Microgrid load forecasting optimization algorithm based on deep learning

Abstract

The present invention relates to the technical field of power grid load prediction and optimization, and in particular to a microgrid load prediction optimization algorithm based on deep learning. The algorithm comprises the following steps: using sensors to collect real-time operation data of a microgrid, dividing the microgrid into different types of nodes based on the collected real-time operation data, and constructing a microgrid topology diagram based on the connection relationship between all nodes; constructing a node load sequence feature and a structural embedding feature for each node based on an input historical load data set and a microgrid topology diagram, and performing operation load prediction of the node based on the influence weight between the nodes; based on the prediction result of the node, the prediction output of the adjacent nodes is integrated to perform collaborative correction of the prediction value; at the same time, according to the error between the prediction output of the adjacent nodes and the actual measurement value, the influence weight between the nodes is dynamically adjusted, and the microgrid load prediction is performed based on the adjusted influence weight.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid load forecasting and optimization, and in particular to a microgrid load forecasting optimization algorithm based on deep learning. Background Art

[0002] With the rapid expansion of distributed energy access and microgrids, power grid systems are placing higher demands on node-level load forecasting accuracy and dynamic response capabilities. However, existing load forecasting methods primarily rely on traditional statistical models (such as ARIMA), shallow machine learning methods (such as SVR and RF), or RNN / LSTM-based sequential neural networks. While these methods can fit load sequences to a certain extent, they generally suffer from the following key technical bottlenecks:

[0003] The electrical connection relationship between microgrid nodes cannot be effectively integrated, and the impact of power flow and load conduction paths between nodes on the current node load evolution is ignored;

[0004] The prediction model cannot be flexibly adjusted according to the behavioral characteristics of different node types (such as power generation, load, and energy storage nodes), and lacks the ability to model structural input features;

[0005] In actual engineering deployment, there is a lack of mechanisms to effectively identify and suppress negative influencing factors in the graph structure, which makes the prediction error susceptible to structural noise. Summary of the Invention

[0006] In view of the above-mentioned problems, the present invention is proposed.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: a microgrid load prediction optimization algorithm based on deep learning, comprising the following steps:

[0008] Use sensors to collect real-time operation data of the microgrid, and based on the collected real-time operation data, divide the microgrid into different types of nodes, and construct a microgrid topology diagram based on the connection relationship between all nodes;

[0009] The influence weights between nodes are analyzed through deep learning algorithms, and the operating load of nodes is predicted using historical loads. Specifically:

[0010] Based on the input historical load data set and microgrid topology diagram, the node load sequence characteristics and structural embedding characteristics are constructed for each node, and the node operating load is predicted based on the influence weights between nodes;

[0011] Based on the influence weights between nodes in the microgrid topology and the load prediction results for the nodes, the global load of the microgrid is collaboratively optimized and predicted, specifically:

[0012] According to the prediction results of the node, the prediction output of the adjacent nodes is integrated to perform collaborative correction of the prediction value. At the same time, according to the error between the prediction output of the adjacent nodes and the actual measurement value, the influence weights between the nodes are dynamically adjusted, and the microgrid load prediction is performed based on the adjusted influence weights.

[0013] As a preferred solution of the microgrid load forecasting optimization algorithm based on deep learning described in the present invention, the method of using sensors to collect real-time operation data of the microgrid is as follows:

[0014] Calculate the active power and reactive power corresponding to the current node, and at the same time, determine the flow direction of the current node, then we have,

[0015] Active power,

[0016]

[0017] reactive power,

[0018]

[0019] in, 、 Respectively represent the voltage data and current data collected at the current node, 、 Indicates the current phase angle and voltage phase angle corresponding to the current node, Indicates the active power corresponding to the current node, Indicates the reactive power corresponding to the current node.

[0020] As a preferred solution of the microgrid load forecasting optimization algorithm based on deep learning described in the present invention, the flow direction of the current node is determined as follows:

[0021] For the current node, select the adjacent nodes of the current node, calculate the active power flow and reactive power flow from the current node to the adjacent nodes respectively, and judge the power flow direction of the current node based on the calculation results, specifically:

[0022] Active power from the current node to the adjacent node,

[0023]

[0024] Reactive power from the current node to the adjacent node,

[0025]

[0026] in, Indicates the voltage data of the current node, Represents the voltage data of adjacent nodes, represents the voltage phase angle of the current node, represents the voltage phase angle of adjacent nodes, represents the impedance between the current node and the adjacent node, Indicates the active power flow from the current node to the adjacent node, Indicates the reactive power flow from the current node to the adjacent node;

[0027] According to the calculation results of active power flow and reactive power flow, the flow direction of the current node is determined, then,

[0028] If the active power flow of the current node satisfies the formula , indicates that the power of the current node flows from the current node to the adjacent node, and vice versa, it means that the power flows from the adjacent node to the current node;

[0029] If the reactive power flow of the current node satisfies the formula , indicating that the reactive power of the current node flows from the current node to the adjacent node, and vice versa, it means that the reactive power flows from the adjacent node to the current node.

