Power distribution network power prediction system containing multi-element source load and method thereof

Abstract

The invention provides an electric power prediction system and method for a power distribution network containing multiple source loads. The system comprises a power distribution network management middle platform, a data acquisition module, a feature extraction module, a correlation analysis module, a model matching module and an electric power prediction module. The data acquisition module acquires power path complexity, power transmission quantity, operation data of a multi-element power supply and load data of a multi-element load; the feature extraction module extracts the operation data and the load data to obtain a power feature vector and a load feature vector; the correlation analysis module analyzes and obtains time correlation feature information and space correlation feature information based on the power supply feature vector and the load feature vector; the model matching module is used for matching a target power prediction model based on the power path complexity and the power transmission quantity; and the power prediction module inputs the time correlation feature information and the space correlation feature information into a target power prediction model to obtain a power prediction result. The power prediction precision of the power distribution network is improved, and safe and stable operation of the power distribution network is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a power prediction system for a distribution network containing multiple sources and loads and a method thereof. Background Art

[0002] In the power system sector, distribution networks directly serve end users, and the accuracy of their power forecasts is crucial for the safe and stable operation of the grid, the rational allocation of resources, and the improvement of economic efficiency. In recent years, with the large-scale integration of distributed energy resources and the rapid growth of new loads such as smart appliances and electric vehicles, distribution networks have exhibited complex characteristics of diverse source-load interactions.

[0003] In current distribution networks, the primary method for power forecasting is based on time series prediction methods, such as the autoregressive moving average model, which primarily relies on the temporal correlation of historical power data. However, in distribution networks with multiple sources and loads, the output of distributed power sources is highly intermittent and random, influenced by natural conditions (such as sunlight and wind speed). Furthermore, the power consumption patterns of new loads are diverse and uncertain. These factors significantly alter the statistical characteristics of power data, making it difficult for traditional time series prediction methods to accurately capture the dynamic changes in data. This results in a significant decrease in the accuracy of power forecasts for distribution networks, making it difficult to ensure the safe and stable operation of distribution networks. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] The present invention aims to provide a power forecasting system and method for a distribution network containing multiple sources and loads, aiming to improve the power forecasting accuracy of the distribution network and ensure the safe and stable operation of the distribution network.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a power forecasting system for a distribution network with multiple sources and loads, comprising: a distribution network management center, a data acquisition module, a feature extraction module, a correlation analysis module, a model matching module, and a power forecasting module; the distribution network management center is connected to the data acquisition module, the feature extraction module, the correlation analysis module, the model matching module, and the power forecasting module, respectively, to manage each module;

[0007] A data acquisition module is used to obtain the power path complexity between the current distribution network and the connected target distribution network and the power transmission volume in each power transmission path direction based on the power grid topology of the current distribution network, as well as the operating data of multiple power sources and load data of multiple loads in the current distribution network;

[0008] A feature extraction module is used to extract features from operating data based on the operating characteristics of multiple power sources to obtain a power feature vector that can reflect the power output, and to extract features from load data based on the load characteristics of multiple loads to obtain a load feature vector that can reflect changes in load demand;

[0009] A correlation analysis module is used to perform correlation analysis based on the power characteristic vector and the load characteristic vector to determine the time correlation characteristic information of the power characteristic vector and the load characteristic vector in time distribution and the spatial correlation characteristic information in spatial distribution;

[0010] a model matching module for performing model matching in a preset model library based on the complexity of the power path and the power transmission amount of each power transmission amount to obtain a target power prediction model;

[0011] The power prediction module is used to input time-related feature information and space-related feature information into the target power prediction model to obtain the power prediction result of the current distribution network output by the target power prediction model; the target power prediction model is trained based on the sample-related feature information and its corresponding power label results.

[0012] As a preferred solution of the power forecasting method for a distribution network containing multiple sources and loads described in the present invention, the method includes: obtaining the power path complexity and power transmission amount in each power transmission path direction between the current distribution network and the connected target distribution network based on the grid topology of the current distribution network, and obtaining the operating data of the multiple power sources and the load data of the multiple loads in the current distribution network;

[0013] Extracting features from the operating data based on the operating characteristics of the multiple power sources to obtain a power feature vector capable of reflecting power output, and extracting features from the load data based on the load characteristics of the multiple loads to obtain a load feature vector capable of reflecting changes in load demand;

[0014] Performing a correlation analysis based on the power source characteristic vector and the load characteristic vector to determine time correlation characteristic information of the power source characteristic vector and the load characteristic vector in time distribution and space correlation characteristic information of the power source characteristic vector and the load characteristic vector in space distribution;

[0015] Performing model matching in a preset model library based on the complexity of the power path and each of the power transmission amounts to obtain a target power prediction model;

[0016] The time-related feature information and the space-related feature information are input into the target power prediction model to obtain the power prediction result of the current distribution network output by the target power prediction model; the target power prediction model is trained based on the sample-related feature information and its corresponding power label results.

[0017] As a preferred solution of the method for power prediction of a distribution network containing multiple sources and loads described in the present invention, wherein: the operating characteristics include power fluctuation characteristics, phase space characteristics and probabilistic characteristics of power supply status;

[0018] The extracting features of the operating data based on the operating characteristics of the multi-source power supply to obtain a power feature vector capable of reflecting the power output includes:

[0019] Determine the power fluctuation entropy feature at any moment based on the power fluctuation characteristics and the power supply active power and the power supply reactive power of the multi-element power supply at any moment;

[0020] Reconstruct the phase space of the voltage signal of the multi-element power source at any moment based on the phase space characteristics combined with the delay time and the embedding dimension to determine the power source phase space characteristics at any moment;

[0021] Determine the power state probability transition characteristics at any moment based on the probability characteristics of the power state and the probability of occurrence of each power state;

[0022] The power fluctuation entropy characteristics, power supply phase space characteristics and probability transfer characteristics at any moment are integrated to obtain the power supply characteristic vector of the multi-element power supply at any moment.

[0023] As a preferred solution of the method for power forecasting of a distribution network containing multiple sources and loads described in the present invention, wherein: the load characteristics include time characteristics, seasonal characteristics and industry characteristics;

[0024] The extracting features of the load data based on the load characteristics of the multi-element load to obtain a load feature vector capable of reflecting changes in load demand includes:

[0025] Decomposing the load active power and negative load reactive power of the time series based on the time characteristics, and obtaining the intrinsic modal component and the residual component based on the decomposition of the load active power and the negative load reactive power, to determine the power energy characteristics of the multi-element load;

[0026] Based on the seasonal characteristics, the load active power and negative load reactive power of the time series are divided into time segments according to seasons, and the power seasonal characteristics of the multivariate load are determined according to the seasonal correlation coefficient of the active power and negative load reactive power in each time segment;

[0027] Clustering different industries based on the industry characteristics, and performing statistical analysis on the load active power and negative load reactive power corresponding to the industry categories obtained by clustering to obtain the power industry clustering characteristics of the multivariate load; the statistical analysis includes mean and variance;

[0028] The power energy characteristics, power seasonal characteristics and power industry clustering characteristics of the multi-load are integrated to obtain a load characteristic vector of the multi-load.

[0029] As a preferred solution of the power forecasting method for a distribution network containing multiple sources and loads described in the present invention, wherein: performing correlation analysis based on the power source characteristic vector and the load characteristic vector to determine the time correlation characteristic information of the power source characteristic vector and the load characteristic vector in time distribution includes:

[0030] At each time point in a time sliding window with a preset window length, feature sequence extraction is performed on the power feature vector and the load feature vector to obtain a power feature vector sequence and a load feature vector sequence, respectively;

[0031] Determining a similarity metric value between the power source feature vector sequence and the load feature vector sequence within the time sliding window based on the vector similarity and cosine similarity between the first feature vector in the power source feature vector sequence and the second feature vector in the load feature vector sequence;

[0032] Determining a time series based on a similarity measure value at each time point in the time sliding window;

[0033] The time-related feature information is determined based on the time series and its sequence features.

