Configuration method, device and system of comprehensive energy

Abstract

The invention discloses a comprehensive energy configuration method, device and system. The method comprises the following steps: acquiring to-be-configured system data; analyzing the to-be-configured system data based on the to-be-configured system data and the data similarity of each historical configuration scheme in the historical configuration scheme library, adjusting the historical configuration schemes, and giving a target configuration scheme; according to the target configuration scheme, obtaining a configuration structure of the comprehensive energy, obtaining a configuration structure change of the comprehensive energy in real time, and adjusting the comprehensive energy operation model to obtain a target operation model; based on the target operation model, obtaining real-time operation data of the comprehensive energy; performing feature extraction on the real-time operation data in combination with a pre-constructed evaluation index library to obtain target evaluation features; the evaluation weights corresponding to the target evaluation features are fused, and an operation evaluation result of the comprehensive energy is given; and according to the operation evaluation result, adjusting the target operation model to complete the configuration of the comprehensive energy, and realizing the improvement of the stability of the comprehensive energy.

Description

Technical Field

[0001] The present invention belongs to the technical field of integrated energy systems, and specifically relates to an integrated energy configuration method, device and system. Background Art

[0002] An integrated energy system (IES) utilizes advanced physical information technology and innovative management models within a specific region to integrate multiple energy sources, achieve coordinated planning and optimized operation among multiple energy subsystems, and effectively improve energy efficiency and promote sustainable energy development while meeting diverse energy demands. IES management regulates energy flows through information flow, optimizes and controls energy equipment within the system, reduces overall system operating costs, improves energy efficiency, and ensures safe and reliable system operation. It is a key guarantee for the stable and efficient operation of IESs and the core of their flexible operation. Furthermore, accurate assessment of IESs is a prerequisite for improving IES efficiency.

[0003] Patent application CN112182887A discloses a comprehensive energy system planning optimization simulation method, including: Step S1, establishing an EQC-based efficiency analysis model based on energy characteristics; Step S2, establishing a four-layer energy center model based on multi-energy coupling analysis of the comprehensive energy system; Step S3, establishing an improved convex programming model based on non-convex factors in the comprehensive energy system; and Step S4, establishing a comprehensive energy system planning optimization model based on economy and efficiency. This multi-objective planning optimization, with economy and efficiency as its goals, achieves a qualitative match in the comprehensive energy utilization process, achieving a comprehensive performance balance across all aspects.

[0004] Currently, integrated energy planning is based on its energy characteristics. However, as integrated energy systems operate under different environments for a period of time, they will undergo certain changes. If a fixed planning scheme is used, the efficiency of integrated energy utilization cannot be guaranteed, and there are certain risks and uncertainties. How to formulate and adjust integrated energy planning schemes and improve the stability of integrated energy is an urgent issue. Summary of the Invention

[0005] In response to the defects existing in the above-mentioned prior art, the present invention provides a method, device and system for configuring integrated energy, the method comprising: obtaining data of a system to be configured; analyzing the data of the system to be configured based on the data similarity between the data of the system to be configured and each historical configuration scheme in a historical configuration scheme library, adjusting the historical configuration scheme and providing a target configuration scheme; obtaining the configuration structure of the integrated energy according to the target configuration scheme, and obtaining the configuration structure changes of the integrated energy in real time, adjusting the integrated energy operation model and obtaining the target operation model; obtaining the real-time operation data of the integrated energy based on the target operation model; extracting features of the real-time operation data in combination with a pre-built evaluation index library to obtain target evaluation features; integrating the evaluation weights corresponding to the target evaluation features to provide an operation evaluation result of the integrated energy; adjusting the target operation model according to the operation evaluation result, completing the configuration of the integrated energy, and improving the stability of the integrated energy. By analyzing the data of the configuration system, the preliminary planning of the integrated energy is completed, the target configuration plan is obtained, and the operation model of the integrated energy is adjusted based on the target configuration plan. The target operation model is given, and adjustments are made in real time based on the operation changes and configuration changes of the integrated energy. To a certain extent, the utilization efficiency of the integrated energy is guaranteed, the risk and uncertainty of the integrated energy are reduced, and the stability of the integrated energy is improved.

[0006] In a first aspect, the present invention provides a method for configuring a comprehensive energy source, comprising the following steps: Get the system data to be configured; Based on the data similarity between the system data to be configured and the historical configuration solutions in the historical configuration solution library, the system data to be configured is analyzed, and the historical configuration solutions are adjusted to provide the target configuration solution. According to the target configuration plan, the configuration structure of the comprehensive energy is obtained, and the changes in the configuration structure of the comprehensive energy are obtained in real time, and the comprehensive energy operation model is adjusted to obtain the target operation model; Based on the target operation model, obtain the real-time operation data of comprehensive energy; Combined with the pre-built evaluation index library, feature extraction is performed on real-time operation data to obtain target evaluation features; The evaluation weights corresponding to the target evaluation characteristics are integrated to provide the comprehensive energy operation evaluation results; According to the operation evaluation results, the target operation model is adjusted to complete the configuration of comprehensive energy.

[0007] Furthermore, the data similarity between the system data to be configured and each historical configuration solution in the historical configuration solution library is obtained by the following steps: Based on the type of system data to be configured, a similarity index analysis strategy is matched, and the local similarity of the historical data corresponding to each historical configuration solution is analyzed separately; Combining the similarity weights corresponding to different types, the local similarities corresponding to each historical configuration scheme are fused to obtain the data similarity corresponding to each historical configuration scheme.

[0008] Furthermore, the types of data to be configured include discrete and continuous types, and the local similarity includes discrete similarity and continuous similarity; Based on the type of system data to be configured, a similarity index analysis strategy is matched and the local similarity of the historical data corresponding to each historical configuration solution is analyzed separately, including: Using a first similarity index, analyzing discrete similarities between discrete system data to be configured and corresponding discrete data in historical configuration solutions; The second similarity index is used to analyze the continuous similarity between the continuous system data to be configured and the corresponding continuous data in the historical configuration scheme.

[0009] Furthermore, similarity weights include discrete similarity weights and continuous similarity weights; Combining the similarity weights corresponding to different types, the local similarities corresponding to each historical configuration scheme are fused to obtain the data similarity corresponding to each historical configuration scheme, specifically including: Normalize the discrete similarity and continuous similarity respectively to obtain the standard discrete similarity and standard continuous similarity; Combining the discrete similarity weight and the continuous similarity weight, the standard discrete similarity and the standard continuous similarity are weighted and summed to obtain the data similarity.

[0010] Furthermore, based on the data similarity between the system data to be configured and each historical configuration solution in the historical configuration solution library, the system data to be configured is analyzed, and the historical configuration solutions are adjusted to provide a target configuration solution, specifically including: Sort the historical configuration plans in the historical configuration plan library according to data similarity, and give the pending historical plans based on the sorting results; Based on a pre-acquired configuration mapping function, the system data to be configured is mapped to obtain a target configuration solution, wherein the pre-acquired configuration mapping function is obtained by analyzing and correcting a historical solution to be determined.

[0011] Furthermore, according to the target configuration scheme, the configuration structure of the comprehensive energy is obtained, which specifically includes: From the target configuration plan, the various energy devices and loads in the integrated energy are obtained, and the nodes of the integrated directed graph are determined; Analyze the operating status of each energy device and load in the comprehensive energy system, and determine the attributes of each node in the comprehensive directed graph; Combined with the connection relationship and connection status of each energy device and load in the comprehensive energy, determine the edges and edge attributes corresponding to each node in the comprehensive directed graph; By integrating the nodes, attributes of the nodes, edges corresponding to the nodes and edge attributes of the comprehensive directed graph, the construction of the comprehensive directed graph is completed and the configuration structure of the comprehensive energy is obtained.

[0012] Furthermore, the configuration structure changes of the integrated energy are obtained in real time, the integrated energy operation model is adjusted, and the target operation model is obtained, which specifically includes: Based on the configuration structure change of comprehensive energy, the configuration structure change is quantified to give the state change of comprehensive energy; According to the state change threshold obtained in advance, the state change amount is judged and a dynamic exploration factor is given. The dynamic exploration factor is used to influence the action selection of the integrated energy at the next moment; Based on the dynamic exploration factor and combined with the system loss function, the comprehensive energy operation model is trained until convergence to obtain the target operation model.