[0030] As a preferred solution of the microgrid load forecasting optimization algorithm based on deep learning described in the present invention, the microgrid is divided into different types of node types as follows:

[0031] If the active power corresponding to the current node satisfies the formula , and the current node's flow direction is outflowing from the current node, which means the current node type is a power generation node;

[0032] If the active power corresponding to the current node satisfies the formula , and the reactive power satisfies the formula ,At the same time, if the current node's flow direction is from the adjacent node to the current node, it means that the current node type is a load node;

[0033] For nodes whose node type is energy storage node, the judgment is based on the charge state of the node, specifically:

[0034] Check whether the current node has capacitance parameters. If the current node cannot collect capacitance parameters, it means that the current node is not an energy storage node.

[0035] If the current node can collect capacitance parameters, it means that the current node type is an energy storage node.

[0036] As a preferred solution of the microgrid load forecasting optimization algorithm based on deep learning described in the present invention, the influence weights between nodes analyzed by the deep learning algorithm are specifically as follows:

[0037] For any node in the microgrid topology , extract the node features corresponding to the node , through the convolutional neural network model to the node The running status of the ;

[0038] For nodes The results of the initial analysis , choose non- Node of Node

[0039] , and perform feature splicing on the two node features, then we have,

[0040]

[0041] in, Representation node The extracted node features, Representation node The extracted node features, Represents the concatenated features, which are used for in-depth analysis of the influence weights between nodes. Then,

[0042] According to the spliced features, the state analysis is performed through the convolutional neural network model, and the analysis results are set as , comparing the two analysis results as well as , based on the comparison results Conduct an in-depth analysis of the influence weights between nodes, specifically:

[0043] If the comparison result satisfies the formula , then it means the node Relative to the node The impact weight is negative;

[0044] If the comparison result satisfies the formula , then it means the node Relative to the node The impact weight is positive;

[0045] If the comparison result satisfies the formula , then it means the node Relative to the node The influence weight is no influence.

[0046] As a preferred solution of the microgrid load forecasting optimization algorithm based on deep learning described in the present invention, the operating load of the node predicted by using historical load is specifically as follows:

[0047] For each node, based on its historical load sequence, the length of the extraction time window is Sliding sequence and construct the historical state matrix;

[0048] Define node type labels and construct an input sample set. Use a graph convolutional network to extract topological features between nodes and perform time series modeling on the extracted topological features to generate initial prediction results.

[0049] Based on the generated initial prediction result set, the prediction results are analyzed and the prediction error is calculated. Based on the prediction error, the graph embedding method is readjusted to generate the second generation prediction value.

[0050] According to the second-generation prediction value set, by introducing a fitting scoring mechanism, the comparison result between the fitting score corresponding to the second-generation prediction value and the fitting score threshold is judged, the adjacency structure is readjusted, and the node load prediction is performed based on the reconstructed adjacency structure.

[0051] As a preferred solution of the microgrid load forecasting optimization algorithm based on deep learning described in the present invention, the generating of the initial forecast result is specifically as follows:

[0052] Input the constructed joint feature tensor into the graph convolution layer to extract the node graph embedding, then we have,

[0053]

[0054] in, represents the normalized adjacency matrix, represents the graph convolution weight, represents the activation function, represents the constructed joint feature tensor, represents the extracted node graph embedding features;

[0055] Embedding graphs into features Input into the LSTM network and obtain the hidden state output, then we have,

[0056]

[0057] in, Representation node The corresponding graph embedding features, Representation node The corresponding hidden state output;

[0058] Based on the hidden state output, generate the initial prediction result, then we have,

[0059]

[0060] in, 、 Represent the weight coefficient and bias term of the initial prediction respectively, Representation node The corresponding hidden state output is, Representation node Initial load forecast results.

[0061] As a preferred solution of the microgrid load forecasting optimization algorithm based on deep learning described in the present invention, the load forecasting of the node is specifically as follows:

[0062] Comparing the fitting scores between the second generation prediction value and the target load curve, we have:

[0063]

[0064] in, Representation node exist The actual load measurement value at the moment, Representation node exist The second generation prediction value at time , Indicates the length of the time window, Represents the fitting score of the second-generation prediction value, which is used to reconstruct the adjacency structure, specifically:

[0065] Setting the Fit Score Threshold , if the calculated fit score satisfies the formula , it means that the adjacency matrix needs to be readjusted to reconstruct the adjacency structure until the fitting score is higher than the set score threshold. Otherwise, it means that there is no need to readjust the adjacency structure.

