A FID-based operation optimization method for distribution networks

Abstract

The present application provides a method for optimizing the operation of a distribution network based on FID, which relates to the technical field of power systems and includes: collecting power supply characteristic data, building a regional power supply network topology, clustering to obtain multiple load sub-regions, constructing a regional load network topology, performing new energy power generation prediction and load prediction to obtain multiple predicted powers and predicted loads, constructing a distribution optimization model, searching for an optimal distribution plan, outputting the optimal distribution plan, and performing distribution optimization control. Through the present application, the technical problems in the prior art, such as large power grid losses and low operation efficiency caused by the large-scale access of distributed power sources and the low level of distribution network automation, can be solved. By accurately matching based on the regional electricity load and the new energy power generation status, the scientificity and rationality of the regional distribution plan setting can be improved, thereby achieving the technical effects of reducing the regional power grid losses and improving the distribution operation efficiency.

Description

Technical Field

[0001] This application relates to the technical field of power systems, and in particular, to a method for optimizing the operation of a distribution network based on FID. Background Art

[0002] With the development of technology, the distribution network is undergoing a digital transformation process, including the application of Internet of Things technology, big data analysis, and the integration of artificial intelligence technology, to improve the operation efficiency and management level of the distribution network. With the rapid development of distributed renewable energy (such as solar energy and wind energy) and new energy vehicles, the role of the distribution network has changed from a simple power distribution network to a multi-energy complementary configuration platform. This requires the distribution network to be able to effectively integrate and manage these distributed resources to achieve more efficient and sustainable energy use. The new power system requires the distribution network to have higher flexibility and reliability to adapt to the changing energy demand and supply patterns. Currently, the large-scale access of distributed power sources brings more uncertainties and randomness to the distribution system, especially distributed power sources based on renewable energy, such as solar energy and wind energy, which have strong randomness due to being greatly affected by weather conditions, affecting the voltage stability and frequency stability of the power grid.

[0003] In summary, in the prior art, there are technical problems of large power grid losses and low operation efficiency due to the large-scale access of distributed power sources and the low automation level of the distribution network. Summary of the Invention

[0004] The purpose of this application is to provide a method for optimizing the operation of a distribution network based on FID to solve the technical problems in the prior art of large power grid losses and low operation efficiency due to the large-scale access of distributed power sources and the low automation level of the distribution network.

[0005] In view of the above problems, the present application provides a method for optimizing the operation of a distribution network based on FID. Among them, the method for optimizing the operation of a distribution network based on FID includes: collecting power supply characteristic data of a target area and building a regional power supply network topology; clustering multiple load points in the target area based on load characteristics to obtain multiple load sub-areas, and building a regional load network topology according to the multiple load sub-areas; based on the regional power supply network topology, predicting new energy power generation within a predetermined window through a power generation prediction channel to obtain multiple predicted powers; based on the regional load network topology, predicting loads within a preset window through load monitoring data and a load database to determine multiple predicted loads; using an FID device, the regional power supply network topology and the regional load network topology to fuse and build a distribution optimization model. In the distribution optimization model, with minimizing grid loss as the optimization index, searching for an optimal distribution plan according to the multiple predicted powers and the multiple predicted loads, and outputting an optimal distribution plan; through the FID device, performing distribution optimization control within a predetermined window according to the optimal distribution plan.

[0006] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0007] By collecting power supply characteristic data of a target area and building a regional power supply network topology; clustering multiple load points in the target area based on load characteristics to obtain multiple load sub-areas, and building a regional load network topology according to the multiple load sub-areas; based on the regional power supply network topology, predicting new energy power generation within a predetermined window through a power generation prediction channel to obtain multiple predicted powers; based on the regional load network topology, predicting loads within a preset window through load monitoring data and a load database to determine multiple predicted loads; using an FID device, the regional power supply network topology and the regional load network topology to fuse and build a distribution optimization model. In the distribution optimization model, with minimizing grid loss as the optimization index, searching for an optimal distribution plan according to the multiple predicted powers and the multiple predicted loads, and outputting an optimal distribution plan; through the FID device, performing distribution optimization control within a predetermined window according to the optimal distribution plan. That is to say, by accurately matching based on the regional electricity load and the new energy power generation status, the scientificity and rationality of the regional distribution plan setting can be improved, thereby achieving the technical effects of reducing the regional grid loss and improving the distribution operation efficiency.