[0013] Furthermore, based on the configuration structure changes of the comprehensive energy, the configuration structure changes are quantified to provide the state change of the comprehensive energy, including: Obtain a comprehensive directed graph at two adjacent moments, analyze the changes in the number of nodes and the number of edges in the comprehensive directed graph corresponding to the two adjacent moments, and obtain the change results; Based on the change results, analyze the changes of each node and edge in the comprehensive directed graph at two adjacent moments to obtain the node change amount and edge change amount; The node change and edge change are combined to obtain the state change.

[0014] Furthermore, the construction of the integrated energy operation model is determined by the following steps: Based on the initial deep learning model, combined with the comprehensive directed graph and the system action space, the expected cumulative reward values corresponding to different states and actions are obtained. The system action space includes various actions performed on each energy device and load in the comprehensive energy. Based on the expected cumulative reward values corresponding to different states and actions, the mean and variance of the expected cumulative reward values are given, and the state distribution of the expected cumulative reward values corresponding to different states and actions is obtained; According to the state distribution of the expected cumulative reward value, the model parameters of the integrated energy operation model are determined, and the construction of the integrated energy operation model is completed.

[0015] Furthermore, combined with the pre-built evaluation index library, feature extraction is performed on the real-time operation data to obtain the target evaluation features, including: Based on the pre-built comprehensive energy indicator library, the corresponding initial assessment features are screened from the real-time operation data; The principal component analysis method is used to analyze the initial evaluation features and give the core features; The target evaluation features are determined based on the similarity between each initial evaluation feature and the core feature.

[0016] Furthermore, principal component analysis was used to analyze the initial evaluation features and provide core features, including: Analyze the linear correlation between each initial evaluation feature and construct the covariance matrix of the initial evaluation features; The covariance matrix is decomposed to give multiple eigenvalues; According to the variance explanation ratio of each eigenvalue and the preset explanation threshold, multiple eigenvalues are screened to obtain the core features.

[0017] Furthermore, the linear correlation between each initial evaluation feature is analyzed, and the covariance matrix of the initial evaluation features is constructed, which specifically includes: According to the elements in each initial evaluation feature and the feature mean of the corresponding initial evaluation feature, the deviation feature of each initial evaluation feature is given; Based on the deviation characteristics of any two initial evaluation characteristics, the mean of the product of the two deviation characteristics is given to obtain the degree of linear correlation between the initial evaluation characteristics; The linear correlation degree between each initial evaluation feature is integrated to construct the covariance matrix of the initial evaluation features.

[0018] Furthermore, based on the variance explanation ratio of each eigenvalue and the preset explanation threshold, multiple eigenvalues are screened to obtain core features, including: According to the variance explanation ratio of each eigenvalue, the cumulative variance explanation ratio corresponding to each eigenvalue is given; Compare the cumulative variance explanation ratio corresponding to each eigenvalue with the preset explanation threshold to obtain the comparison result; According to the comparison results, the eigenvalues are screened and the eigenvectors corresponding to the screened eigenvalues are given; The standard data matrix is projected into a vector matrix composed of eigenvectors corresponding to the filtered eigenvalues to obtain the core features.

[0019] Furthermore, the target operation model includes a reward function, wherein the reward function is composed of an integrated energy generation part, an integrated energy storage part, and an integrated energy low-carbon operation part; Based on the results of the operation evaluation, the target operation model is adjusted to complete the configuration of comprehensive energy, including: If the operation evaluation result is not up to standard, analyze the sub-scores of each part of the operation evaluation result; Analyze the compliance of each sub-score and adjust the coefficients of the corresponding parts of the reward function in the target operation model based on the compliance; Based on the adjusted reward function, the target operation model is trained until convergence; Run the model according to the target after training, run the comprehensive energy, and complete the configuration of the comprehensive energy.

[0020] In a second aspect, the present invention further provides a comprehensive energy configuration device, which adopts any of the above comprehensive energy configuration methods, including: Perception module, used to obtain system data to be configured; Storage module, used to store historical configuration solution library; The calculation module is used to analyze the data of the system to be configured based on the data similarity between the system data to be configured and the various historical configuration schemes in the historical configuration scheme library, and adjust the historical configuration schemes to provide a target configuration scheme; according to the target configuration scheme, the configuration structure of the comprehensive energy is obtained, and the configuration structure changes of the comprehensive energy are obtained in real time, and the comprehensive energy operation model is adjusted to obtain the target operation model; based on the target operation model, the real-time operation data of the comprehensive energy is obtained; in combination with the pre-built evaluation index library, the real-time operation data is extracted to obtain the target evaluation features; the evaluation weights corresponding to the target evaluation features are integrated to provide the operation evaluation results of the comprehensive energy; according to the operation evaluation results, the target operation model is adjusted to complete the configuration of the comprehensive energy.

[0021] In a third aspect, the present invention further provides a comprehensive energy configuration system, comprising: a network device, an execution device, and the comprehensive energy configuration device as described above.

[0022] The present invention provides a comprehensive energy configuration method, device, and system, which have at least the following beneficial effects: (1) By analyzing the data of the configuration system, the preliminary planning of the integrated energy is completed, the target configuration plan is obtained, and the operation model of the integrated energy is adjusted based on the target configuration plan. The target operation model is given, and adjustments are made in real time based on the operation changes and configuration changes of the integrated energy. To a certain extent, the utilization efficiency of the integrated energy is guaranteed, the risk and uncertainty of the integrated energy are reduced, and the stability of the integrated energy is improved.

[0023] (2) By integrating the initial deep learning model, the system action space, and the reward function, the state distribution of the expected cumulative reward values corresponding to different system actions and different system states is analyzed, and the uncertainty of the system state and system action is modeled to achieve dynamic adjustment of the model parameters.

[0024] (3) By analyzing the configuration system data, based on data similarity, historical configuration plans with high similarity are screened from the historical configuration plan library, and the target configuration plan is determined. While ensuring the comprehensive energy configuration effect, the comprehensive energy configuration efficiency is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A flowchart of a method for configuring integrated energy provided by an embodiment of the present invention; Figure 2 A flowchart of a target configuration solution provided in an embodiment of the present invention; Figure 3 A flowchart of obtaining a target operation model provided by an embodiment of the present invention; Figure 4 A flowchart for determining target evaluation features provided by an embodiment of the present invention; Figure 5 A flowchart for determining evaluation weights provided in an embodiment of the present invention; Figure 6 A flowchart for configuring integrated energy provided by an embodiment of the present invention; Figure 7 This is a structural block diagram of the comprehensive energy configuration device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0026] To better understand the above technical solution, the following will be described in detail with reference to the accompanying drawings and specific implementation methods. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0027] The terms used in the embodiments of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The singular forms "a," "an," "the," and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, and unless the context clearly indicates otherwise, "a plurality" generally includes at least two.

[0028] It should also be noted that the terms "include," "comprises," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a product or device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such product or device. In the absence of further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the product or device comprising the element.

[0029] The integrated energy system couples multiple energy sources such as electricity, gas, and heat, and realizes multi-energy synergy in production, transmission, distribution, utilization, storage and other links. It can improve the overall energy utilization efficiency and system operation flexibility, and has been widely developed.

[0030] Currently, there is a lack of comprehensive control methods for the entire energy process, from planning to management and control. This results in a lack of timely control when changes occur, leading to high energy consumption and poor intelligence in the operation of integrated energy. Ensuring comprehensive energy control throughout the entire process in a complex and dynamic environment is an urgent issue that needs to be addressed.

[0031] For integrated energy planning involving data collection and analysis, a storage-and-computing architecture is employed. This architecture integrates sensing, storage, and computing functions with in-situ data processing (computation and storage fusion) into small devices at the hardware level, creating intelligent sensing devices that integrate sensing, storage, and computing. These intelligent sensing devices, integrating sensing, storage, and computing, are deployed in energy supply networks, energy coupling devices, energy storage devices, and user terminals to enable near-source, efficient processing of integrated energy system status data (critical calculations are performed within the storage medium), supporting system planning and management.

[0032] like Figure 1 As shown, the embodiment of the present invention provides a method for configuring comprehensive energy, and the specific steps are as follows: S101: Acquire system data to be configured.