[0066] For the adjusted prediction model, regenerate the third-generation prediction value according to the prediction method of the second-generation prediction value, and calculate the fitting score corresponding to the third-generation prediction value , compare the fitting score corresponding to the third-generation prediction value with the fitting score corresponding to the second-generation prediction value until the comparison result satisfies the formula So far, it means that the load forecast of the node is accurate, and the third-generation prediction value is used as the load forecast result of the node.

[0067] As a preferred solution of the microgrid load forecasting optimization algorithm based on deep learning described in the present invention, the collaborative correction of the predicted value is specifically as follows:

[0068] For nodes The prediction result is based on its adjacent node set , integrate the prediction output of adjacent nodes and make collaborative corrections to the prediction values, then we have,

[0069]

[0070] in, Representation node The predicted value after integrating the adjacent nodes, Representation node The third generation prediction value of Representation node The third generation prediction value of Representation node With node The influence weight between Representation node The adjacent node set of the node is the topological structure. A set of nodes that can form edge relationships between them. represents the synergy correction coefficient, specifically:

[0071] If the node The prediction result of the adjacent node set is compared with the prediction result of the adjacent node set to satisfy the formula , indicating that the node prediction is disturbed by the adjacent nodes, and the correction coefficient is increased to achieve the correction of the adjacent nodes.

[0072] As a preferred solution of the microgrid load forecasting optimization algorithm based on deep learning of the present invention, the microgrid load forecasting based on the adjusted influence weight is specifically as follows:

[0073] For nodes The predicted value after integrating adjacent nodes The error between the actual measured value and the weight adjustment factor between each pair of nodes is calculated, then:

[0074]

[0075] in, represents the adjustment coefficient, Representation node The predicted value after integrating the adjacent nodes, Representation node The third generation prediction value of node Actual measurement results of load, Representation node With node The weight adjustment factor between nodes is used to achieve dynamic adjustment of the weights between nodes, then,

[0076]

[0077] in, Representation node With node The influence weight between Representation node With node The weight adjustment factor between Representation node With node The adjusted impact weights between

[0078] Repeat the collaborative correction of the predicted value by the integrated adjacent nodes and the dynamic adjustment process of the weight until the calculated weight adjustment factor is less than the set weight adjustment factor threshold. At this time, the prediction result set generated in the current iteration is:

[0079]

[0080] in, represents the optimization result of microgrid load forecasting, represents the load forecast optimization result of the first node, Indicates the The load forecast optimization results of each node.

[0081] Beneficial effects of the present invention:

[0082] By constructing a topology-time joint modeling framework that integrates graph convolutional networks and GRU structures, this paper realizes the joint modeling of topology information and time series load data between nodes in a microgrid, effectively capturing the complex correlation between nodes and significantly improving the model's adaptability to sudden load changes and prediction accuracy.

[0083] By introducing a "structure-time" joint embedding tensor construction mechanism and combining node types (such as energy storage nodes, load nodes, and power generation nodes) and their flow direction characteristics, adaptive generation and classification modeling of prediction samples based on different node functional characteristics are achieved, enhancing the model's generalization capability.

[0084] Through the error-guided dynamic adjustment of adjacency weights and structural optimization mechanism, automatic learning of load transmission relationships between nodes and self-optimization of network structures are achieved, significantly improving the model consistency score and generalization stability in multi-node and multi-load collaborative prediction scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0086] Figure 1 This is a schematic diagram of the overall method steps of the microgrid load prediction optimization algorithm based on deep learning in the present invention. DETAILED DESCRIPTION

[0087] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0088] Example 1

[0089] Reference Figure 1 , which is the first embodiment of the present invention, provides a microgrid load forecasting optimization algorithm based on deep learning, including the following steps:

[0090] S1: Collect microgrid operation data and construct a microgrid topology diagram based on the collected operation data.