[0034] As a preferred embodiment of the power forecasting system and method for a distribution network containing multiple sources and loads described in the present invention, the method further comprises: performing correlation analysis based on the power source characteristic vector and the load characteristic vector to determine spatial correlation characteristic information of the power source characteristic vector and the load characteristic vector in spatial distribution, including:

[0035] Taking each power source point in the multiple power sources and each load point in the multiple loads as the center, determine a first spatial neighborhood range of each power source point and a second spatial neighborhood range of each load point;

[0036] Obtaining a set of load points within a first spatial neighborhood of each power point, and a set of power points within a second spatial neighborhood of each load point;

[0037] Determining a first difference metric value based on a power source characteristic vector of each power source point and a load characteristic vector of any load point in the load point set, and determining a second difference metric value based on the load characteristic vector of each load point and a power source characteristic vector of any power source point in the power source point set;

[0038] determining a first average difference metric value for each power source point based on the first difference metric value for each power source point, and determining a second average difference metric value for each load point based on the second difference metric value for each load point;

[0039] The first average difference metric value of each power source point and the second average difference metric value of each load point are fused to obtain the spatial correlation feature information.

[0040] As a preferred embodiment of the power forecasting system and method for a distribution network with multiple sources and loads described in the present invention, the method includes: performing model matching in a preset model library based on the power path complexity and each power transmission amount to obtain a target power forecasting model, including:

[0041] Normalizing the power path complexity and each of the power transmission amounts, and performing row fusion on the normalized results to obtain a first eigenvector;

[0042] Determining the degree of adaptation of the first eigenvector to the second eigenvector of each power prediction model based on the degree of difference between the first eigenvector and the second eigenvector corresponding to each power prediction model in the preset model library; the second eigenvector of each power prediction model is generated by normalizing the power path complexity and power transmission amount of each power prediction model under different scenarios;

[0043] determining a final matching degree between the first eigenvector and the second eigenvector of each power prediction model based on the adaptation degree and the model reliability vector of the power prediction model;

[0044] The final matching degrees are traversed, and the power prediction model corresponding to the final matching degree with the largest value is determined as the target power prediction model.

[0045] As a preferred embodiment of the power forecasting system and method for a distribution network with multiple sources and loads described in the present invention, the method of obtaining the power path complexity and power transmission amount in each power transmission path direction between the current distribution network and the connected target distribution network based on the grid topology of the current distribution network includes:

[0046] determining a power transmission path from a first node in the current distribution network to a second node in the target distribution network based on the grid topology;

[0047] determining the power path complexity based on the voltage level and power complexity of each node in the power path, and the path length and path rated capacity between two adjacent nodes in the power path;

[0048] determining a power transmission tendency factor of the power transmission path based on the voltage level of the first node, the voltage level of the second node, and the path reactance and path resistance of the power transmission path, and determining a power transmission path direction according to the power transmission tendency factor;

[0049] An amount of power transmission of the power transmission path in a direction of the power transmission path is determined based on a power transmission tendency factor, a path reactance, and a path resistance of the power transmission path.

[0050] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, it implements the steps of a power prediction system for a distribution network containing multiple sources and loads and a method thereof.

[0051] A computer-readable storage medium stores a computer program thereon, characterized in that when the computer program is executed by a processor, it implements the steps of a power prediction system for a distribution network containing multiple sources and loads and a method thereof.

[0052] Beneficial effects of the present invention: The power prediction system for a distribution network containing multiple sources and loads provided by an embodiment of the present invention extracts a power characteristic vector that can reflect the power output from the operating data of the multiple power sources according to the operating characteristics, and extracts a load characteristic vector that can reflect the change in load demand from the load data of the multiple loads according to the load characteristics. Therefore, it can better capture the changing rules of power output and load demand. At the same time, by analyzing the associated characteristic information of the power characteristic vector and the load characteristic vector in time distribution and spatial distribution, it can better capture the complex nonlinear relationship between power output and load demand. On the other hand, the target power prediction model adapted to the current situation can be accurately obtained through the complexity of the power path and the amount of each power transmission, and then the power prediction result can be accurately predicted based on the target power prediction model combined with the associated characteristic information. Therefore, the embodiment of the present invention improves the power prediction accuracy of the distribution network and ensures the safe and stable operation of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] 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:

[0054] Figure 1 It is a structural diagram of a power forecasting system for a distribution network containing multiple sources and loads provided by the present invention;

[0055] Figure 2It is a flow chart of the power forecasting method for a distribution network with multiple sources and loads provided by the present invention;

[0056] Figure 3 An embodiment diagram of an electronic device provided by an embodiment of the present invention;

[0057] Figure 4 An embodiment diagram of a computer-readable storage medium provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0058] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments 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 persons in this field without creative work should fall within the scope of protection of the present invention.

[0059] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0060] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0061] The present invention is described in detail with reference to schematic diagrams. For ease of illustration, cross-sectional views of device structures may be partially enlarged and not to scale when describing embodiments of the present invention. Furthermore, the schematic diagrams are merely illustrative and should not limit the scope of the present invention. Furthermore, in actual production, the three-dimensional dimensions of length, width, and depth should be included.

[0062] In the description of the present invention, it should be noted that the terms "upper, lower, inner, and outer" and other references to orientations or positional relationships are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first, second, or third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0063] In this disclosure, unless otherwise specified or limited, the terms "mounted," "connected," and "connected" should be interpreted broadly. For example, they may refer to fixed, removable, or integral connections. They may also refer to mechanical, electrical, or direct connections, indirect connections through an intermediary, or internal communication between two components. Those skilled in the art will understand the specific meanings of these terms in this disclosure.

[0064] Example 1

[0065] Reference Figure 1-Figure 4 , which is the first embodiment of the present invention, provides a power forecasting system for a distribution network with multiple sources and loads and a method thereof, including:

[0066] Optional, see Figure 1 As shown, Figure 1 The diagram is a schematic diagram of the structure of a power forecasting system for a distribution network with multiple sources and loads provided by the present invention. The power forecasting system for a distribution network with multiple sources and loads includes a distribution network management center, a data acquisition module, a feature extraction module, a correlation analysis module, a model matching module, and a power forecasting module. The distribution network management center in the embodiment of the present invention is connected to the data acquisition module, feature extraction module, correlation analysis module, model matching module, and power forecasting module, respectively. Therefore, the distribution network management center can manage each module.

[0067] Optionally, the data acquisition module obtains the power path complexity of the current distribution network and the connected target distribution network and the power transmission amount in each power transmission path direction based on the grid topology of the current distribution network, as well as the operating data of multiple power sources and load data of multiple loads in the current distribution network.

[0068] Optionally, the feature extraction module extracts features from the operating data based on the operating characteristics of the multiple power sources to obtain a power feature vector that can reflect the power output, and extracts features from the load data based on the load characteristics of the multiple loads to obtain a load feature vector that can reflect changes in load demand.

[0069] Optionally, the correlation analysis module performs correlation analysis based on the power source characteristic vector and the load characteristic vector to determine time correlation characteristic information of the power source characteristic vector and the load characteristic vector in time distribution, as well as space correlation characteristic information of the power source characteristic vector and the load characteristic vector in space distribution.

[0070] Optionally, an embodiment of the present invention pre-constructs a preset model library, wherein the preset model library includes multiple power prediction models, each power prediction model is adapted to different power grid conditions, and each power prediction model is trained based on sample-related feature information and its corresponding power label results.

[0071] Therefore, the model matching module performs model matching in a preset model library according to the complexity of the power path and the power transmission amount of each power transmission amount to obtain a target power prediction model.

[0072] Optionally, the power prediction module inputs the time-related feature information and the space-related feature information into the target power prediction model to obtain the power prediction result of the current distribution network output by the target power prediction model.