[0033] It's important to understand that the system data to be configured is required during comprehensive energy planning. This includes, but is not limited to, construction area, geographic location, land use planning, photovoltaic resources, wind resources, geothermal resources, current energy supply and consumption data, environmental data (such as sunshine intensity, wind speed, humidity, and temperature), energy market conditions (such as energy prices), and cooling, heating, and electricity load statistics. These load statistics can be presented as duration curves for electricity, heating, cooling, and gas loads. Other data may include the infrastructure status of power distribution and gas networks, typical daily load demands in different seasons, and specific parameters for energy storage equipment, photovoltaic power plants, and substations. The appropriate data is used based on the specific focus.

[0034] S102: Based on the data similarity between the system data to be configured and each historical configuration solution in the historical configuration solution library, the system data to be configured is analyzed, and the historical configuration solutions are adjusted to provide a target configuration solution.

[0035] Before analyzing data similarity, we must first understand the historical configuration plan library. Each historical configuration plan in the library is based on planning data for different regions and situations, as well as the corresponding configuration plan. A configuration plan analyzes and configures the planning data for the target area. Planning data includes construction area, geographic location, land use planning, photovoltaic resources, wind resources, geothermal resources, current energy supply and consumption data, environmental data (such as sunshine intensity, wind speed, humidity, and temperature), energy market conditions (such as energy prices), and cooling, heating, and electricity load statistics. Cooling, heating, and electricity load statistics can be presented as duration curves for electricity, heating, cooling, and gas loads. Other data may include the infrastructure status of the power distribution and gas grid, typical daily load demand in different seasons, and specific parameters for energy storage devices, photovoltaic power plants, substations, and other systems. The appropriate data is used based on the specific focus. Planning data is used to determine the target area's electricity demand (e.g., peak hours, base load), thermal demand (e.g., heating and hot water supply), cooling demand (e.g., air conditioning and refrigeration), and available energy resources. These include renewable energy sources such as solar, wind, and geothermal energy, traditional energy sources such as natural gas and electricity, and waste heat recovery from industrial processes. Based on the target area's specific conditions, appropriate energy technologies and equipment are selected, such as solar photovoltaic panels for power generation, wind turbines for wind power generation, ground-source heat pumps for heating and cooling, and energy storage devices for balancing supply and demand. Disparate energy technologies and equipment are integrated into a unified system, combining multiple energy sources to improve energy supply stability. Energy supply is adjusted based on changes in energy demand, and future energy demand is forecasted to optimize energy production and scheduling, resulting in a comprehensive configuration plan.

[0036] Different planning data will result in different corresponding configuration plans. The greater the difference in planning data, the lower the referenceability of the corresponding configuration plan. Therefore, by calculating the data similarity between the current system data to be configured and the planning data in each historical configuration plan, the referenceability of each historical configuration plan in the historical configuration plan library is determined. The higher the data similarity, the closer the system data to be configured is to the planning data of the historical configuration plan, and the higher the referenceability of the corresponding configuration plan. Based on the above analysis, the data similarity between the system data to be configured and each historical configuration plan in the historical configuration plan library is given. The specific steps are as follows: Based on the type of system data to be configured, a similarity index analysis strategy is matched, and the local similarity of the historical data corresponding to each historical configuration solution is analyzed separately; Combining the similarity weights corresponding to different types, the local similarities corresponding to each historical configuration scheme are fused to obtain the data similarity corresponding to each historical configuration scheme.

[0037] In the embodiments provided herein, all data is classified as continuous or discrete based on the value of the system data to be configured. Specifically, the types of data to be configured include both discrete and continuous data, and local similarity includes both discrete and continuous similarity. It will be appreciated that different similarity metrics can more accurately reflect the degree of similarity between data of different types. In other embodiments, the same similarity metric can be used for all system data to be configured, and this is not a limitation.

[0038] Furthermore, based on the type of system data to be configured, a similarity index analysis strategy is matched, and the local similarity of the historical data corresponding to each historical configuration solution is analyzed separately, specifically including: Using a first similarity index, analyzing discrete similarities between discrete system data to be configured and corresponding discrete data in historical configuration solutions; The second similarity index is used to analyze the continuous similarity between the continuous system data to be configured and the corresponding continuous data in the historical configuration scheme.

[0039] In the embodiment provided by the present invention, the above-mentioned first similarity index is the Jaccard coefficient, and the above-mentioned second similarity index is the Euclidean distance. In other embodiments, other similarity indexes may be used, such as Manhattan distance, Hamming distance, Mahalanobis distance, and Pearson correlation coefficient, etc., without limitation. Among them, the Jaccard coefficient is used to measure the ratio of the intersection to the union of two sets. The larger the Jaccard similarity coefficient, the more similar the two sets are. Euclidean distance is a metric for calculating the straight-line distance between two vectors. The smaller the Euclidean distance, the more similar the two vectors are. Manhattan distance calculates the sum of the absolute distances between two vectors on each coordinate axis. The smaller the Manhattan distance, the more similar the two vectors are. Hamming distance is used to measure the number of different characters between two strings of equal length. The smaller the Hamming distance, the more similar the two strings are. Mahalanobis distance takes into account the covariance matrix of the data and can better reflect the actual distance between data points. The Pearson correlation coefficient is used to measure the linear relationship between two variables. The value range is [-1,1], where 1 indicates a perfect positive correlation and -1 indicates a perfect negative correlation.

[0040] In a specific example, the continuous data in the system data to be configured i is calculated Corresponding continuous data in historical configuration scheme j The Euclidean distance is the continuous similarity, which is specifically expressed as:

[0041] in, is the continuous data in the system data i to be configured Corresponding continuous data in historical configuration scheme j Euclidean distance, p is continuous data Dimensions, Continuous data The kth element in Corresponding continuous data in historical configuration scheme j The kth element in the system data to be configured, n is the continuous data in the system data i The number of groups, m is the number of historical configuration schemes in the historical configuration scheme.

[0042] In a specific example, the discrete data set in the system data to be configured i is calculated Corresponding discrete data set in historical configuration scheme j The Jaccard coefficient is the discrete similarity, which is specifically expressed as:

[0043] in, is the discrete data set in the system data i to be configured Corresponding discrete data set in historical configuration scheme j Jaccard coefficient.

[0044] Furthermore, by combining the similarity weights corresponding to different types, the local similarities corresponding to each historical configuration scheme are fused to obtain the data similarity corresponding to each historical configuration scheme. The similarity weights include discrete similarity weights and continuous similarity weights, specifically including: Normalize the discrete similarity and continuous similarity respectively to obtain the standard discrete similarity and standard continuous similarity; Combining the discrete similarity weight and the continuous similarity weight, the standard discrete similarity and the standard continuous similarity are weighted and summed to obtain the data similarity.

[0045] In a specific implementation, the continuous similarity, ie, the Euclidean distance, is normalized to a range of [0, 1].

[0046]

[0047] in, is the continuous data in the system data i to be configured Corresponding continuous data in historical configuration scheme j The standard continuous similarity of For all The maximum value in .

[0048] Since the Jaccard coefficient ranges from [0, 1], there is no need to normalize the Jaccard coefficient.

[0049] Data similarity is specifically expressed as:

[0050] in, is the data similarity between the system data to be configured i and the historical configuration scheme j in the historical configuration scheme library, is the continuous similarity weight, is the discrete similarity weight.

[0051] It can be understood that the continuous similarity weight is the influence of continuous data on data similarity, and the discrete similarity weight is the influence of discrete data on data similarity. The values of the continuous similarity weight and the discrete similarity weight are set according to actual conditions and are not limited thereto.

[0052] Furthermore, based on the data similarity between the system data to be configured and each historical configuration scheme in the historical configuration scheme library, the system data to be configured is analyzed, and the historical configuration scheme is adjusted to give the target configuration scheme. Figure 2 , specifically including: Sort the historical configuration plans in the historical configuration plan library according to data similarity, and give the pending historical plans based on the sorting results; Based on a pre-acquired configuration mapping function, the system data to be configured is mapped to obtain a target configuration solution, wherein the pre-acquired configuration mapping function is obtained by analyzing and correcting a historical solution to be determined.

[0053] It can be understood that, based on the sorting results, historical configuration plans with higher data similarity are screened out to obtain pending historical plans. The pending historical plans have a higher reference value for configuring system data. Using the pending historical plans to determine the configuration mapping function can ensure the configuration effect of the target configuration plan and improve the overall energy utilization efficiency.