[0091] Specifically, the microgrid operation data is collected by deploying a sensor network, and the real-time operation data of the microgrid is collected by sensors in the sensor network. The sensor network divides the microgrid into multiple nodes and deploys sensors at each node. The microgrid topology is based on the node type in the microgrid and the connection edges between different nodes. The node type is determined by deploying sensors in the microgrid operation nodes and judging the current node type based on the real-time operation data of the microgrid collected by the sensors, including power generation nodes, power storage nodes and load nodes. The connection edges are the connection relationships between different nodes. The structure composed of the connection edges of all nodes is the microgrid topology diagram, which is specifically constructed as follows:

[0092] Divide the microgrid into nodes, then we have ,in, represents the set of nodes in the microgrid, Represents the first node, Indicates the Nodes, and deploy sensors at each node to collect real-time operation data of each node, including current data, current phase angle, voltage data, voltage phase angle,

[0093] Based on the data collected by each node sensor, the type of the current node is determined, specifically:

[0094] Based on the real-time operation data of the microgrid collected by the node, the active power and reactive power corresponding to the current node are calculated. At the same time, the flow direction of the current node is determined. Then,

[0095] Active power,

[0096]

[0097] reactive power,

[0098]

[0099] in, 、 Respectively represent the voltage data and current data collected at the current node, 、 Indicates the current phase angle and voltage phase angle corresponding to the current node, Indicates the active power corresponding to the current node, Indicates the reactive power corresponding to the current node;

[0100] According to the power corresponding to the node, the current flow direction of the node is determined, then,

[0101] For the current node, select the adjacent nodes of the current node, calculate the active power flow and reactive power flow from the current node to the adjacent nodes respectively, and judge the power flow direction of the current node based on the calculation results, specifically:

[0102] Active power from the current node to the adjacent node,

[0103]

[0104] Reactive power from the current node to the adjacent node,

[0105]

[0106] in, Indicates the voltage data of the current node, Represents the voltage data of adjacent nodes, represents the voltage phase angle of the current node, represents the voltage phase angle of adjacent nodes, represents the impedance between the current node and the adjacent node, Indicates the active power flow from the current node to the adjacent node, Indicates the reactive power flow from the current node to the adjacent node;

[0107] According to the calculation results of active power flow and reactive power flow, the flow direction of the current node is determined, then,

[0108] If the active power flow of the current node satisfies the formula , indicates that the power of the current node flows from the current node to the adjacent node, and vice versa, it means that the power flows from the adjacent node to the current node;

[0109] If the reactive power flow of the current node satisfies the formula , indicating that the reactive power of the current node flows from the current node to the adjacent node, and vice versa, it means that the reactive power flows from the adjacent node to the current node;

[0110] Determine the node type based on the active power, reactive power, and power flow direction corresponding to the current node, as follows:

[0111] If the active power corresponding to the current node satisfies the formula , and the current node's flow direction is outflowing from the current node, which means the current node type is a power generation node;

[0112] If the active power corresponding to the current node satisfies the formula , and the reactive power satisfies the formula ,At the same time, if the current node's flow direction is from the adjacent node to the current node, it means that the current node type is a load node;

[0113] For nodes whose node type is energy storage node, the judgment is based on the charge state of the node, specifically:

[0114] Check whether the current node has capacitance parameters. If the current node cannot collect capacitance parameters, it means that the current node is not an energy storage node.

[0115] If the current node can collect capacitance parameters, it means that the current node type is an energy storage node.

[0116] Based on different node types, a microgrid topology diagram is constructed, specifically:

[0117] According to the microgrid nodes, and the identified node types, then we have, ,in, represents the set of load nodes, represents the set of discharge nodes, Represents the energy storage node set, and the total number of the three node types is indivual, represents the set of nodes in the microgrid;

[0118] According to the calculated flow direction, all nodes are connected based on the graph structure. Once the edge relationships between all nodes are determined, the microgrid topology diagram is completed.

[0119] S2: Based on the in-depth analysis of the microgrid topology diagram, the degree of correlation between topological nodes is determined.

[0120] Specifically, the in-depth analysis based on the microgrid topology diagram includes analyzing the degree of correlation between nodes based on a deep learning algorithm and predicting the operating load of the nodes based on historical load, which is specifically implemented as follows:

[0121] The degree of correlation between nodes is analyzed based on deep learning algorithms. This is to use deep learning algorithms to conduct in-depth analysis of the edges of the topological structure and determine the degree of correlation between nodes. The specific implementation is as follows:

[0122] According to the constructed microgrid topology diagram ,in, represents the set of nodes in the microgrid, Represents the set of connecting edges between nodes, Represents the set of influence weights between nodes;

[0123] Influence weights between nodes , representing a node For Node The influence weight of nodes is analyzed based on the deep learning algorithm, specifically:

[0124] For any node in the microgrid topology , extract the node features corresponding to the node , through the convolutional neural network model to the node The running status of the Since the convolutional neural network model is a well-known technology, it will not be described in detail here;

[0125] For nodes The results of the initial analysis , choose non- Node of Node , and perform feature splicing on the two node features, then we have,

[0126]

[0127] in, Representation node The extracted node features, Representation node The extracted node features, Represents the concatenated features, which are used for in-depth analysis of the influence weights between nodes. Then,