[0008] The above description is only an overview of the technical solution of this application. In order to understand the technical means of this application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of this application more obvious and understandable, the following specific embodiments of this application are specifically given. It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of this application, nor is it used to limit the scope of this application. Other features of this application will become easily understandable through the following specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in this application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0010] Figure 1 It is a schematic flow chart of a method for optimizing the operation of a distribution network based on FID in this application;

[0011] Figure 2 It is a schematic flow chart of determining multiple predicted loads in a method for optimizing the operation of a distribution network based on FID in this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0012] By providing a method for optimizing the operation of a distribution network based on FID, this application solves the technical problems in the prior art that due to the large-scale access of distributed power sources and the low level of distribution network automation, the power grid loss is large and the operation efficiency is low. By accurately matching based on the regional electricity load and the new energy generation status, the scientificity and rationality of the regional power distribution plan setting can be improved, thereby achieving the technical effects of reducing the regional power grid loss and improving the power distribution operation efficiency.

[0013] Next, the technical solutions in this application will be clearly and completely described with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments of this application. It should be understood that this application is not limited by the exemplary embodiments described here. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of this application. Additionally, it should be noted that for the convenience of description, only the parts related to this application are shown in the drawings rather than all.

[0014] Embodiment, please refer to the attached Figure 1 This application provides a method for optimizing the operation of a distribution network based on FID. Among them, the method for optimizing the operation of a distribution network based on FID specifically includes the following steps:

[0015] Step 1: Collect the power supply characteristic data of the target area and build the regional power supply network topology.

[0016] Specifically, collect the power supply characteristic data in the target area, such as power supply equipment, equipment location coordinates, and distribution line characteristics. Power supply equipment refers to all power source points in the area, including traditional thermal power plants and new energy generation equipment, such as wind power generation and solar power generation equipment. The equipment location coordinates record the exact positions of these power supply equipment in the physical space. The distribution line characteristics are a key part of the power supply characteristic data, including the resistivity, cross-sectional area, and length of the line. The resistivity is calculated based on parameters such as the material and cross-sectional area of the line and determines the power loss during power transmission. The cross-sectional area and length of the line affect the power transmission capacity of the line. A larger cross-sectional area and a shorter length help reduce energy loss and improve power transmission efficiency. Taking the power supply equipment as the graph nodes and the distribution lines as the edges of the graph, determine the positions of the nodes in the topology according to the equipment location coordinates, and generate an intuitive regional power supply network topology. By building an accurate power supply network topology, the operating state of the power grid can be better understood.

[0017] Step 2: Cluster multiple load points in the target area based on load characteristics to obtain multiple load sub-areas, and build a regional load network topology according to the multiple load sub-areas.

[0018] Specifically, collect the load characteristics of each load point in the target area, including load access points, load sizes, load types (such as residential, commercial, industrial), load usage patterns (such as daytime peak, nighttime low peak), load volatility (such as seasonal changes, sudden peaks), and load persistence, etc. Use clustering algorithms, such as the K-means algorithm, to analyze the collected load characteristic data, cluster similar load points together, and generate multiple load sub-areas. After clustering, the load points within each sub-area have similar electricity consumption characteristics. For example, in an urban area, by analyzing the load point characteristics of residential areas, commercial areas, and industrial areas, they can be classified into different load sub-areas respectively. For example, factories with daytime electricity peaks can be classified into one sub-area, and residences with nighttime electricity peaks can be classified into another sub-area. Then, build a load network topology according to the load sub-areas. Each load sub-area is regarded as a node, and the power lines connecting these nodes are used as edges. Use graph computing methods to build a regional load network topology map according to the geographical location information of the load points. This topology map shows the connection relationships between different load sub-areas and their interaction methods with the power grid. Through clustering analysis, the load demand of each load sub-area can be predicted more accurately, thereby improving the accuracy of load prediction.

[0019] Step 3: Based on the regional power supply network topology, perform new energy power generation prediction within a predetermined window through the power generation prediction channel to obtain multiple predicted powers.