[0054] The above configuration mapping function is obtained through model training. In this example, the decision tree model is used. Other models can be used in other implementations and there is no limitation on this. First, set the model parameters of the decision tree model to complete the initialization of the decision tree model. The setting of the model parameters is set according to actual needs and there is no limitation on this. and configuration scheme data As the source domain, the system data to be configured As the target domain. It is understandable that the planning data in the source domain and the system data to be configured in the target domain are highly similar in distribution or feature form. If the distribution is relatively close, they can be aligned through a certain mapping, and the rule of "planning data → configuration solution data" in the source domain can be transferred to the target domain. The training is performed until convergence, and a configuration mapping function is obtained. After the configuration mapping function is obtained, the system data to be configured is substituted into the configuration mapping function to obtain the target configuration solution.

[0055] S103: According to the target configuration plan, the configuration structure of the integrated energy is obtained, and the changes in the configuration structure of the integrated energy are obtained in real time, and the integrated energy operation model is adjusted to obtain the target operation model.

[0056] Furthermore, according to the target configuration scheme, the configuration structure of the comprehensive energy is obtained, which specifically includes: From the target configuration plan, the various energy devices and loads in the integrated energy are obtained, and the nodes of the integrated directed graph are determined; Analyze the operating status of each energy device and load in the comprehensive energy system, and determine the attributes of each node in the comprehensive directed graph; Combined with the connection relationship and connection status of each energy device and load in the comprehensive energy, determine the edges and edge attributes corresponding to each node in the comprehensive directed graph; By integrating the nodes, attributes of the nodes, edges corresponding to the nodes and edge attributes of the comprehensive directed graph, the construction of the comprehensive directed graph is completed and the configuration structure of the comprehensive energy is obtained.

[0057] In a specific embodiment, a directed graph is used. To represent the integrated energy system, a comprehensive directed graph is obtained. N is a node set in the comprehensive directed graph, and each node in the node set represents various energy devices and loads in the integrated energy system. For example, N1 is a photovoltaic power station, N2 is a wind turbine, N3 is a gas generator, N4 is a kerosene generator, and N5 is a hydrogen energy storage device. For various types of power loads. The attribute vector of each node includes the power generation power, load power, energy storage status and operating status of each node. Among them, the power generation power is only applicable to power generation equipment in the integrated energy system, such as photovoltaic and wind turbines, the load power is only applicable to load nodes in the integrated energy system, and the energy storage status is only applicable to energy storage equipment. In this example, the value of the operating status is 1 or 0, 1 indicates that the energy equipment or load corresponding to the node is operating normally, and 0 indicates that the energy equipment or load corresponding to the node is out of service. In a specific example, the values of the load power and energy storage status in the attribute vector of a power generation equipment can be 0, empty, or other forms can be used to indicate that the attribute value is unavailable, and there is no limitation on this. In addition, for power generation equipment, the corresponding attribute vector only contains power generation power and operating status. Similarly, for loads, the corresponding attribute vector only includes load power and operating status. For energy storage equipment, the corresponding attribute vector only includes energy storage status and operating status.

[0058] E is the edge set that represents the connection between nodes in the comprehensive directed graph, such as power transmission lines, energy conversion channels, etc. Each edge represents the connection from node i to node j. Each edge The attribute vector includes transmission capacity, transmission loss and operation status, where the operation status takes the value of 1 or 0, 1 indicates that the corresponding edge is operating normally, and 0 indicates that the corresponding edge has a fault and is not operating normally.

[0059] By constructing a comprehensive directed graph based on the planning structure of the integrated energy system, the operating status and operating parameters of each energy device and load and their connection relationship in the integrated energy system can be obtained in a timely and clear manner. It can also capture faults or topology changes in the integrated energy system in a timely manner, providing a basis for timely adjustment of the operation of the integrated energy system and ensuring its stable operation.

[0060] Furthermore, the configuration structure changes of comprehensive energy are obtained in real time, the comprehensive energy operation model is adjusted, and the target operation model is obtained. Figure 3 , specifically including: Based on the configuration structure change of comprehensive energy, the configuration structure change is quantified to give the state change of comprehensive energy; According to the state change threshold obtained in advance, the state change amount is judged and a dynamic exploration factor is given. The dynamic exploration factor is used to influence the action selection of the integrated energy at the next moment; Based on the dynamic exploration factor and combined with the system loss function, the comprehensive energy operation model is trained until convergence to obtain the target operation model.

[0061] Furthermore, based on the configuration structure changes of the comprehensive energy, the configuration structure changes are quantified to provide the state changes of the comprehensive energy, including: Obtain a comprehensive directed graph at two adjacent moments, analyze the changes in the number of nodes and the number of edges in the comprehensive directed graph corresponding to the two adjacent moments, and obtain the change results; Based on the change results, analyze the changes of each node and edge in the comprehensive directed graph at two adjacent moments to obtain the node change amount and edge change amount; The node change and edge change are combined to obtain the state change.

[0062] In a specific embodiment, the comprehensive directed graph at two adjacent moments is obtained, that is, the comprehensive directed graph at moment t is obtained. And the comprehensive directed graph at time t+1 . Analyze the changes in the number of nodes in the two comprehensive directed graphs and get ,in, is the node set from time t+1 Find the node set at time t There are no nodes in is the node set at time t Find the node set at time t+1 If there is no node in ,Right now It is not an empty set, indicating that there are additions or deletions of nodes in the comprehensive directed graph between time t and time t+1. Represents the empty set.

[0063] Continuing to analyze the changes in the number of edges in the two comprehensive directed graphs, we get ,in, is the edge set from time t+1 Find the edge set at time t There is no edge in is the edge set from time t Find the edge set at time t+1 There are no edges in . If , indicating the addition or deletion of edges in the comprehensive directed graph between time t and time t+1.

[0064] if or , then it is determined that the configuration structure of the comprehensive energy has changed, that is, the change result is a change. and , it is determined that the configuration structure of the comprehensive energy has not changed, that is, the change result is no change.

[0065] When the configuration structure of the integrated energy changes, the node changes and edge changes are calculated to obtain the state changes.

[0066] In the embodiment provided by the present invention, Euclidean distance is used to quantify the impact of configuration structure changes.

[0067] The node change is specifically expressed as:

[0068] Edge variation, specifically expressed as:

[0069] Furthermore, the state change is specifically expressed as:

[0070] in, is the state change corresponding to the comprehensive directed graph from time t to time t+1, is the node change corresponding to the comprehensive directed graph from time t to time t+1, is the edge change corresponding to the comprehensive directed graph from time t to time t+1, is the second norm, is the attribute vector of node i at time t, For the edge The attribute vector at time t, E is the set of edges in the comprehensive directed graph, and n is the number of nodes in the comprehensive directed graph.

[0071] The state change threshold is determined based on the mean and standard deviation of the state change amount. The mean and standard deviation of the state change amount are calculated by obtaining the state change amount under the normal operation of the integrated energy system in the past. In one specific example, the state change threshold is obtained by summing the mean value with the product of a predetermined empirical coefficient and the standard deviation. In other embodiments, other methods may also be used, and this is not limited to this.

[0072] Specifically, the construction of the comprehensive energy operation model is determined by the following steps: Based on the initial deep learning model, combined with the comprehensive directed graph and the system action space, the expected cumulative reward values corresponding to different states and actions are obtained. The system action space includes various actions performed on each energy device and load in the comprehensive energy. Based on the expected cumulative reward values corresponding to different states and actions, the mean and variance of the expected cumulative reward values are given, and the state distribution of the expected cumulative reward values corresponding to different states and actions is obtained; According to the state distribution of the expected cumulative reward value, the model parameters of the integrated energy operation model are determined, and the construction of the integrated energy operation model is completed.

[0073] In the embodiment provided by the present invention, firstly, the comprehensive directed graph, the system action space, and the model parameters are initialized. , target model parameters And the experience replay buffer. Among them, the model parameters Used to estimate the Q-value function , target model parameters Used to stabilize the Q value function During the training process, the experience replay buffer is used to store the experience of the agent interacting with the environment, including state, action, reward, and next state. In this example, the Q value function is the expected cumulative reward value.

[0074] It's important to understand that in reinforcement learning, an agent is an entity that can perceive its environment and autonomously take actions to achieve specific goals. Agents learn optimal behavioral strategies through interaction with their environment. Agents are autonomous, able to operate without direct human intervention and make decisions based on internal logic. Agents possess perceptual capabilities, using sensors to detect environmental conditions. Agents possess decision-making capabilities, selecting the optimal action strategy based on perceived environmental information. Agents are adaptable, able to adjust their behavior based on environmental changes. Agents are interactive, able to interact with the environment and with other agents.