[0128] According to the spliced features, the state analysis is performed through the convolutional neural network model, and the analysis results are set as , comparing the two analysis results as well as , based on the comparison results, a deep analysis of the influence weights between nodes is conducted, specifically:

[0129] If the comparison result satisfies the formula , then it means the node Relative to the node The influence weight of is negative, then the weight coefficient value satisfies the formula ,The specific value is set by the implementer according to the actual application scenario;

[0130] If the comparison result satisfies the formula , then it means the node Relative to the node The influence weight of is positive, then the weight coefficient value satisfies the formula ,The specific value is set by the implementer according to the actual application scenario;

[0131] If the comparison result satisfies the formula , then it means the node Relative to the node The influence weight of is no influence, then the weight coefficient value satisfies the formula .

[0132] It should be noted that for the in-depth analysis of the influence weights between nodes, a state analysis is performed on any node each time. After the state analysis is completed, the features of the latter node and the former node (the node that has completed the analysis) are spliced, and the spliced features are subjected to a second state analysis. The results of the two state analyses are compared, and an in-depth analysis of the influence weights between nodes is performed based on the analysis results, which provides a basis for judging the value of the weight coefficient until the state analysis corresponding to all nodes is completed. The selection of the previous and next nodes is based on the connecting edges of the topological structure.

[0133] Furthermore, the operation load of the node is predicted based on the historical load. The node load sequence characteristics and structural embedding characteristics are constructed for each node according to the input historical load data set and the microgrid topology diagram, and the operation load of the node is predicted based on the influence weight between nodes. The specific implementation is as follows:

[0134] For each node, based on its historical load sequence, the length of the extraction time window is The sliding sequence and constructing the historical state matrix, then we have,

[0135] For any node , set the corresponding historical load sequence, then we have,

[0136]

[0137] in, Representation node In time Load data at the moment, including active power , reactive power ;

[0138] For any node , the construction time window length is The sliding time slice set of , then we have,

[0139]

[0140] in, Representation node The corresponding historical state matrix, represents the length of the divided time step, Indicates the total length of the time window, Represents one row in the historical state matrix;

[0141] According to the active power, reactive power and node type label (generation / load / energy storage) of the node, a set of predicted input samples is generated for each node, then,

[0142] Define node type labels , the values are -1, 0, and 1 respectively. When the value is -1, it indicates a load node; when the value is 0, it indicates an energy storage node; when the value is 1, it indicates a power generation node;

[0143] Constructing an input sample set ,in, Indicates the node type label, represents the time step length, Represents the feature dimension of each time step, if the formula is satisfied , which indicates that the characteristics of the current time step include,active power, reactive power and node type;

[0144] For the generated set of node historical state samples, a graph convolutional network is used to extract the topological features between nodes, and the extracted topological features are modeled for time series to generate the initial prediction results:

[0145] Input the constructed joint feature tensor into the graph convolution layer to extract the node graph embedding, then we have,

[0146]

[0147] in, represents the normalized adjacency matrix, represents the graph convolution weight, represents the activation function, represents the constructed joint feature tensor, represents the extracted node graph embedding features;

[0148] Embedding graphs into features Input into the LSTM network and obtain the hidden state output, then we have,

[0149]

[0150] in, Representation node The corresponding graph embedding features, Representation node The corresponding hidden state output;

[0151] Based on the hidden state output, generate the initial prediction result, then we have,

[0152]

[0153] in, 、 Represent the weight coefficient and bias term of the initial prediction respectively, Representation node The corresponding hidden state output is, Representation node Initial load forecast results.

[0154] For the generated initial set of prediction results, we perform error analysis on the prediction results and readjust the graph embedding method based on the prediction error to generate optimized prediction results (second-generation prediction values):

[0155] Back propagation of the error between the first generation predicted value and the actual value, we have,

[0156]

[0157] in, Representation node Initial load forecast results, Representation node Actual measurement results of load, represents the error of the first generation prediction value;

[0158] Based on the error of the first generation prediction value, the adjacency weight in the graph structure is dynamically updated to correct the influence relationship between nodes. Then,

[0159]

[0160] in, represents the normalized adjacency matrix, represents the adjusted normalized adjacency matrix, represents the error change of the first generation prediction value, Indicates the attention weight, which is set by the implementer according to the actual application scenario;

[0161] Based on the adjusted normalized adjacency matrix, the first prediction value is subjected to secondary extraction of graph embedding features, and then,

[0162]

[0163] in, represents the adjusted normalized adjacency matrix, represents the graph convolution weight, represents the activation function, represents the constructed joint feature tensor, Represents the secondary extracted node graph embedding features;