[0073] The embodiment of the present invention extracts a power characteristic vector that can reflect the power output from the operating data of multiple power sources based on the operating characteristics, and extracts a load characteristic vector that can reflect the change in load demand from the load data of multiple loads based on the load characteristics. This can better capture the changing patterns of power output and load demand. At the same time, by analyzing the associated characteristic information of the power characteristic vector and the load characteristic vector in terms of time distribution and spatial distribution, it can better capture the complex nonlinear relationship between power output and load demand. The target power prediction model adapted to the current situation can be accurately obtained by measuring the complexity of the power path and the amount of power transmission. Then, the power prediction results can be accurately predicted based on the target power prediction model combined with the associated characteristic information, thereby improving the power prediction accuracy of the distribution network and ensuring the safe and stable operation of the distribution network.

[0074] Optional, see Figure 2 , Figure 2 : This is a flow chart of the method for predicting power in a distribution network with multiple sources and loads provided by the present invention. In the embodiment of the present invention, the method for predicting power in a distribution network with multiple sources and loads is executed by a power prediction system. Therefore, the method for predicting power in a distribution network with multiple sources and loads includes:

[0075] Step 10: Based on the grid topology of the current distribution network, the complexity of the power paths between the current distribution network and the connected target distribution network and the power transmission amount in each power transmission path direction are obtained, as well as the operating data of the multiple power sources and the load data of the multiple loads in the current distribution network.

[0076] Optionally, in the embodiment of the present invention, the topological structure of the current distribution network and the target distribution network connected thereto is a graph G = (V, E), where V is a node set including various power equipment nodes such as power nodes, load nodes, and substation nodes; E is a set of edges, and each edge e ij ∈E represents the power transmission path connecting node i and node j. For each edge e ij Assign attributes, such as path length L ij , path reactance X ij , path resistance R ij and the path rated transmission capacity C ij etc.; assign attributes to each node, such as the voltage level U of the node i wait.

[0077] Therefore, the power prediction system can obtain the power path complexity of the current distribution network and the connected target distribution network and the power transmission amount in each power transmission path direction based on the path length, path reactance, path resistance and path rated transmission capacity of each transmission path in the power grid topology, as well as the voltage level of each node, as described in steps 101 to 104.

[0078] Furthermore, the power forecasting system analyzes the multiple power sources and multiple loads in the current distribution network to obtain the operating data of the multiple power sources and the load data of the multiple loads. Among them, the multiple power sources include distributed photovoltaic devices, wind power devices and energy storage devices, and the multiple loads include industrial loads, residential loads and electric vehicle charging loads. The operating data of the multiple power sources include basic electrical parameters, power generation data, power status information and control parameters, etc., and the load data of the multiple loads include power consumption data, load curve data, load classification data, power consumption time information and power quality data, etc.

[0079] Step 20 , extracting features from the operating data based on the operating characteristics of the multiple power sources to obtain a power feature vector that can reflect the power output, and extracting features from the load data based on the load characteristics of the multiple loads to obtain a load feature vector that can reflect the change in load demand.

[0080] Optionally, the operating characteristics of the multi-element power supply in the embodiments of the present invention include power fluctuation characteristics, phase space characteristics, and probabilistic characteristics of the power supply state. The power fluctuation characteristic refers to the fluctuation of the power output of the power supply over time. The phase space characteristic refers to the state of the power supply at different times. The probabilistic characteristics of the power supply state refer to the probability of a power supply failure at different times.

[0081] The load characteristics of the multi-element load in the embodiments of the present invention include time characteristics, seasonal characteristics, and industry characteristics. Time characteristics refer to the regular changes in load demand over different time scales. Seasonal characteristics refer to seasonal variations in load demand due to factors such as climate and temperature. Industry characteristics refer to variations in load demand across different industries due to differences in production processes, working hours, and electrical equipment.

[0082] Therefore, the power prediction system extracts features from the operating data based on the power fluctuation characteristics, phase space characteristics, and probabilistic characteristics of the power supply state of the multi-source power supply to obtain a power supply feature vector that can reflect the power supply output, as described in steps 201 to 204.

[0083] Furthermore, the power forecasting system extracts features from the load data according to the time characteristics, seasonal characteristics, and industry characteristics of the multi-element loads to obtain a load feature vector that can reflect changes in load demand, as specifically described in steps 205 to 208 .

[0084] Step 30 : performing correlation analysis based on the power source characteristic vector and the load characteristic vector to determine time correlation characteristic information of the power source characteristic vector and the load characteristic vector in time distribution, and space correlation characteristic information of the power source characteristic vector and the load characteristic vector in space distribution.

[0085] Furthermore, the power forecasting system performs a correlation analysis on the power source characteristic vector and the load characteristic vector in terms of their temporal distribution to obtain temporal correlation characteristic information of the power source characteristic vector and the load characteristic vector in terms of their temporal distribution, as specifically described in steps 301 to 304. The power forecasting system also performs a correlation analysis on the power source characteristic vector and the load characteristic vector in terms of their spatial distribution to determine spatial correlation characteristic information of the power source characteristic vector and the load characteristic vector in terms of their spatial distribution, as specifically described in steps 305 to 309.

[0086] Step 40 : performing model matching in a preset model library based on the complexity of the power path and each power transmission amount to obtain a target power prediction model.

[0087] Optionally, embodiments of the present invention pre-build a preset model library, wherein the preset model library includes multiple power prediction models, each adapted to different power grid conditions, and each trained based on sample-related feature information and its corresponding power label results. Therefore, the power prediction system performs model matching within the preset model library based on power path complexity and individual power transmission amounts, obtaining a target power prediction model that best matches the current situation, as described in steps 401 to 404.

[0088] Step 50: Input the time-related characteristic information and the space-related characteristic information into the target power prediction model to obtain the power prediction result of the current distribution network output by the target power prediction model.

[0089] Furthermore, the power prediction system inputs the time-related characteristic information and the space-related characteristic information into the target power prediction model to obtain the power prediction results of the current distribution network output by the target power prediction model, wherein the power prediction results in the embodiment of the present invention include power generation prediction, power load prediction, power balance prediction, power flow prediction, power quality prediction, etc.

[0090] Power generation forecasting involves predicting the total power generation of various diverse power sources (such as solar, wind, hydro, and thermal power) within the current distribution network over a future time period. Power load forecasting involves predicting the power load of different types of users (such as industrial, commercial, and residential) within the current distribution network's coverage area. For example, based on the temporal correlation of power consumption in a city's commercial district and the spatial correlation between commercial venues and surrounding power sources, the power load of that district can be predicted on weekend nights. Power balance forecasting involves analyzing the power balance of the distribution network over a specific future time period based on the results of power generation and load forecasting. This involves determining whether the power generated by the power source can meet the power demand of the load, as well as any potential power surpluses or shortfalls. Power flow forecasting involves predicting the flow of power within the distribution network, including the direction and magnitude of active and reactive power on each line. For example, based on the distribution of power sources and loads around a transmission line and their temporal variations, the power flow along that line can be predicted during peak hours. Power quality prediction involves predicting power quality indicators of the current distribution network, such as voltage deviation, frequency deviation, and harmonic content. By combining temporal and spatial correlation characteristics, we can analyze changing trends in power quality across different time periods and regions. For example, we can predict the potential voltage fluctuations and increased harmonic content in an industrial zone during a specific production period due to the large number of nonlinear loads.

[0091] The embodiment of the present invention extracts a power characteristic vector that can reflect the power output from the operating data of multiple power sources based on the operating characteristics, and extracts a load characteristic vector that can reflect the change in load demand from the load data of multiple loads based on the load characteristics. This can better capture the changing patterns of power output and load demand. At the same time, by analyzing the associated characteristic information of the power characteristic vector and the load characteristic vector in terms of time distribution and spatial distribution, it can better capture the complex nonlinear relationship between power output and load demand. The target power prediction model adapted to the current situation can be accurately obtained through the complexity of the power path and the amount of power transmission. Then, the power prediction results can be accurately predicted based on the target power prediction model combined with the associated characteristic information, thereby improving the power prediction accuracy of the distribution network and ensuring the safe and stable operation of the distribution network.