[0075] Then, experience collection is performed. At each time step , obtain the current operating status of the integrated energy system , select an action And execute, observe the next state and the reward function , and finally Stored to the experience replay buffer.

[0076] action The specific expression of is:

[0077] in, is the mean of the Q values, is the variance of the Q value, is a dynamic adjustment factor.

[0078] Among them, the above reward function , specifically obtained through the following steps: Based on the current operating status and dispatching operations of the integrated energy system, analyze the operating status and carbon emissions of the integrated energy system, and determine the system power weight coefficient, system energy storage weight coefficient, and system low-carbon weight coefficient; According to the connection relationship between each energy device and load in the comprehensive energy, the operation status and carbon emission status of each energy device and load are analyzed to determine the device power weight coefficient and the device energy storage weight coefficient; The reward function of comprehensive energy is determined by combining the carbon emission factors of various energy devices and loads and the operating conditions of various energy devices and loads in the comprehensive energy system at the current moment.

[0079] The above-mentioned system power weight coefficient, system energy storage weight coefficient and system low-carbon weight coefficient are pre-set based on the overall architecture of the integrated energy system, and the equipment power weight coefficient and equipment energy storage weight coefficient are pre-set based on different energy equipment and loads in the integrated energy system.

[0080] By constructing a comprehensive energy reward function, we quantify the merits of the current state and actions of the comprehensive energy system. The resulting reward function value guides the comprehensive energy operation model's regulation of the comprehensive energy system at the next moment. Within the comprehensive energy system, the reward function comprehensively considers factors such as power balance, energy utilization efficiency, equipment operating costs, and system stability. Based on the power generation and load power of each node, we consider the power balance of the comprehensive energy system, ensuring a balance between power generation and load power to avoid overloads or power shortages. We also consider the operating costs of the comprehensive energy system and minimize power generation and transmission costs. Based on the energy storage status of each energy storage device, we avoid frequent switching of device states and ensure stable system operation under various operations. We also consider the carbon emissions of different energy sources to maximize the utilization of renewable energy and reduce reliance on fossil fuels.

[0081] Reward Function It includes the integrated energy generation part, the integrated energy storage part and the integrated energy low-carbon operation part. The integrated energy generation part is determined by the following steps: Analyze the power generation and load power of various energy equipment and loads, and provide the power difference between the power generation and load power; Combined with the equipment power weight coefficients corresponding to various energy equipment and loads, the squares of the power differences are weighted and summed, and the system power weight coefficients of the comprehensive energy are integrated to obtain the comprehensive energy power generation part.

[0082] The comprehensive energy storage part is determined by the following steps: The energy storage status of various energy storage devices at the current moment is analyzed, and the absolute value of the energy storage status is weighted and summed in combination with the device energy storage weight coefficient corresponding to each type of energy storage device. The system energy storage weight coefficient of the comprehensive energy is integrated to obtain the comprehensive energy storage part.

[0083] The comprehensive energy low-carbon operation part is determined through the following steps: Analyze the power generation of various energy sources at the current moment, combine the carbon emission factors of various energy sources, perform weighted summation of the power generation, integrate the system low-carbon weight coefficient of comprehensive energy, and obtain the low-carbon operation part of comprehensive energy.

[0084] Furthermore, the reward function , specifically expressed as:

[0085] in, is the reward function value of comprehensive energy at time t, is the system power weight coefficient, is the equipment power weight coefficient of the i-th power generation equipment, is the system energy storage weight coefficient, is the energy storage weight coefficient of the j-th energy storage device, is the system low carbon weight coefficient, is the carbon emission factor of the kth energy type, is the power generation of the kth type of energy at time t, p is the total number of energy types, is the power generation of node i at time t, is the load power of node i at time t, is the energy storage state of node j at time t.

[0086] In the embodiment provided by the present invention, the exploration rate Selecting random actions and exploring unknown action spaces helps the agent try new strategies when facing unknown or uncertain environments, and avoid relying solely on the currently known optimal strategy. Select the best action, based on the current Q value and uncertainty, to select an action that is most likely to result in the maximum reward. As a basis for decision making, For action a in state The estimated value of is the variance of the action estimate. It will be adjusted dynamically over time. In the early stage, the agent will usually conduct more exploration (i.e. As learning progresses, the intelligent body will rely more on the knowledge it has learned. Increase, In this example, the initial value of the exploration rate can be set according to the actual situation, and then combined with the set decay rate, the exploration rate is adjusted to explore the unknown action space and find the optimal action.

[0087] In a specific example, in a microgrid, due to overcast weather, solar power generation is insufficient to meet demand. This could be increasing the output of wind turbines or discharging energy from battery storage to ensure load demand is met and system stability and reliability are ensured.

[0088] Dynamic exploration factor, specifically expressed as:

[0089] in, is the dynamic exploration factor at time t, is the initial exploration weight, is the influence coefficient of the configuration change of the integrated energy system on the dynamic exploration factor, It is the state change corresponding to the comprehensive directed graph from time t to time t+1.

[0090] After completing the experience collection, experience replay and network update are performed. Randomly sample a portion of experience samples from the experience replay buffer. , and calculate the target Q value of the sampled experience.

[0091]

[0092] in, is the target Q value corresponding to the j-th experience obtained by sampling, is the discount factor, For instant rewards, Indicates taking actions for all possible next states Take the maximum value, Take action for the next state of the jth experience The discount factor indicates the importance of future rewards and ranges from [0,1]. In this example, the discount factor ranges from [0.9,0.99].

[0093] Optimize model parameters through backpropagation and gradient descent And regularly update the target model parameters .

[0094]

[0095] in, is the learning rate, used to control model parameters Update step size, is the loss function About model parameters The gradient of the target model parameter Update once every fixed number of steps or fixed time steps to stabilize the training process of the comprehensive energy operation model.

[0096] During the training process of the integrated energy operation model, a system loss function is constructed to optimize the state distribution of model parameters, enabling the integrated energy operation model to more accurately fit the data and capture uncertainty, while avoiding overfitting without sacrificing the expressive power of the integrated energy operation model.

[0097] Get the system loss function, including: Analyze the differences in model parameters in the integrated energy operation model, analyze the changes in Q values corresponding to the model parameters, and construct a likelihood loss function; According to the prior distribution and approximate posterior distribution of the model parameters in the integrated energy operation model, the KL divergence of the prior distribution and the approximate posterior distribution is given, and the KL divergence loss function is obtained; The likelihood loss function and the KL divergence loss function are combined to obtain the system loss function.

[0098] The system loss function is specifically expressed as:

[0099] in, is the system loss function, is the likelihood loss function, is the function value of the KL divergence loss function KL(), which represents the prior distribution of model parameters in the integrated energy operation model and the approximate posterior distribution the Kullback-Leibler (KL) divergence between two probability distributions, is the regularization weight, is the absolute value of the experience sample set B, is the target Q value corresponding to the jth experience sample in the experience sample set B, In state and actions The Q value in this case is is the prior distribution of model parameters in the integrated energy operation model, is the approximate posterior distribution of the model parameters in the comprehensive energy operation model.

[0100] Likelihood loss function is used for fitting The KL divergence loss function is used to constrain the state distribution of model parameters, measure the difference between two probability distributions, and compare the similarity of two probability distributions.

[0101] In other embodiments, when there is an error in the integrated directed graph, first, the integrated directed graph is updated based on the new topological structure. and the system action space A, ensuring that all newly added or deleted nodes and edges are correctly represented in the state vector and action set. Secondly, the experience replay buffer is updated. The experience replay buffer is cleared or partially cleared to prevent data from the old topology from interfering with the learning of the new strategy. New operating data of the integrated energy system is collected. Under the new topology, the system interacts with the environment again, gradually filling the new experience replay buffer. Finally, training is performed using the new experience data. The process of experience collection, experience replay, and network update continues, using the new data to optimize the integrated energy operation model and adapt it to the new topology to achieve optimal control of the integrated energy system.

[0102] S104: Based on the target operation model, obtain real-time operation data of the comprehensive energy.