[0164] And based on the secondary extraction results of the graph embedding features, the corresponding hidden state output is obtained, then,

[0165]

[0166] in, Representation node The corresponding secondary extracted graph embedding features, Representation node Hidden state output corresponding to the secondary extraction result;

[0167] The second generation prediction of node load is performed based on the acquired hidden state data, then,

[0168]

[0169] in, 、 They represent the weight coefficient and bias term of the initial prediction, which are set by the implementer according to the actual application scenario. Representation node The hidden state output corresponding to the secondary extraction result, Representation node Secondary load prediction results;

[0170] Based on the second-generation prediction value set, a fitting scoring mechanism is introduced to determine the comparison between the fitting score corresponding to the second-generation prediction value and the fitting score threshold. The adjacency structure is readjusted and the node load forecast (third-generation prediction value) is performed based on the reconstructed adjacency structure. Specifically,

[0171] Comparing the fitting scores between the second generation prediction value and the target load curve, we have:

[0172]

[0173] in, Representation node exist The actual load measurement value at the moment, Representation node exist The second generation prediction value at time , Indicates the length of the time window, Represents the fitting score of the second-generation prediction value, which is used to reconstruct the adjacency structure, specifically:

[0174] Setting the Fit Score Threshold , if the calculated fit score satisfies the formula , it means that the adjacency matrix needs to be readjusted to reconstruct the adjacency structure until the fitting score is higher than the set score threshold. Otherwise, it means that there is no need to readjust the adjacency structure. The specific value of the adjustment of the adjacency matrix is set by the implementer according to the actual application scenario.

[0175] For the adjusted prediction model, regenerate the third-generation prediction value according to the prediction method of the second-generation prediction value, and calculate the fitting score corresponding to the third-generation prediction value , compare the fitting score corresponding to the third-generation prediction value with the fitting score corresponding to the second-generation prediction value until the comparison result satisfies the formula So far, it means that the load forecast of the node is accurate, and the third-generation prediction value is used as the load forecast result of the node.

[0176] S3: Based on the influence weights between nodes and the load forecast results of the nodes, the microgrid load forecast is optimized.

[0177] Specifically, the microgrid load forecast optimization is to collaboratively optimize and predict the global load of the microgrid based on the influence weights between nodes in the microgrid topology and the load forecast results for the nodes. The specific implementation is as follows:

[0178] According to the influence weights between nodes and the topological structure of nodes, a global influence weight matrix is constructed , represents the total number of nodes in the microgrid node set, where , positive values indicate positive impact, negative values indicate negative impact, and 0 indicates no impact;

[0179] According to the node load prediction results, obtain the load of each node at the prediction time. The third generation forecast value of the load is constructed and the initial forecast set of the global load is constructed.

[0180]

[0181] in, The constructed global load initial forecast set, represents the third generation prediction value of the first node, Indicates the The third-generation predicted value of nodes;

[0182] For nodes The prediction result is based on its adjacent node set , integrate the prediction output of adjacent nodes and make collaborative corrections to the prediction values, then we have,

[0183]

[0184] in, Representation node The predicted value after integrating the adjacent nodes, Representation node The third generation prediction value of Representation node The third generation prediction value of Representation node With node The influence weight between Representation node The adjacent node set of the node is the topological structure. A set of nodes that can form edge relationships between them. Indicates the synergy correction coefficient, which is set by the implementer according to the actual application scenario. Specifically:

[0185] If the node The prediction result of the adjacent node set is compared with the prediction result of the adjacent node set to satisfy the formula , indicating that the node prediction is disturbed by the adjacent nodes, and the correction coefficient is increased to achieve the correction of the adjacent nodes;

[0186] For nodes The predicted value after integrating adjacent nodes The error between the actual measured value and the weight adjustment factor between each pair of nodes is calculated, then:

[0187]

[0188] in, represents the adjustment coefficient, Representation node The predicted value after integrating the adjacent nodes, Representation node The third-generation prediction value of node Actual measurement results of load, Representation node With node The weight adjustment factor between nodes is used to achieve dynamic adjustment of the weights between nodes, then,

[0189]

[0190] in, Representation node With node The influence weight between Representation node With node The weight adjustment factor between Representation node With node The adjusted impact weights between

[0191] Repeat the collaborative correction of the predicted value by the integrated adjacent nodes and the dynamic adjustment process of the weight until the calculated weight adjustment factor is less than the set weight adjustment factor threshold (set by the implementer according to the actual application scenario). At this time, the set of prediction results generated in the current iteration is:

[0192]

[0193] in, represents the optimization result of microgrid load forecasting, represents the load forecast optimization result of the first node, Indicates the The load forecast optimization results of each node.

