermediate features corresponding to the X cross-downsampled features respectively, and a power transmission state change knowledge vector corresponding to the X initial distributed sensor metering information respectively is generated based on the X state mapping intermediate features.

9. The power transmission data analysis platform system according to claim 6, It is characterized in that The branch network weight optimization of the AI ​​knowledge mining branch is performed through X target algorithm debugging errors to obtain a cross knowledge mining module, including: The local algorithm debugging errors corresponding to the X target algorithm debugging errors are generated by using the X target algorithm debugging errors and the branch network weights in the AI ​​knowledge mining branch. Based on the X local algorithm debugging errors and the preset debugging hyperparameters, the loss change partial derivative variables are generated. The branch network weights of the AI ​​knowledge mining branch are optimized according to the loss change partial derivative variables to obtain a cross knowledge mining module.

10. The power transmission data analysis platform system according to claim 6, It is characterized in that The generating of a cross decision tree network corresponding to each virtual power plant resource coordination system based on the cross knowledge mining module and the distributed lightweight algorithm and edge decision tree branches in the X virtual power plant resource coordination systems includes: Performing a weighted operation on the branch network weights of the distributed lightweight algorithms and the branch network weights of the edge decision tree branches in the X virtual power plant resource coordination systems to obtain an algorithm optimization indication feature, and sending the algorithm optimization indication feature to the X virtual power plant resource coordination systems, so that each virtual power plant resource coordination system performs neuron weight optimization on the pre-matched distributed lightweight algorithm and edge decision tree branch according to the algorithm optimization indication feature to obtain a lightweight crossover algorithm and a crossover decision branch; When the cross-knowledge mining module meets the debugging index of the middle-office system, the cross-knowledge mining module is combined with the lightweight cross-algorithm and cross-decision branches pre-matched with each virtual power plant resource coordination system to obtain a cross-decision tree network corresponding to each virtual power plant resource coordination system; wherein the cross-knowledge mining module is combined with the lightweight cross-algorithm and cross-decision branches pre-matched with the target virtual power plant resource coordination system to obtain a cross-decision tree network corresponding to the target virtual power plant resource coordination system; The algorithm optimization indication feature includes knowledge mining optimization variables and decision processing optimization variables; the weighted operation is performed on the branch network weights of the distributed lightweight algorithms in the X virtual power plant resource coordination systems and the branch network weights of the edge decision tree branches to obtain the algorithm optimization indication feature, including: Obtaining the branch network weights of the distributed lightweight algorithms in the X virtual power plant resource coordination systems, performing weighted operations on the branch network weights of the X distributed lightweight algorithms, and obtaining knowledge mining optimization variables corresponding to the X virtual power plant resource coordination systems respectively; the knowledge mining optimization variables are used to be sent to the corresponding virtual power plant resource coordination system, and are used to optimize the neuron weights of the distributed lightweight algorithms corresponding to the corresponding virtual power plant resource coordination system, and obtain a lightweight crossover algorithm; The branch network weights of the edge decision tree branches in the X virtual power plant resource coordination systems are obtained, and the branch network weights of the X edge decision tree branches are weighted to obtain the decision processing optimization variables corresponding to the X virtual power plant resource coordination systems respectively; the decision processing optimization variables are used to be sent to the corresponding virtual power plant resource coordination system, and are used to optimize the neuron weights of the edge decision tree branches corresponding to the corresponding virtual power plant resource coordination system to obtain cross decision branches.

11. An electronic device, It is characterized in that The method comprises at least one processor (210) and a memory (230); the memory (230) stores computer-executable instructions; the at least one processor (210) executes the computer-executable instructions stored in the memory (230), so that the at least one processor (210) executes the power transmission data analysis method based on a virtual power plant as described in any one of claims 1 to 5 above; Also included is a transceiver (220), the processor (210) controls the transceiver (220) to interact with other devices; A transceiver sends information or data to other devices, or receives information or data sent by other devices.

12. A readable storage medium, It is characterized in that The readable storage medium stores programs or instructions, and when the programs or instructions are executed by the processor, the steps of the power transmission data analysis method based on the virtual power plant as described in any of the above claims 1-5 are implemented.

Images