[0103] It is understood that based on the actual needs of the integrated energy system, namely the system data to be configured, a corresponding target configuration scheme is obtained. Based on the target configuration scheme, the configuration structure of the integrated energy system can be obtained to complete the construction of the integrated energy system. The integrated energy system then actually operates based on the integrated energy operation model. During operation, the configuration structure of the integrated energy system changes. For example, if problems occur in certain lines or certain equipment is damaged, the integrated energy operation model is adjusted based on the changes in the configuration structure to provide a target operation model. The integrated energy system continues to operate according to the operation strategy provided by the target operation model. The target operation model is the basis for the current operation strategy of the integrated energy system, and the integrated energy operation model is the basis for the previous operation strategy. When the configuration structure of the integrated energy system does not change, the integrated energy operation model and the target operation model are the same.

[0104] Therefore, when the integrated energy system operates according to the operational strategy specified by the target operation model, real-time operational data for the integrated energy system can be obtained. This real-time operational data includes the power generation, load power, energy storage status, and operational status of various energy devices and loads within the integrated energy system, as well as energy production data (such as the output of various energy sources and the operating parameters of production equipment), energy consumption data (such as energy consumption and consumption patterns by user or region), energy conversion data (such as conversion efficiency and losses of conversion equipment), storage data (such as the storage capacity of energy storage equipment and charging and discharging parameters), and environmental data (such as ambient temperature and humidity).

[0105] S105: Combined with the pre-built evaluation index library, feature extraction is performed on the real-time operation data to obtain target evaluation features.

[0106] Reference Figure 4 , specifically including: Based on the pre-built comprehensive energy indicator library, the corresponding initial assessment features are screened from the real-time operation data; The principal component analysis method is used to analyze the initial evaluation features and give the core features; The target evaluation features are determined based on the similarity between each initial evaluation feature and the core feature.

[0107] Among them, the above-mentioned comprehensive energy index database collects historical operating data of the comprehensive energy system and conducts in-depth analysis using data mining technology and statistical analysis methods; by analyzing the fluctuation patterns of energy supply and consumption in different seasons and under different operating conditions, key parameters closely related to system performance are extracted as indicators; cluster analysis and other methods are used to identify high-incidence patterns of energy equipment failures, and relevant characteristics are incorporated into the equipment operation index system; and successful evaluation cases of similar comprehensive energy systems at home and abroad are widely studied to select indicators.

[0108] By building a comprehensive energy indicator library, we can initially screen out evaluation features related to the operation of the integrated energy system from a large amount of operational data. This initial screening reduces the computational complexity of subsequent core feature analysis, improving the efficiency of the evaluation process while ensuring the accuracy of the integrated energy system assessment.

[0109] Furthermore, principal component analysis was used to analyze the initial evaluation features and provide core features, including: Analyze the linear correlation between each initial evaluation feature and construct the covariance matrix of the initial evaluation features; The covariance matrix is decomposed to give multiple eigenvalues; According to the variance explanation ratio of each eigenvalue and the preset explanation threshold, multiple eigenvalues are screened to obtain the core features.

[0110] Furthermore, the linear correlation between each initial evaluation feature is analyzed, and the covariance matrix of the initial evaluation features is constructed, which specifically includes: According to the elements in each initial evaluation feature and the feature mean of the corresponding initial evaluation feature, the deviation feature of each initial evaluation feature is given; Based on the deviation characteristics of any two initial evaluation characteristics, the mean of the product of the two deviation characteristics is given to obtain the degree of linear correlation between the initial evaluation characteristics; The linear correlation degree between each initial evaluation feature is integrated to construct the covariance matrix of the initial evaluation features.

[0111] Specifically, the degree of linear correlation between the initial evaluation features is expressed as:

[0112] in, is the linear correlation degree between the u-th initial evaluation feature and the v-th initial evaluation feature, m is the number of samples, is the kth eigenvalue in the uth initial evaluation feature, is the feature mean of the u-th initial evaluation feature, is the kth eigenvalue in the vth initial evaluation feature, is the feature mean of the vth initial evaluation feature.

[0113] Deviation features refer to the matrix composed of the differences between each eigenvalue and the feature mean in the initial evaluation features.

[0114] The covariance matrix is specifically expressed as:

[0115] Where C is the covariance matrix, is the matrix composed of the standardized initial evaluation features, m is the number of samples, Transpose the matrix.

[0116] Furthermore, multiple eigenvalues are specifically expressed as:

[0117] Where C is the covariance matrix of the initial evaluation features, is a diagonal matrix, and each element on the diagonal of the diagonal matrix is the eigenvalue. is the identity matrix, is the transposed matrix of V, V is the eigenvector matrix, and the elements in the eigenvector matrix are the eigenvectors corresponding to the eigenvalues.

[0118] In a specific embodiment, the covariance matrix C is subjected to eigendecomposition to obtain the eigenvalues and the corresponding eigenvector The eigendecomposition is specifically expressed as ,in, is a diagonal matrix whose diagonal elements are eigenvalues , and arranged in descending order. V is the eigenvector The matrix composed of .

[0119] Furthermore, based on the variance explanation ratio of each eigenvalue and the preset explanation threshold, multiple eigenvalues are screened to obtain core features, including: According to the variance explanation ratio of each eigenvalue, the cumulative variance explanation ratio corresponding to each eigenvalue is given; Compare the cumulative variance explanation ratio corresponding to each eigenvalue with the preset explanation threshold to obtain the comparison result; According to the comparison results, the eigenvalues are screened and the eigenvectors corresponding to the screened eigenvalues are given; The standard data matrix is projected into a vector matrix composed of eigenvectors corresponding to the filtered eigenvalues to obtain the core features.

[0120] In a specific implementation, the eigenvalues are arranged in descending order, and the core features are selected based on the cumulative variance explained ratio.

[0121] Calculate the cumulative proportion of variance explained for each eigenvalue:

[0122] in, is the cumulative variance explained by the k-th eigenvalue, is the proportion of variance explained by the i-th eigenvalue, is the i-th eigenvalue, and m is the number of eigenvalues.

[0123] In a specific example, the preset interpretation threshold is , select The minimum value k is used as the number of principal components. After selecting k principal components, the standardized data matrix Projection to the matrix consisting of the first k eigenvectors On the top, get the core features , specifically expressed as:

[0124] in, is composed of the first k eigenvectors Matrix, core features for The matrix, core features Elements in represents the score of the i-th sample on the j-th principal component.

[0125] Specifically, based on the similarity between the initial evaluation features and the core features, combined with the preset similarity range, the initial evaluation features are screened to give the target evaluation features; Combined with the information content of the target evaluation feature, the evaluation weight corresponding to the target evaluation feature is given.

[0126] In the embodiment provided by the present invention, the similarity between each initial evaluation feature and the core feature is represented by the Pearson correlation coefficient. In other embodiments, other methods can be used to determine the similarity between each evaluation feature and the core feature, which is not limited to this.

[0127] In the embodiments provided by the present invention, a similarity range is set to determine whether the similarity is within the preset similarity range. If so, the initial evaluation feature is used as the target evaluation feature. Conversely, if the similarity is not within the preset similarity range, the initial evaluation feature is eliminated and not used as the target evaluation feature for evaluating the integrated energy system. The target evaluation features obtained after screening are closely related to the core features, can deeply reflect the key characteristics and operational nature of the integrated energy system, effectively avoid interference from irrelevant or weakly related indicators, significantly improve the pertinence and effectiveness of the evaluation, and ensure that the evaluation focuses on core elements and accurately understands system performance.

[0128] S106: The evaluation weights corresponding to the target evaluation features are integrated to provide the comprehensive energy operation evaluation results.

[0129] Among them, reference Figure 5 , the evaluation weight corresponding to the target evaluation feature is obtained through the following steps: According to the value of each element in the target evaluation feature, the information content of the target evaluation feature is given; Combined with the information content of the target evaluation features, the degree of variation of the target evaluation features is analyzed; Based on the degree of variation of all target evaluation features and combined with the degree of variation of the target evaluation features, the evaluation weight corresponding to the target evaluation features is given.

[0130] Furthermore, the evaluation weight corresponding to the target evaluation feature is specifically expressed as:

[0131] in, is the evaluation weight corresponding to the j-th target evaluation feature, is the information content of the jth target feature weight, m is the number of target evaluation features, Evaluate the degree of variation of features for the target.

[0132] In the embodiment provided by the present invention, the entropy weight method is used to set the evaluation weights. The information entropy of each target evaluation feature is calculated to reflect the information amount of each target evaluation feature. The weights are allocated entirely based on the intrinsic information characteristics of the data to ensure the objectivity and scientificity of the weight setting.