[0194] Furthermore, if the functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0195] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0196] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, and then editing, interpreting, or processing in another suitable manner as necessary, and then storing it in a computer memory.

[0197] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A microgrid load forecasting optimization algorithm based on deep learning, characterized by: The following steps are included: Use sensors to collect real-time operation data of the microgrid, and based on the collected real-time operation data, divide the microgrid into different types of nodes, and construct a microgrid topology diagram based on the connection relationship between all nodes; The influence weights between nodes are analyzed through deep learning algorithms, and the operating load of nodes is predicted using historical loads. Specifically: Based on the input historical load data set and microgrid topology diagram, the node load sequence characteristics and structural embedding characteristics are constructed for each node, and the node operating load is predicted based on the influence weights between nodes; Based on the influence weights between nodes in the microgrid topology and the load prediction results for the nodes, the global load of the microgrid is collaboratively optimized and predicted, specifically: According to the prediction results of the node, the prediction output of the adjacent nodes is integrated to perform collaborative correction of the prediction value. At the same time, according to the error between the prediction output of the adjacent nodes and the actual measurement value, the influence weights between the nodes are dynamically adjusted, and the microgrid load prediction is performed based on the adjusted influence weights.

2. The microgrid load forecasting optimization algorithm based on deep learning according to claim 1 is characterized in that: The use of sensors to collect real-time microgrid operation data is specifically as follows: Calculate the active power and reactive power corresponding to the current node, and at the same time, determine the flow direction of the current node, then we have, Active power, reactive power, in, 、 Respectively represent the voltage data and current data collected at the current node, 、 Indicates the current phase angle and voltage phase angle corresponding to the current node, Indicates the active power corresponding to the current node, Indicates the reactive power corresponding to the current node.

3. The microgrid load forecasting optimization algorithm based on deep learning according to claim 2 is characterized in that: The flow direction of the current node is determined as follows: For the current node, select the adjacent nodes of the current node, calculate the active power flow and reactive power flow from the current node to the adjacent nodes respectively, and judge the power flow direction of the current node based on the calculation results, specifically: Active power from the current node to the adjacent node, Reactive power from the current node to the adjacent node, in, Indicates the voltage data of the current node, Represents the voltage data of adjacent nodes, represents the voltage phase angle of the current node, represents the voltage phase angle of adjacent nodes, represents the impedance between the current node and the adjacent node, Indicates the active power flow from the current node to the adjacent node, Indicates the reactive power flow from the current node to the adjacent node; According to the calculation results of active power flow and reactive power flow, the flow direction of the current node is determined, then, If the active power flow of the current node satisfies the formula , indicates that the power of the current node flows from the current node to the adjacent node, and vice versa, it means that the power flows from the adjacent node to the current node; If the reactive power flow of the current node satisfies the formula , indicating that the reactive power of the current node flows from the current node to the adjacent node, and vice versa, it means that the reactive power flows from the adjacent node to the current node.

4. The microgrid load forecasting optimization algorithm based on deep learning according to claim 3 is characterized in that: The microgrid is divided into different types of node types as follows: If the active power corresponding to the current node satisfies the formula , and the current node's flow direction is outflowing from the current node, which means the current node type is a power generation node; If the active power corresponding to the current node satisfies the formula , and the reactive power satisfies the formula ,At the same time, if the current node's flow direction is from the adjacent node to the current node, it means that the current node type is a load node; For nodes whose node type is energy storage node, the judgment is based on the charge state of the node, specifically: Check whether the current node has capacitance parameters. If the current node cannot collect capacitance parameters, it means that the current node is not an energy storage node. If the current node can collect capacitance parameters, it means that the current node type is an energy storage node.

5. The microgrid load forecasting optimization algorithm based on deep learning according to claim 4 is characterized in that: The influence weights between nodes analyzed by deep learning algorithm are as follows: For any node in the microgrid topology , extract the node features corresponding to the node , through the convolutional neural network model to the node The running status of the ; For nodes The results of the initial analysis , choose non- Node of Node , and perform feature splicing on the two node features, then we have, in, Representation node The extracted node features, Representation node The extracted node features, Represents the concatenated features, which are used for in-depth analysis of the influence weights between nodes. Then, According to the spliced features, the state analysis is performed through the convolutional neural network model, and the analysis results are set as , comparing the two analysis results as well as , based on the comparison results Conduct an in-depth analysis of the influence weights between nodes, specifically: If the comparison result satisfies the formula , then it means the node Relative to the node The impact weight is negative; If the comparison result satisfies the formula , then it means the node Relative to the node The impact weight is positive; If the comparison result satisfies the formula , then it means the node Relative to the node The influence weight is no influence.