[0092] In one embodiment, steps 101 to 104 are described as follows:

[0093] Step 101: Determine a power transmission path from a first node in a current distribution network to a second node in a target distribution network based on a power grid topology.

[0094] Optionally, the power prediction system determines a power transmission path from a first node s in the current distribution network to a second node t in the target distribution network according to the power grid topology, wherein the power transmission path includes multiple nodes. Therefore, the power transmission path can be expressed as es,t ={v1,v2,...,v n}, where v1 = s, v n =t.

[0095] Step 102 : determining the power path complexity based on the voltage level and power complexity of each node in the power path, and the path length and path rated capacity between two adjacent nodes in the power path.

[0096] Optionally, the power path complexity in an embodiment of the present invention is used to measure the complexity of different transmission paths from the current distribution network to the target distribution network. Therefore, the power prediction system obtains the voltage level and power complexity of each node in the power path, as well as the path length and path rated capacity between two adjacent nodes in the power path, where the power complexity is the number of edges connected to the node.

[0097] Furthermore, the power prediction system determines the power path complexity based on the voltage level and power complexity of each node in the power path, as well as the path length and path rated capacity between two adjacent nodes in the power path. The specific formula is as follows:

[0098]

[0099] Among them, PC(P) represents the power path complexity, Represents node v k With node v k+1 The path length between Represents node v k With node v k+1 The rated capacity of the path between Represents node v k+1 The power complexity (with node v k+1 The number of connected edges), U vk Represents node v k The voltage level, Represents node v k+1 Voltage level, U base Indicates the preset reference voltage value.

[0100] Step 103: Determine a power transmission tendency factor of the power transmission path based on the voltage level of the first node, the voltage level of the second node, and the path reactance and path resistance of the power transmission path, and determine the power transmission path direction according to the power transmission tendency factor.

[0101] Optionally, the power prediction system obtains the voltage level U of the first node s s , the voltage level U of the second node t t and power transmission path e s,tThe path reactance X s,t and the path resistance R s,t .

[0102] Furthermore, the power prediction system calculates the voltage level U of the first node s according to the voltage level U of the first node s. s , the voltage level U of the second node t t and power transmission path e s,t The path reactance X s,t and the path resistance R s,t , calculate the power transmission path e s,t Power Transmission Propensity Factor (PTF) s,t , the specific formula is as follows:

[0103]

[0104] Among them, PTF s,t represents the power transmission path e between node s and node t s,t The power transmission tendency factor PTF is, j represents the imaginary unit. s,t If the real part of is greater than 0, it means that the power transmission path e s,t The power transmission direction is from node s to node t; if the power transmission tendency factor PTF s,t The real part of is less than 0, which means that the power transmission path e s,t The power transmission direction is from node t to node s.

[0105] By calculating the power transmission tendency factor PTF of all edges s,t By analyzing the current power distribution network and the target power distribution network connected to it, the directions of the power transmission paths can be determined.

[0106] Step 104 : determining the power transmission amount of the power transmission path in the power transmission path direction based on the power transmission tendency factor, the path reactance, and the path resistance of the power transmission path.

[0107] Furthermore, the power prediction system is based on the power transmission path e s,t Power Transmission Propensity Factor (PTF) s,t , path reactance X s,t and the path resistance R s,t , determine the power transmission path e s,t The specific formula for the amount of power transmitted in the direction of its power transmission path is as follows:

[0108]

[0109] Among them, P s,t Indicates the power transmission path e s,t The amount of power transmitted in the direction of its power transmission path.

[0110] The embodiments of the present invention accurately determine the power path complexity and the amount of power transmission from the power grid topology. This allows the target power forecast model to be accurately derived based on the power path complexity and the amount of power transmission. Furthermore, the target power forecast model, combined with associated feature information, accurately predicts power forecast results. Consequently, the embodiments of the present invention improve the accuracy of power forecasts for distribution networks and ensure their safe and stable operation.

[0111] In one embodiment, steps 201 to 204 are described as follows.

[0112] Step 201 : determining a power fluctuation entropy characteristic at any moment based on the power supply active power and the power supply reactive power of the multiple power sources at any moment based on the power fluctuation characteristic.

[0113] Optionally, the power forecasting system performs preliminary processing on various types of operating data collected from multiple power sources, including basic electrical parameters (such as the instantaneous value of voltage, the effective value of current, the fluctuation range of frequency, the real-time value of active power, the amplitude of reactive power, and the size of apparent power, etc.), data on changes in generated power over time, and real-time status information of power sources (such as power-on time, number of fault alarms, power supply temperature, etc.). For example, for parts with missing values, the weighted average method based on adjacent data is used to fill them; for outliers, a detection model based on data density is constructed to identify them. If the data density around a data point is lower than a specific threshold, it is determined to be an outlier and corrected.

[0114] Furthermore, considering that the power output power will fluctuate over time, the power forecasting system predicts the power output power at any time t according to the power fluctuation characteristics. s The power fluctuation entropy is calculated based on the active power of the power supply to determine the power fluctuation entropy at any time t s The active power fluctuation entropy characteristics are as follows:

[0115]

[0116] Among them, H P (t s ) represents any time t s The active power fluctuation entropy characteristics, represents the time sliding window [(t s -n s ),t s ]The average value of internal active power, n s Indicates the length of the time sliding window.

[0117] Furthermore, the power forecasting system predicts the power fluctuation characteristics at any time t sThe power fluctuation entropy is calculated based on the reactive power of the power supply to determine the power fluctuation entropy at any time t s The reactive power fluctuation entropy characteristics are as follows:

[0118]

[0119] Among them, H Q (t s ) represents any time t s The reactive power fluctuation entropy characteristics, represents the time sliding window [(t s -n s ),t s ]The average value of the internal reactive power, n s Indicates the length of the time sliding window.

[0120] Therefore, at any time t s Active power fluctuation entropy characteristic H P (t s ) and reactive power fluctuation entropy characteristics H Q (t s ) is the multi-source power supply at any time t s Determine the power fluctuation entropy characteristics H(t s ).

[0121] Step 202 : reconstruct the phase space of the voltage signal of the multi-element power source at any moment based on the phase space characteristics combined with the delay time and the embedding dimension to determine the power source phase space characteristics at any moment.

[0122] Furthermore, the power forecasting system combines the phase space characteristics with the delay time and embedding dimension to predict the multi-source power supply at any time t s The voltage signal v(t s ) to reconstruct the phase space and construct the multi-source power supply at any time t s The phase space vector under , the specific formula is as follows:

[0123]

[0124] in, It means that the multi-source power supply at any time t s The phase space vector under s Denotes the delay time, d s represents the embedding dimension.

[0125] Furthermore, the power forecasting system is based on the multi-source power supply at any time t s The phase space vector under the condition of , calculate the distance characteristics of the phase space vector The specific formula is as follows:

[0126]

[0127] Where w1 and w2 represent the indices of the phase space vector.

[0128] Furthermore, the power forecasting system sets a preset number of minimum distance features The value is determined as the multi-source power supply at any time t s The power supply phase space characteristics under

[15] are shown in FIG1 , wherein the preset number is set according to actual conditions, such as 5, 10, etc. Therefore, the power supply phase space characteristics can reflect the distribution and evolution of the voltage signal of the multi-source power supply in the phase space at any moment.

[0129] Step 203 : Determine the power state probability transition feature at any moment based on the probability characteristics of the power state and the probability of occurrence of each power state.

[0130] Optionally, the power supply status information in the embodiment of the present invention, such as normal operation, minor fault, major fault, etc., exists in different probability forms. b states, and the probability p of each state occurring is calculated based on historical data i (t s ),i=1,2,...,L b .