[0133] Information entropy reflects the amount of information contained in the target evaluation feature. The larger the information entropy of the target evaluation feature, the smaller the degree of variation of the target evaluation feature, the less information it provides, and the smaller its role in the comprehensive evaluation. Conversely, the smaller the information entropy of the target evaluation feature, the greater the degree of variation of the target evaluation feature, the more information it provides, and the greater the weight should be.

[0134] The evaluation weights are determined by the information entropy of the target evaluation characteristics, so that the weight distribution is completely based on the information characteristics of the data itself, avoiding the interference of human factors, ensuring the objectivity and scientific nature of the weight setting, and making the evaluation results more convincing and reliable.

[0135] After determining the evaluation weights, the target evaluation features and their corresponding evaluation weights are weighted and summed to obtain the comprehensive energy operation evaluation results.

[0136] S107: According to the operation evaluation results, adjust the target operation model to complete the configuration of comprehensive energy.

[0137] Further, refer to Figure 6 ,The target operation model includes a reward function, where the reward function is composed of an integrated energy generation part, an integrated energy storage part, and an integrated energy low-carbon operation part; Based on the results of the operation evaluation, the target operation model is adjusted to complete the configuration of comprehensive energy, including: If the operation evaluation result is not up to standard, analyze the sub-scores of each part of the operation evaluation result; Analyze the compliance of each sub-score and adjust the coefficients of the corresponding parts of the reward function in the target operation model based on the compliance; Based on the adjusted reward function, the target operation model is trained until convergence; Run the model according to the target after training, run the comprehensive energy, and complete the configuration of the comprehensive energy.

[0138] It can be understood that the target operation model described above is obtained by adjusting and training the integrated energy operation model. Since the integrated energy operation model uses a reward function during its training and adjustment process, adjustments to the target operation model also utilize the reward function. Based on this, various components of the reward function are adjusted based on the evaluation results. Based on the adjusted reward function, the target operation model is further adjusted and trained to obtain a new target operation model. The integrated energy system will operate according to the operating strategy specified by the new target operation model, generating new real-time operating data and then new evaluation results. This process repeats until the evaluation results meet the requirements of the integrated energy system, i.e., the evaluation results meet the standards.

[0139] In one specific embodiment, the operational evaluation result ranges from 0 to 1 and is graded based on the performance of the integrated energy system. For example, five grades can be set: excellent (operational evaluation result between 0.8 and 1), good (operational evaluation result between 0.6 and 0.8), medium (operational evaluation result between 0.4 and 0.6), poor (operational evaluation result between 0.2 and 0.4), and poor (operational evaluation result between 0 and 0.2). This grading helps to intuitively understand the relative position and performance level of the integrated energy system among similar systems. The basis for the different grades can refer to industry averages, performance data of advanced benchmark systems, and relevant policies, standards, and regulations. In the examples provided in this application, excellent and good are considered to meet the standards, while medium, poor, and poor are considered to fail to meet the standards. The operational evaluation result is an evaluation of the overall performance of the integrated energy system. If the operational evaluation result meets the standards, but the parameter settings of some equipment in the integrated energy system do not meet the standards, the overall results will be used as the guide, and no adjustments will be made to the small non-compliant parts. If the operation evaluation result is not up to standard, it means that there is a problem with the overall operation configuration of the integrated energy. At this time, it is necessary to analyze which part does not meet the standard, adjust the reward function according to the non-compliance situation, and give a new target operation model to complete the configuration of the integrated energy.

[0140] As can be seen from the aforementioned analysis process, the operational data of the integrated energy system can be divided into three parts: data related to power generation equipment, data related to energy storage equipment, and data reflecting the carbon emissions of power generation equipment, energy storage equipment, and various energy sources. In other words, each target assessment feature used to determine the operational evaluation result can also be divided into at least one of these three parts. In a specific example, when the operational evaluation result is substandard, the compliance of each target assessment feature is analyzed. Each target assessment feature has a corresponding compliance analysis table. If the value corresponding to the target assessment feature falls within the substandard range, the target assessment feature is substandard. Conversely, if the value corresponding to the target assessment feature falls within the standard range, the target assessment feature is standard. The standard compliance analysis table corresponding to each target assessment feature is pre-set and varies across different integrated energy systems, so this is not a limitation. Each substandard target assessment feature is analyzed to determine whether it belongs to the power generation equipment, energy storage equipment, or carbon emissions component. Based on the number or proportion of each substandard target assessment feature in the corresponding component, the adjustment parameter corresponding to the reward function is determined. In different implementations, the above-mentioned sub-scores can be the characteristic values of each target evaluation feature, the number or proportion of each non-compliant target evaluation feature in the corresponding part, and there is no limitation on this. For example, if there are only 5 non-compliant target evaluation features belonging to power generation equipment, the corresponding system power weight coefficient in the reward function will be adjusted upward. For another example, there are a total of 10 target evaluation features belonging to power generation equipment, and 5 of them do not meet the standard, and the proportion of non-compliant ones is 0.5. According to the adjustment parameter corresponding to 0.5, the system power weight coefficient in the reward function is adjusted. If the non-compliant target evaluation features involve multiple parts, the coefficients corresponding to different parts are adjusted separately. The correspondence between the number or proportion of target evaluation features in different parts and the corresponding adjustment parameters is pre-set according to the actual scenario.

[0141] Based on the results of the operational assessment, specific decision-making recommendations are provided for the optimization and development of the integrated energy system. In terms of energy supply, if the supply stability is insufficient, consideration can be given to adding backup energy supply equipment, optimizing energy scheduling strategies, or strengthening interconnection with surrounding energy systems to improve supply reliability. In the energy consumption link, if the energy consumption structure is unreasonable and over-reliant on a certain type of highly polluting or high-cost energy, the transition from energy-saving equipment to clean energy can be promoted. For equipment operation, if the equipment failure rate is high, it is recommended to increase investment in equipment maintenance, establish an equipment life cycle management system, and plan equipment upgrades in advance. In terms of environmental indicators, if carbon emissions exceed the standard, measures such as promoting the construction of renewable energy projects and implementing carbon capture and storage technology pilots can be taken to reduce the environmental impact of the system and achieve the sustainable development of the integrated energy system.

[0142] Reference Figure 7The embodiment of the present invention provides a comprehensive energy configuration device, which adopts any one of the above comprehensive energy configuration methods, including: Perception module, used to obtain system data to be configured; Storage module, used to store historical configuration solution library; The calculation module is used to analyze the data of the system to be configured based on the data similarity between the system data to be configured and the various historical configuration schemes in the historical configuration scheme library, and adjust the historical configuration schemes to provide a target configuration scheme; according to the target configuration scheme, the configuration structure of the comprehensive energy is obtained, and the configuration structure changes of the comprehensive energy are obtained in real time, and the comprehensive energy operation model is adjusted to obtain the target operation model; based on the target operation model, the real-time operation data of the comprehensive energy is obtained; in combination with the pre-built evaluation index library, the real-time operation data is extracted to obtain the target evaluation features; the evaluation weights corresponding to the target evaluation features are integrated to provide the operation evaluation results of the comprehensive energy; according to the operation evaluation results, the target operation model is adjusted to complete the configuration of the comprehensive energy.

[0143] In addition, the present invention also provides a comprehensive energy configuration system, including: a network device, an execution device, and the comprehensive energy configuration device as described above.

[0144] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0145] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment or part of code, which contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the boxes can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can be implemented using a dedicated hardware-based system that performs the specified functions or operations, or can be implemented using a combination of dedicated hardware and computer instructions.

[0146] The units described in the embodiments of the present disclosure may be implemented in software or hardware, wherein the name of a unit does not necessarily limit the unit itself.

[0147] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the invention. Thus, the present invention is intended to include such changes and modifications as fall within the scope of the claims and their equivalents.

Claims

1. A method for configuring comprehensive energy, characterized in that: include: Get the system data to be configured; Based on the data similarity between the system data to be configured and the historical configuration solutions in the historical configuration solution library, the system data to be configured is analyzed, and the historical configuration solutions are adjusted to provide the target configuration solution. According to the target configuration plan, the configuration structure of the comprehensive energy is obtained, and the changes in the configuration structure of the comprehensive energy are obtained in real time, and the comprehensive energy operation model is adjusted to obtain the target operation model; Based on the target operation model, obtain the real-time operation data of comprehensive energy; Combined with the pre-built evaluation index library, feature extraction is performed on real-time operation data to obtain target evaluation features; The evaluation weights corresponding to the target evaluation characteristics are integrated to provide the comprehensive energy operation evaluation results; According to the operation evaluation results, the target operation model is adjusted to complete the configuration of comprehensive energy.