6. The microgrid load forecasting optimization algorithm based on deep learning according to claim 5 is characterized in that: The operation load of the node predicted by using historical load is specifically as follows: For each node, based on its historical load sequence, the length of the extraction time window is Sliding sequence and construct the historical state matrix; Define node type labels and construct an input sample set. Use a graph convolutional network to extract topological features between nodes and perform time series modeling on the extracted topological features to generate initial prediction results. Based on the generated initial prediction result set, the prediction results are analyzed and the prediction error is calculated. Based on the prediction error, the graph embedding method is readjusted to generate the second generation prediction value. According to the second-generation prediction value set, by introducing a fitting scoring mechanism, the comparison result between the fitting score corresponding to the second-generation prediction value and the fitting score threshold is judged, the adjacency structure is readjusted, and the node load prediction is performed based on the reconstructed adjacency structure.

7. The microgrid load forecasting optimization algorithm based on deep learning according to claim 6 is characterized in that: The initial prediction results are generated as follows: Input the constructed joint feature tensor into the graph convolution layer to extract the node graph embedding, then we have, in, represents the normalized adjacency matrix, represents the graph convolution weight, represents the activation function, represents the constructed joint feature tensor, represents the extracted node graph embedding features; Embedding graphs into features Input into the LSTM network and obtain the hidden state output, then we have, in, Representation node The corresponding graph embedding features, Representation node The corresponding hidden state output; Based on the hidden state output, generate the initial prediction result, then we have, in, 、 Represent the weight coefficient and bias term of the initial prediction respectively, Representation node The corresponding hidden state output is, Representation node Initial load forecast results.

8. The microgrid load forecasting optimization algorithm based on deep learning according to claim 7 is characterized in that: The load forecast of the node is as follows: Comparing the fitting scores between the second generation prediction value and the target load curve, we have: in, Representation node exist The actual load measurement value at the moment, Representation node exist The second generation prediction value at time , Indicates the time window length, Represents the fitting score of the second-generation prediction value, which is used to reconstruct the adjacency structure. Specifically: Setting the Fit Score Threshold , if the calculated fit score satisfies the formula , it means that the adjacency matrix needs to be readjusted to reconstruct the adjacency structure until the fitting score is higher than the set score threshold. Otherwise, it means that there is no need to readjust the adjacency structure. For the adjusted prediction model, regenerate the third-generation prediction value according to the prediction method of the second-generation prediction value, and calculate the fitting score corresponding to the third-generation prediction value , compare the fitting score corresponding to the third-generation prediction value with the fitting score corresponding to the second-generation prediction value until the comparison result satisfies the formula So far, it means that the load forecast of the node is accurate, and the third-generation prediction value is used as the load forecast result of the node.

9. The microgrid load forecasting optimization algorithm based on deep learning according to claim 8 is characterized in that: The collaborative correction of the predicted value is specifically as follows: For nodes The prediction result is based on its adjacent node set , integrate the prediction output of adjacent nodes and make collaborative corrections to the prediction values, then we have, in, Representation node The predicted value after integrating the adjacent nodes, Representation node The third generation prediction value of Representation node The third-generation prediction value of Representation node With node The influence weight between Representation node The adjacent node set of the node is the topological structure. A set of nodes that can form edge relationships between them. represents the synergy correction coefficient, specifically: If the node The prediction result of the adjacent node set is compared with the prediction result of the adjacent node set to satisfy the formula , indicating that the node prediction is disturbed by the adjacent nodes, and the correction coefficient is increased to achieve the correction of the adjacent nodes.

10. The microgrid load forecasting optimization algorithm based on deep learning according to claim 9 is characterized in that: The microgrid load forecasting based on the adjusted impact weight is specifically as follows: For nodes The predicted value after integrating adjacent nodes The error between the actual measured value and the weight adjustment factor between each pair of nodes is calculated, then: in, represents the adjustment coefficient, Representation node The predicted value after integrating the adjacent nodes, Representation node The third-generation prediction value of node Actual measurement results of load, Representation node With node The weight adjustment factor between nodes is used to achieve dynamic adjustment of the weights between nodes, then, in, Representation node With node The influence weight between Representation node With node The weight adjustment factor between Representation node With node The adjusted impact weights between Repeat the collaborative correction of the predicted value by the integrated adjacent nodes and the dynamic adjustment process of the weight until the calculated weight adjustment factor is less than the set weight adjustment factor threshold. At this time, the prediction result set generated in the current iteration is: in, represents the optimization result of microgrid load forecasting, represents the load forecast optimization result of the first node, Indicates the Load forecast optimization results for each node.

Images