[0131] Furthermore, the power forecasting system combines the probability characteristics of the power supply state with the probability p of each state to occur. i (t s ), determine any time t s The power state probability transition characteristics under , the specific formula is as follows:

[0132]

[0133] Among them, Th ij (t s ) represents any time t s Power state probability transition characteristics under .

[0134] In step 204 , the power fluctuation entropy feature, the power phase space feature, and the probability transfer feature at any moment are integrated to obtain the power feature vector of the multi-element power source at any moment.

[0135] Furthermore, the power forecasting system will s The power fluctuation entropy characteristic H(t s ), power supply phase space characteristics and probability transfer feature Th ij (t s ) are integrated to obtain the multi-source power supply at any time t s The power eigenvector under Specifically:

[0136]

[0137] The embodiment of the present invention extracts a power supply characteristic vector that can reflect the power supply output from the operating data of the multi-source power supply based on the operating characteristics. Therefore, it can better capture the changing rules of the power supply output, improve the power prediction accuracy of the distribution network, and ensure the safe and stable operation of the distribution network.

[0138] In one embodiment, steps 205 to 208 are described as follows:

[0139] Step 205 : Decompose the load active power and negative load reactive power of the time series based on the time characteristics, and obtain the intrinsic mode components and residual components based on the decomposition of the load active power and negative load reactive power to determine the power energy characteristics of the multi-element load.

[0140] Optionally, the power forecasting system collects various data of multiple loads, such as active power P(t u ), reactive power Q(t u ), voltage V(t u ), current I(t u ) and the corresponding timestamp t u etc. to remove noise interference to ensure the accuracy of the data.

[0141] Optionally, considering that load data has obvious time characteristics, the power forecasting system can calculate the load active power P(t u ) is decomposed to obtain the load active power P(t u ) of multiple intrinsic mode components and residual components, such as the load active power P(t u ) is decomposed into N u eigenmode components and a residual component, namely Furthermore, the power forecasting system determines the active power energy ratio characteristics of each intrinsic mode component in the load active power based on multiple intrinsic mode components and residual components of the load active power. The specific formula is as follows:

[0142]

[0143] in, Represents the i-th intrinsic mode component IMFP in the load active power i (t) represents the active power energy ratio characteristic, and T1 represents the time series length of the load active power.

[0144] Furthermore, the power forecasting system can predict the load reactive power Q(tu ) is decomposed to obtain the load reactive power Q(t u ) multiple eigenmode components and residual components, in one embodiment, the time series load reactive power Q(t u ) is decomposed into N u eigenmode components and a residual component, namely Furthermore, the power forecasting system determines the reactive power energy ratio characteristics of each intrinsic mode component in the load reactive power based on multiple intrinsic mode components and residual components of the load reactive power. The specific formula is as follows:

[0145]

[0146] in, Represents the i-th intrinsic mode component IMFQ in the load reactive power i (t) represents the reactive power energy ratio characteristic, and T2 represents the time series length of the load reactive power.

[0147] Furthermore, the power prediction system determines the active power energy proportion characteristics and the reactive power energy proportion characteristics as the power energy characteristics of the multi-element load.

[0148] Step 206 , based on seasonal characteristics, the load active power and negative load reactive power of the time series are time segmented according to seasons, and the power seasonal characteristics of the multivariate load are determined according to the seasonal correlation coefficient of the active power and negative load reactive power of each time segment.

[0149] Furthermore, in order to reflect the seasonal characteristics of the load and construct the power seasonal characteristics, the power forecasting system divides the load active power of the time series into time segments according to the season with a year as a cycle according to the seasonal characteristics to obtain the active power of each time segment. In one embodiment, it can be divided into 4 time segments according to the season.

[0150] Furthermore, the power forecasting system determines the seasonal correlation coefficients of the active power in different time segments based on the load data of the active power in different time segments, and determines the seasonal correlation coefficients of the active power in different time segments as the power seasonal characteristics of the active power of the multivariate load. The specific formula is as follows:

[0151]

[0152] Among them, s1 represents the time segment of the s1 season, and s2 represents the time segment of the s2 season. represents the seasonal correlation coefficient of active power in season s1 and season s2, T s Indicates the length of each season's time segment, P s1(t) represents the load data of active power in season s1, The mean value of the load data of active power in season s1, P s2 (t) represents the load data of active power in season s2, Indicates the mean value of the load data of active power in season s2.

[0153] Similarly, the power forecasting system divides the load reactive power of the time series into time segments according to the seasons based on seasonal characteristics to obtain the reactive power of each time segment.

[0154] Furthermore, the power forecasting system determines the seasonal correlation coefficients of the reactive power in different time segments based on the reactive power load data in different time segments, and determines the seasonal correlation coefficients of the reactive power in different time segments as the power seasonal characteristics of the reactive power of the multivariate load. The specific formula is as follows:

[0155]

[0156] Among them, s1 represents the time segment of the s1 season, and s2 represents the time segment of the s2 season. The seasonal correlation coefficient of reactive power between seasons s1 and s2, T s Indicates the length of each season's time segment, Q s1 (t) represents the reactive power load data in season s1, represents the mean value of reactive power load data in season s1, Q s2 (t) represents the reactive power load data in season s2, Represents the mean value of reactive power load data in season s2.

[0157] Step 207 , clustering different industries based on industry characteristics, and performing statistical analysis on the load active power and negative load reactive power corresponding to the clustered industry categories to obtain power industry clustering characteristics of multiple loads.

[0158] Optionally, based on the industry characteristics of the load, the power forecasting system clusters different industries to obtain different industry categories. Therefore, the local density of each load sample point is first determined. The specific formula is as follows:

[0159]

[0160] Among them, ρ i Represents the load sample point x i The local density, ρ i Represents the local density of the i-th load sample point; i and j represent the index of the load sample point, N hIndicates the number of samples of load data, where samples refer to load data samples of different industries; d ij Represents the load sample point x i and load sample point x j The distance between c Indicates the cutoff distance.

[0161] After calculating the local density ρ of all load sample points i Finally, the power forecasting system identifies points with high local density and a large distance from points with even higher local density as initial cluster centers. After determining the initial cluster centers, the load data is clustered. The power forecasting system assigns each load data sample to the category of the cluster center closest to it and updates the location of the cluster center until the cluster center no longer changes significantly or the preset number of iterations is reached, resulting in the classification of each industry.

[0162] Furthermore, statistical analysis is performed on the active power of loads and reactive power of negative loads corresponding to different industry categories obtained through clustering. For example, for each industry category, the mean, variance, etc. of the active power and reactive power of all load data samples in the category are calculated to obtain the power industry clustering characteristics of multivariate loads.

[0163] Step 208 : The power energy characteristics, power seasonal characteristics, and power industry clustering characteristics of the multi-element load are integrated to obtain a load characteristic vector of the multi-element load.

[0164] Furthermore, the power forecasting system integrates the power energy characteristics, power seasonal characteristics and power industry clustering characteristics of multiple loads to obtain the load characteristic vector of multiple loads.

[0165] The embodiment of the present invention extracts a load characteristic vector that can reflect the change of load demand from the load data of the multi-element load according to the load characteristics. Therefore, it can better capture the changing pattern of load demand, improve the power prediction accuracy of the distribution network, and ensure the safe and stable operation of the distribution network.

[0166] In one embodiment, steps 301 to 304 are described as follows:

[0167] Step 301 : performing feature sequence extraction on the power source feature vector and the load feature vector at each time point in a time sliding window of a preset window length, to obtain a power source feature vector sequence and a load feature vector sequence, respectively.

[0168] Optionally, since the collection time intervals of power data and load data may be different, it is necessary to unify the power data and load data to the same time scale. Therefore, the power feature vector and load characteristic vector Time alignment processing is performed, wherein the embodiment of the present invention adopts a linear interpolation method to complete the data with inconsistent time intervals so that the two feature vectors correspond one to one at the time point.