2. The method for configuring comprehensive energy according to claim 1, wherein: The data similarity between the system data to be configured and each historical configuration scheme in the historical configuration scheme library is obtained through the following steps: Based on the type of system data to be configured, a similarity index analysis strategy is matched, and the local similarity of the historical data corresponding to each historical configuration solution is analyzed separately; Combining the similarity weights corresponding to different types, the local similarities corresponding to each historical configuration scheme are fused to obtain the data similarity corresponding to each historical configuration scheme.

3. The method for configuring comprehensive energy according to claim 2, wherein: The types of data to be configured include discrete and continuous types, and the local similarity includes discrete similarity and continuous similarity; Based on the type of system data to be configured, a similarity index analysis strategy is matched and the local similarity of the historical data corresponding to each historical configuration solution is analyzed separately, including: Using a first similarity index, analyzing discrete similarities between discrete system data to be configured and corresponding discrete data in historical configuration solutions; The second similarity index is used to analyze the continuous similarity between the continuous system data to be configured and the corresponding continuous data in the historical configuration scheme.

4. The method for configuring comprehensive energy according to claim 3, wherein: Similarity weights include discrete similarity weights and continuous similarity weights; Combining the similarity weights corresponding to different types, the local similarities corresponding to each historical configuration scheme are fused to obtain the data similarity corresponding to each historical configuration scheme, specifically including: Normalize the discrete similarity and continuous similarity respectively to obtain the standard discrete similarity and standard continuous similarity; Combining the discrete similarity weight and the continuous similarity weight, the standard discrete similarity and the standard continuous similarity are weighted and summed to obtain the data similarity.

5. The method for configuring comprehensive energy according to claim 1, wherein: Based on the data similarity between the system data to be configured and the historical configuration solutions in the historical configuration solution library, the system data to be configured is analyzed, and the historical configuration solutions are adjusted to provide the target configuration solution, including: Sort the historical configuration plans in the historical configuration plan library according to data similarity, and give the pending historical plans based on the sorting results; Based on a pre-acquired configuration mapping function, the system data to be configured is mapped to obtain a target configuration solution, wherein the pre-acquired configuration mapping function is obtained by analyzing and correcting a historical solution to be determined.

6. The method for configuring comprehensive energy according to claim 1, wherein: According to the target configuration plan, the configuration structure of comprehensive energy is obtained, which specifically includes: From the target configuration plan, the various energy devices and loads in the integrated energy are obtained, and the nodes of the integrated directed graph are determined; Analyze the operating status of each energy device and load in the comprehensive energy system, and determine the attributes of each node in the comprehensive directed graph; Combined with the connection relationship and connection status of each energy device and load in the comprehensive energy, determine the edges and edge attributes corresponding to each node in the comprehensive directed graph; By integrating the nodes, attributes of the nodes, edges corresponding to the nodes and edge attributes of the comprehensive directed graph, the construction of the comprehensive directed graph is completed and the configuration structure of the comprehensive energy is obtained.

7. The method for configuring comprehensive energy according to claim 6, wherein: Obtain the configuration structure changes of the integrated energy in real time, adjust the integrated energy operation model, and obtain the target operation model, including: Based on the configuration structure change of comprehensive energy, the configuration structure change is quantified to give the state change of comprehensive energy; According to the state change threshold obtained in advance, the state change amount is judged and a dynamic exploration factor is given. The dynamic exploration factor is used to influence the action selection of the integrated energy at the next moment; Based on the dynamic exploration factor and combined with the system loss function, the comprehensive energy operation model is trained until convergence to obtain the target operation model.

8. The method for configuring comprehensive energy according to claim 7, wherein: Based on the configuration structure changes of comprehensive energy, the changes in the configuration structure are quantified and the state changes of comprehensive energy are given, including: Obtain a comprehensive directed graph at two adjacent moments, analyze the changes in the number of nodes and the number of edges in the comprehensive directed graph corresponding to the two adjacent moments, and obtain the change results; Based on the change results, analyze the changes of each node and edge in the comprehensive directed graph at two adjacent moments to obtain the node change amount and edge change amount; The node change and edge change are combined to obtain the state change.

9. The method for configuring comprehensive energy according to claim 6, wherein: The construction of the comprehensive energy operation model is determined by the following steps: Based on the initial deep learning model, combined with the comprehensive directed graph and the system action space, the expected cumulative reward values corresponding to different states and actions are obtained. The system action space includes various actions performed on each energy device and load in the comprehensive energy. Based on the expected cumulative reward values corresponding to different states and actions, the mean and variance of the expected cumulative reward values are given, and the state distribution of the expected cumulative reward values corresponding to different states and actions is obtained; According to the state distribution of the expected cumulative reward value, the model parameters of the integrated energy operation model are determined, and the construction of the integrated energy operation model is completed.

10. The method for configuring comprehensive energy according to claim 1, wherein: Combined with the pre-built evaluation index library, feature extraction is performed on real-time operation data to obtain target evaluation features, including: Based on the pre-built comprehensive energy indicator library, the corresponding initial assessment features are screened from the real-time operation data; The principal component analysis method is used to analyze the initial evaluation features and give the core features; The target evaluation features are determined based on the similarity between each initial evaluation feature and the core feature.

11. The method for configuring comprehensive energy according to claim 10, wherein: The principal component analysis method is used to analyze the initial evaluation features and give the core features, including: Analyze the linear correlation between each initial evaluation feature and construct the covariance matrix of the initial evaluation features; The covariance matrix is decomposed to give multiple eigenvalues; According to the variance explanation ratio of each eigenvalue and the preset explanation threshold, multiple eigenvalues are screened to obtain the core features.

12. The method for configuring comprehensive energy according to claim 11, wherein: Analyze the linear correlation between each initial evaluation feature and construct the covariance matrix of the initial evaluation features, including: According to the elements in each initial evaluation feature and the feature mean of the corresponding initial evaluation feature, the deviation feature of each initial evaluation feature is given; Based on the deviation characteristics of any two initial evaluation characteristics, the mean of the product of the two deviation characteristics is given to obtain the degree of linear correlation between the initial evaluation characteristics; The linear correlation degree between each initial evaluation feature is integrated to construct the covariance matrix of the initial evaluation features.

13. The method for configuring comprehensive energy according to claim 11, wherein: Based on the variance explanation ratio of each eigenvalue and the preset explanation threshold, multiple eigenvalues are screened to obtain core features, including: According to the variance explanation ratio of each eigenvalue, the cumulative variance explanation ratio corresponding to each eigenvalue is given; Compare the cumulative variance explanation ratio corresponding to each eigenvalue with the preset explanation threshold to obtain the comparison result; According to the comparison results, the eigenvalues are screened and the eigenvectors corresponding to the screened eigenvalues are given; The standard data matrix is projected into a vector matrix composed of eigenvectors corresponding to the filtered eigenvalues to obtain the core features.

14. A comprehensive energy configuration device, characterized in that: The method for configuring the integrated energy according to any one of claims 1 to 13 comprises: Perception module, used to obtain system data to be configured; Storage module, used to store historical configuration solution library; The calculation module is used to analyze the data of the system to be configured based on the data similarity between the system data to be configured and the various historical configuration schemes in the historical configuration scheme library, and adjust the historical configuration schemes to provide a target configuration scheme; according to the target configuration scheme, the configuration structure of the comprehensive energy is obtained, and the configuration structure changes of the comprehensive energy are obtained in real time, and the comprehensive energy operation model is adjusted to obtain the target operation model; based on the target operation model, the real-time operation data of the comprehensive energy is obtained; in combination with the pre-built evaluation index library, the real-time operation data is extracted to obtain the target evaluation features; the evaluation weights corresponding to the target evaluation features are integrated to provide the operation evaluation results of the comprehensive energy; according to the operation evaluation results, the target operation model is adjusted to complete the configuration of the comprehensive energy.

15. A comprehensive energy configuration system, characterized in that: include: A network device, an execution device, and a configuration device for the integrated energy as claimed in claim 14.

Images