[0169] Furthermore, in order to analyze the power characteristic vector and load characteristic vector For the correlation between different time intervals, the embodiment of the present invention sets a time sliding window with a preset window length, wherein the preset window length of the time sliding window is W time units. W , from t W -W+1 to time point t W In the time sliding window, the power feature vectors are extracted respectively. Power characteristic vector sequence And extract the load characteristic vector The load characteristic vector sequence Specifically:

[0170]

[0171] Step 302 : Determine a similarity metric value between the power source feature vector sequence and the load feature vector sequence within a time sliding window based on the vector similarity and cosine similarity between the first feature vector in the power source feature vector sequence and the second feature vector in the load feature vector sequence.

[0172] Furthermore, in order to measure the power characteristic vector sequence and load characteristic vector sequence Similarity metric within the time sliding window, the power forecasting system calculates the power feature vector sequence The first eigenvectors and load eigenvector sequences in The vector similarity of each second eigenvector in is as follows:

[0173]

[0174] in, Represents the power characteristic vector sequence The first eigenvector of , Represents the load characteristic vector sequence The i-th second eigenvector in , represents the first eigenvector With the second eigenvector The degree of similarity between the vectors, m represents the dimension of the feature vector, represents the first eigenvector The jth component of represents the second eigenvector The jth component of .

[0175] Furthermore, the power forecasting system calculates the power source characteristic vector sequence The first eigenvectors and load eigenvector sequences in The cosine similarity of each second eigenvector in The specific formula is as follows:

[0176]

[0177] Furthermore, the power forecasting system uses the vector similarity and cosine similarity Calculate the comprehensive similarity between the first eigenvector and the second eigenvector. The specific formula is as follows:

[0178]

[0179] Among them, S i represents the first eigenvector With the second eigenvector The comprehensive similarity between them, α represents the adjustment parameter.

[0180] Furthermore, the power prediction system calculates the similarity S between the first eigenvector and the second eigenvector. i , calculate the power eigenvector sequence and load characteristic vector sequence Similarity measure S within the time sliding window window , the specific formula is as follows:

[0181]

[0182] Step 303: Determine a time series based on the similarity metric value of each time point in the time sliding window.

[0183] Step 304: Determine time-related feature information based on the time series and its sequence features.

[0184] Furthermore, for each time point in the sliding time window, the power forecasting system calculates the similarity metric value S corresponding to each time point window , and get the time series S window (t), where the time series S window (t) reflects the power characteristic vector Power characteristic vector sequence The degree of similarity at different points in time.

[0185] Furthermore, in order to further extract the key features of time correlation, the time series Swindow (t) is processed and the time series S is calculated. window (t) first-order derivative and second-order derivative, get the time series S window The sequence characteristics of (t) are as follows:

[0186]

[0187] in, represents the first-order derivative, represents the second-order derivative, and Δt represents the time interval.

[0188] Furthermore, the power forecasting system converts the time series S window (t), first-order derivative and the second-order derivative Determine the final time-related feature information.

[0189] By analyzing the time-correlated characteristic information of the power source characteristic vector and the load characteristic vector in the time distribution, the embodiment of the present invention can better capture the complex nonlinear relationship between power output and load demand, improve the power prediction accuracy of the distribution network, and ensure the safe and stable operation of the distribution network.

[0190] In one embodiment, steps 305 to 309 are described as follows:

[0191] Step 305 : Taking each power source point in the multi-source power source and each load point in the multi-load power source as the center, determine a first spatial neighborhood range of each power source point and a second spatial neighborhood range of each load point.

[0192] Optionally, the multiple power sources and multiple loads in the embodiment of the present invention have their own location coordinates in geographic space, and the actual geographic location information (such as longitude and latitude) of the multiple power sources and multiple loads is mapped into a two-dimensional plane coordinate system. For each power source point and load point, in addition to the existing power source characteristic vector and load characteristic vector In addition, the embodiment of the present invention also assigns the power coordinates (x G ,y G ), assign the load coordinate (x L ,y L ). At the same time, the power characteristic vector and load characteristic vector Normalization is performed so that the values of each dimension are within the same order of magnitude.

[0193] Furthermore, the power forecasting system determines a first spatial neighborhood range of each power source point and a second spatial neighborhood range of each load point with each power source point and each load point as the center, wherein the spatial neighborhood range can be a circular area with a radius of r and each power source point and each load point as the center.

[0194] Step 306: Obtain a set of load points within a first spatial neighborhood of each power point, and a set of power points within a second spatial neighborhood of each load point.

[0195] Furthermore, for each power point Power i , the power forecasting system determines the set of all load points within its first spatial neighborhood For each load point i , the power forecasting system determines the set of all power points within its second spatial neighborhood

[0196] Step 307: Determine a first difference metric value based on the power characteristic vector of each power point and the load characteristic vector of any load point in its load point set, and determine a second difference metric value based on the load characteristic vector of each load point and the power characteristic vector of any power point in its power point set.

[0197] Furthermore, in order to measure the spatial correlation between each power point and the load points in its first spatial neighborhood, and the correlation between each load point and the power points in its second spatial neighborhood, for each power point Power i , the power forecast system is based on the power of each power point i The first difference metric is calculated by the power source eigenvector and the load eigenvector of any load point p in the load point set within the first spatial neighborhood. The specific formula is as follows:

[0198]

[0199] in, Indicates power point Power i The power characteristic vector and the load characteristic vector of load point p The first difference metric between them, n represents the dimension of the feature vector, G iz Indicates power point Power i The power characteristic vector The zth component of L pz Represents the load characteristic vector of load point p The zth component of .

[0200] Furthermore, each load point Loadi , the power forecasting system is based on each load point Load i The load characteristic vector of and the power characteristic vector of any power point l in the power point set within the second spatial neighborhood range are used to calculate the second difference measure value. The specific formula is as follows:

[0201]

[0202] in, Indicates each load point Load i The load characteristic vector and the power characteristic vector of power point l The second difference measure between iz Indicates load point Load i The load characteristic vector The zth component of G lz Represents the power characteristic vector of power point l The zth component of .

[0203] Step 308 : determining a first average difference metric value of each power point based on the first difference metric value of each power point, and determining a second average difference metric value of each load point based on the second difference metric value of each load point.

[0204] Furthermore, the power prediction system determines a first average difference metric value of each power source point according to the first difference metric value of each power source point. The specific formula is as follows:

[0205]

[0206] in, Represents a set of load points The number of medium load points.

[0207] Furthermore, the power forecasting system determines the second average difference metric value of each load point according to the second difference metric value of each load point. The specific formula is as follows:

[0208]

[0209] in, Represents a collection of power points The number of power points.

[0210] Step 309 : Fusing the first average difference metric value of each power source point and the second average difference metric value of each load point to obtain spatial correlation feature information.

[0211] Furthermore, the power forecasting system calculates the first average difference metric value of all power supply points. and the second mean difference measure for all load points The spatial correlation feature information is obtained by fusion. In one embodiment, it includes V power points and O load points. Therefore, the spatial correlation feature information can be expressed as:

[0212]

[0213] By analyzing the spatial correlation characteristic information of the power eigenvector and the load eigenvector in spatial distribution, the embodiment of the present invention can better capture the complex nonlinear relationship between power output and load demand, improve the power prediction accuracy of the distribution network, and ensure the safe and stable operation of the distribution network.

[0214] In one embodiment, steps 401 to 404 are described as follows.

[0215] Step 401 : normalize the power path complexity and each power transmission amount, and perform row fusion on the normalized results to obtain a first eigenvector.

[0216] Optionally, the power prediction system calculates the power path complexity PC(e s,t ) and the power transmission amount P in each power transmission path direction s,t In one embodiment, there are m different power transmission paths, and the normalized power path complexity is PC norm (e s,t ), the normalized power transmission amount is

[0217] Furthermore, the power forecasting system will PC norm (e s,t )and Fusion is performed to obtain the first eigenvector Here, m represents the number of the starting node, and d represents the number of the ending node.

[0218] Step 402 : Based on the difference between the first eigenvector and the second eigenvector corresponding to each power prediction model in the preset model library, determine the degree of adaptation of the first eigenvector to the second eigenvector of each power prediction model.

[0219] Optionally, the second eigenvector corresponding to each power prediction model in the preset model library of the embodiment of the present invention is is the number of power forecast models in the model library, where the second eigenvector of each power forecast model is It is also generated by normalizing the power path complexity and power transmission volume of each power prediction model in different scenarios.

[0220] Furthermore, the power forecasting system determines the first eigenvector and the second eigenvector corresponding to each power prediction model in the preset model library The specific formula is as follows:

[0221]

[0222] Among them, d xk represents the first eigenvector and the kth second eigenvector The degree of difference in feature dimension, M represents the first eigenvector The dimension, X i represents the first eigenvector The i-th element of Y ki represents the kth second eigenvector The i-th element of .

[0223] Furthermore, the power forecasting system is based on the first eigenvector and the second eigenvector corresponding to each power prediction model in the preset model library The degree of difference between The second eigenvector corresponding to each power forecast model The specific formula is as follows:

[0224]

[0225] Among them, a k represents the first eigenvector and the kth second eigenvector It should be noted that the fitness a k The larger the value of and the kth second eigenvector The higher the matching degree, the stronger the adaptability of the power forecasting model to the current input data.

[0226] Step 403 : determining a final matching degree between the first eigenvector and the second eigenvector of each power prediction model based on the adaptation degree and the model reliability vector of the power prediction model.

[0227] Furthermore, in order to accurately determine the best matching target power forecast model, it is necessary to consider the differences in the reliability of different power forecast models in actual applications. Therefore, the power forecast system is based on the first eigenvector The second eigenvector corresponding to each power forecast model The degree of adaptation and the model reliability vector of the power forecast model Among them, rk Represents the reliability coefficient of the kth power prediction model, and determines the final matching degree between the first eigenvector and the second eigenvector of each power prediction model. The specific formula is as follows:

[0228]

[0229] Among them, s k represents the first eigenvector The final matching degree with the k-th power forecast model.

[0230] Step 404 , traverse the final matching degrees, and determine the power prediction model corresponding to the final matching degree with the largest value as the target power prediction model.

[0231] Furthermore, the power forecasting system calculates the final matching degree s for each k Traverse and find the final matching degree s with the largest value k The corresponding power forecast model will take the maximum final matching degree s k The corresponding power forecast model is determined as the target power forecast model. That is:

[0232]

[0233] in, Indicates the index of the best matching model, index The corresponding power prediction model is the target power prediction model matched in the model library according to the current power path complexity and power transmission volume.

[0234] The present invention uses the complexity of power paths and the amount of power transmitted to accurately determine a target power forecast model tailored to the current situation. This model, combined with associated feature information, accurately predicts power forecast results. Consequently, the present invention improves the accuracy of power forecasts for distribution networks and ensures their safe and stable operation.

[0235] See also Figure 3 , Figure 3 This is a diagram of an embodiment of an electronic device provided by an embodiment of the present invention. Figure 3 As shown, an embodiment of the present invention provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, the following steps are implemented:

[0236] Based on the grid topology of the current distribution network, the complexity of the power paths between the current distribution network and the connected target distribution network and the power transmission amount in each power transmission path direction are obtained, as well as the operating data of the multiple power sources and the load data of the multiple loads in the current distribution network are obtained;

[0237] Extracting features from the operating data based on the operating characteristics of the multiple power sources to obtain a power feature vector that can reflect the power output; and extracting features from the load data based on the load characteristics of the multiple loads to obtain a load feature vector that can reflect changes in load demand;

[0238] Performing correlation analysis based on the power source characteristic vector and the load characteristic vector to determine the time correlation characteristic information of the power source characteristic vector and the load characteristic vector in time distribution, as well as the spatial correlation characteristic information of the power source characteristic vector and the load characteristic vector in space distribution;

[0239] Based on the complexity of the power path and the amount of power transmission, a model is matched in a preset model library to obtain a target power prediction model;

[0240] The time-related feature information and the space-related feature information are input into the target power prediction model to obtain the power prediction result of the current distribution network output by the target power prediction model; the target power prediction model is trained based on the sample-related feature information and its corresponding power label results.

[0241] See also Figure 4 , Figure 4 Detailed description of an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. Figure 4 As shown, this embodiment provides a computer-readable storage medium 400 on which a computer program 311 is stored. When the computer program 311 is executed by a processor, the following steps are implemented:

[0242] Based on the grid topology of the current distribution network, the complexity of the power paths between the current distribution network and the connected target distribution network and the power transmission amount in each power transmission path direction are obtained, as well as the operating data of the multiple power sources and the load data of the multiple loads in the current distribution network are obtained;

[0243] Extracting features from the operating data based on the operating characteristics of the multiple power sources to obtain a power feature vector that can reflect the power output; and extracting features from the load data based on the load characteristics of the multiple loads to obtain a load feature vector that can reflect changes in load demand;

[0244] Performing correlation analysis based on the power source characteristic vector and the load characteristic vector to determine the time correlation characteristic information of the power source characteristic vector and the load characteristic vector in time distribution, as well as the spatial correlation characteristic information of the power source characteristic vector and the load characteristic vector in space distribution;

[0245] Based on the complexity of the power path and the amount of power transmission, a model is matched in a preset model library to obtain a target power prediction model;

[0246] The time-related feature information and the space-related feature information are input into the target power prediction model to obtain the power prediction result of the current distribution network output by the target power prediction model; the target power prediction model is trained based on the sample-related feature information and its corresponding power label results.

[0247] On the other hand, the present invention further provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the power forecasting method for a distribution network with multiple sources and loads provided by the above methods. The power forecasting method for a distribution network with multiple sources and loads includes:

[0248] Based on the grid topology of the current distribution network, the complexity of the power paths between the current distribution network and the connected target distribution network and the power transmission amount in each power transmission path direction are obtained, as well as the operating data of the multiple power sources and the load data of the multiple loads in the current distribution network are obtained;

[0249] Extracting features from the operating data based on the operating characteristics of the multiple power sources to obtain a power feature vector that can reflect the power output; and extracting features from the load data based on the load characteristics of the multiple loads to obtain a load feature vector that can reflect changes in load demand;

[0250] Performing correlation analysis based on the power source characteristic vector and the load characteristic vector to determine the time correlation characteristic information of the power source characteristic vector and the load characteristic vector in time distribution, as well as the spatial correlation characteristic information of the power source characteristic vector and the load characteristic vector in space distribution;

[0251] Based on the complexity of the power path and the amount of power transmission, a model is matched in a preset model library to obtain a target power prediction model;

[0252] The time-related feature information and the space-related feature information are input into the target power prediction model to obtain the power prediction result of the current distribution network output by the target power prediction model; the target power prediction model is trained based on the sample-related feature information and its corresponding power label results.

[0253] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0254] Example 2

[0255] The second embodiment of the present invention is different from the first embodiment in that:

[0256] If the functions are implemented in the form of 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 part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0257] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered 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.

[0258] 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), a fiber optic 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, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0259] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

Claims

1. A power forecasting system for a distribution network containing multiple sources and loads, characterized in that: It includes a distribution network management center, a data acquisition module, a feature extraction module, a correlation analysis module, a model matching module, and a power forecasting module. The distribution network management center is connected to the data acquisition module, the feature extraction module, the correlation analysis module, the model matching module, and the power forecasting module to manage each module. A data acquisition module is used to obtain the power path complexity between the current distribution network and the connected target distribution network and the power transmission volume in each power transmission path direction based on the power grid topology of the current distribution network, as well as the operating data of multiple power sources and load data of multiple loads in the current distribution network; A feature extraction module is used to extract features from operating data based on the operating characteristics of multiple power sources to obtain a power feature vector that can reflect the power output, and to extract features from load data based on the load characteristics of multiple loads to obtain a load feature vector that can reflect changes in load demand; A correlation analysis module is used to perform correlation analysis based on the power characteristic vector and the load characteristic vector to determine the time correlation characteristic information of the power characteristic vector and the load characteristic vector in time distribution and the spatial correlation characteristic information in spatial distribution; a model matching module for performing model matching in a preset model library based on the complexity of the power path and the power transmission amount of each power transmission amount to obtain a target power prediction model; The power prediction module is used to input time-related feature information and space-related feature information into the target power prediction model to obtain the power prediction result of the current distribution network output by the target power prediction model; the target power prediction model is trained based on the sample-related feature information and its corresponding power label results.

2. A method for predicting power in a distribution network with multiple sources and loads, implemented based on the power prediction system for a distribution network with multiple sources and loads as claimed in claim 1, characterized in that: Based on the grid topology of the current distribution network, the complexity of the power paths between the current distribution network and the connected target distribution network and the power transmission amount in each power transmission path direction are obtained, as well as the operating data of the multiple power sources and the load data of the multiple loads in the current distribution network are obtained; Extracting features from the operating data based on the operating characteristics of the multiple power sources to obtain a power feature vector capable of reflecting power output, and extracting features from the load data based on the load characteristics of the multiple loads to obtain a load feature vector capable of reflecting changes in load demand; Performing a correlation analysis based on the power source characteristic vector and the load characteristic vector to determine time correlation characteristic information of the power source characteristic vector and the load characteristic vector in time distribution and space correlation characteristic information of the power source characteristic vector and the load characteristic vector in space distribution; Performing model matching in a preset model library based on the complexity of the power path and each of the power transmission amounts to obtain a target power prediction model; The time-related feature information and the space-related feature information are input into the target power prediction model to obtain the power prediction result of the current distribution network output by the target power prediction model; the target power prediction model is trained based on the sample-related feature information and its corresponding power label results.

3. The method for power forecasting in a distribution network with multiple sources and loads according to claim 2, characterized in that: The operating characteristics include power fluctuation characteristics, phase space characteristics and probabilistic characteristics of power supply status; The extracting features of the operating data based on the operating characteristics of the multi-source power supply to obtain a power feature vector capable of reflecting the power output includes: Determine the power fluctuation entropy feature at any moment based on the power fluctuation characteristics and the power supply active power and the power supply reactive power of the multi-element power supply at any moment; Reconstruct the phase space of the voltage signal of the multi-element power source at any moment based on the phase space characteristics combined with the delay time and the embedding dimension to determine the power source phase space characteristics at any moment; Determine the power state probability transition characteristics at any moment based on the probability characteristics of the power state and the probability of occurrence of each power state; The power fluctuation entropy characteristics, power supply phase space characteristics and probability transfer characteristics at any moment are integrated to obtain the power supply characteristic vector of the multi-element power supply at any moment.

4. The method for power forecasting in a distribution network with multiple sources and loads according to claim 2, wherein: The load characteristics include time characteristics, seasonal characteristics and industry characteristics; The extracting features of the load data based on the load characteristics of the multi-element load to obtain a load feature vector capable of reflecting changes in load demand includes: Decomposing the load active power and negative load reactive power of the time series based on the time characteristics, and obtaining the intrinsic modal component and the residual component based on the decomposition of the load active power and the negative load reactive power, to determine the power energy characteristics of the multi-element load; Based on the seasonal characteristics, the load active power and negative load reactive power of the time series are divided into time segments according to seasons, and the power seasonal characteristics of the multivariate load are determined according to the seasonal correlation coefficient of the active power and negative load reactive power in each time segment; Clustering different industries based on the industry characteristics, and performing statistical analysis on the load active power and negative load reactive power corresponding to the industry categories obtained by clustering to obtain the power industry clustering characteristics of the multivariate load; the statistical analysis includes mean and variance; The power energy characteristics, power seasonal characteristics and power industry clustering characteristics of the multi-load are integrated to obtain a load characteristic vector of the multi-load.

5. The method for power forecasting in a distribution network with multiple sources and loads according to claim 2, characterized in that: Performing a correlation analysis based on the power source characteristic vector and the load characteristic vector to determine time correlation characteristic information of the power source characteristic vector and the load characteristic vector in time distribution includes: At each time point in a time sliding window with a preset window length, feature sequence extraction is performed on the power feature vector and the load feature vector to obtain a power feature vector sequence and a load feature vector sequence, respectively; Determining a similarity metric value between the power source feature vector sequence and the load feature vector sequence within the time sliding window based on the vector similarity and cosine similarity between the first feature vector in the power source feature vector sequence and the second feature vector in the load feature vector sequence; Determining a time series based on a similarity measure value at each time point in the time sliding window; The time-related feature information is determined based on the time series and its sequence features.

6. The method for power forecasting in a distribution network with multiple sources and loads according to claim 2, characterized in that: Performing a correlation analysis based on the power source characteristic vector and the load characteristic vector to determine spatial correlation characteristic information of the power source characteristic vector and the load characteristic vector in spatial distribution includes: Taking each power source point in the multiple power sources and each load point in the multiple loads as the center, determine a first spatial neighborhood range of each power source point and a second spatial neighborhood range of each load point; Obtaining a set of load points within a first spatial neighborhood of each power point, and a set of power points within a second spatial neighborhood of each load point; Determining a first difference metric value based on a power source characteristic vector of each power source point and a load characteristic vector of any load point in the load point set, and determining a second difference metric value based on the load characteristic vector of each load point and a power source characteristic vector of any power source point in the power source point set; determining a first average difference metric value for each power source point based on the first difference metric value for each power source point, and determining a second average difference metric value for each load point based on the second difference metric value for each load point; The first average difference metric value of each power source point and the second average difference metric value of each load point are fused to obtain the spatial correlation feature information.

7. The method for power forecasting in a distribution network with multiple sources and loads according to claim 2, characterized in that: The performing model matching in a preset model library based on the complexity of the power path and each of the power transmission amounts to obtain a target power prediction model includes: Normalizing the power path complexity and each of the power transmission amounts, and performing row fusion on the normalized results to obtain a first eigenvector; Determining the degree of adaptation of the first eigenvector to the second eigenvector of each power prediction model based on the degree of difference between the first eigenvector and the second eigenvector corresponding to each power prediction model in the preset model library; the second eigenvector of each power prediction model is generated by normalizing the power path complexity and power transmission amount of each power prediction model under different scenarios; determining a final matching degree between the first eigenvector and the second eigenvector of each power prediction model based on the adaptation degree and the model reliability vector of the power prediction model; The final matching degrees are traversed, and the power prediction model corresponding to the final matching degree with the largest value is determined as the target power prediction model.

8. The method for predicting power in a distribution network containing multiple sources and loads according to any one of claims 2 to 7, characterized in that: The obtaining, based on the grid topology of the current distribution network, the complexity of the power paths between the current distribution network and the connected target distribution network and the power transmission amount in each power transmission path direction includes: determining a power transmission path from a first node in the current distribution network to a second node in the target distribution network based on the grid topology; determining the power path complexity based on the voltage level and power complexity of each node in the power path, and the path length and path rated capacity between two adjacent nodes in the power path; determining a power transmission tendency factor of the power transmission path based on the voltage level of the first node, the voltage level of the second node, and the path reactance and path resistance of the power transmission path, and determining a power transmission path direction according to the power transmission tendency factor; An amount of power transmission of the power transmission path in a direction of the power transmission path is determined based on a power transmission tendency factor, a path reactance, and a path resistance of the power transmission path.

9. An electronic device comprising: A memory and a processor, characterized in that a computer software program is stored on the memory, and when the processor reads and executes the computer software program, the power prediction method for a distribution network containing multiple sources and loads as described in any one of claims 2 to 8 is implemented.

10. A non-transitory computer-readable storage medium, characterized in that The storage medium stores a computer software program, which, when executed by a processor, implements the power prediction method for a distribution network containing multiple sources and loads according to any one of claims 2 to 8.

Images