[0020] Specifically, collect the attribute information of each device in the regional power supply network topology, randomly select a device, perform homologous information retrieval in the power big data to obtain a sample data set. Divide the sample data set into N equal parts, and using the method of sampling with replacement, each time select N parts of data to generate the first training set. Repeat this process N times to obtain N different training sets. Use the N training sets to perform supervised training on the BP neural network, using the sample power generation factor set as the input and the sample power generation as the supervision signal. Perform cross-validation to ensure the generalization ability of the model, and obtain N converged power generation prediction units. Based on the N converged power generation prediction units, parallelly build the first power generation prediction model, output the average value of the output results of the N prediction units to reduce the prediction error. Establish the mapping relationship between the first power supply device and the first power generation prediction model, and according to the classification decision principle, use the mapping relationship to construct the power generation prediction channel. The power generation prediction channel includes multiple power generation prediction units, which correspond one-to-one with the new energy power supply devices. Each new energy power supply device has a corresponding power generation prediction unit, and these units make predictions according to the attributes of the device (such as type, specification, working principle, etc.). Through these prediction units, the expected power generation of each new energy power supply device in the future period can be obtained. According to these predictions, multiple predicted power values can be obtained. These predicted power values represent the electric power that each new energy power generation device is expected to generate in the future period. Since there may be multiple new energy power generation devices in the power grid, multiple predicted power values will be obtained. By understanding the expected power of new energy power generation in advance, grid operators can better respond to load changes.

[0021] Step 4: Based on the regional load network topology, perform load prediction within a preset window through the load monitoring data and the load database to determine multiple predicted loads.

[0022] Specifically, in the regional load network topology, randomly select the first load sub-region, and call the load database with the preset window and the first load sub-region as constraints. Retrieve and obtain the historical starting load set and the historical load set, perform mean calculation and deviation calculation, and according to the predetermined deviation threshold, eliminate the unreliable data greater than the threshold, and use the remaining data as the sample data, that is, the sample starting load set and the sample load set. Configure the gradually decreasing weights according to the time series characteristics. As time goes by, the contribution of the sample load data to the current moment load prediction gradually decreases. The sample starting load set and the sample load set are weighted and fused to obtain the standard starting load and the standard load.

[0023] Load forecasting within a preset window is performed through load monitoring data and a load database. The load monitoring data refers to the power grid load data monitored in real time, while the load database contains historical and real-time load data, which reflect the power demand of the load sub-region at different time points. By combining these data, the load demand for a future period can be predicted. To improve the accuracy of load forecasting, the first real-time starting load of the first load sub-region is obtained through non-intrusive power load monitoring method, then the first starting load deviation between the first real-time starting load and the standard starting load is calculated, and based on this deviation analysis, the first fluctuation coefficient is determined. Finally, the standard load is compensated according to the first fluctuation coefficient to obtain the first predicted load. Adding it to multiple predicted loads gives a more comprehensive load forecasting result. Combining real-time load data and historical load data can more accurately predict the load demand at each time point within the preset window.

[0024] Step Five: Use the FID device, the regional power supply network topology, and the regional load network topology to fuse and construct a distribution optimization model. In the distribution optimization model, with minimizing the power grid loss as the optimization index, search for an optimal distribution plan according to the multiple predicted powers and the multiple predicted loads, and output the optimal distribution plan.

[0025] Specifically, the FID device is a device used to monitor the frequency change of the power grid. By installing the FID device in the power grid, the frequency change of the power grid is monitored in real time, so as to understand the operating state and load situation of the power grid. The regional power supply network topology reflects the connection relationship between various power supply devices in the power grid, while the regional load network topology reflects the connection relationship between various load sub-regions in the power grid. These two topology diagrams are the basis for distribution optimization because they provide the structure and load distribution of the power grid. In the distribution optimization model, find one or more distribution plans that have the minimum power grid loss while meeting the predicted load. Search for an optimal distribution plan according to the multiple predicted powers and the multiple predicted loads, simulate and evaluate different distribution plans, and find the optimal plan. The optimal distribution plan is the plan with the minimum loss, which can improve the operating efficiency and stability of the power grid while meeting the predicted load. By finding a distribution plan that meets the predicted load and has the minimum power grid loss, the power grid resources can be allocated more efficiently and the operating efficiency of the power grid can be improved.

[0026] Step Six: Through the FID device, perform distribution optimization control within the predetermined window according to the optimal distribution plan.

[0027] Specifically, first, in the power distribution optimization model, by simulating and evaluating different power distribution schemes, the scheme with the minimum loss, the highest efficiency, and the best reliability is selected. For example, in the power distribution optimization model, different device operation modes, load distribution strategies, etc. can be considered to generate multiple initial power distribution schemes, and then the optimal scheme is selected through an optimization algorithm. Next, through the FID device, the frequency data of the power grid is obtained in real time, and the power grid is controlled according to the optimal power distribution scheme. For example, in an urban area, assuming that the optimal power distribution scheme is determined through the power distribution optimization model, the FID device installed in the power grid is used to monitor the frequency data of the power grid in real time. According to the optimal power distribution scheme, the operating state of the power grid is adjusted, such as adjusting the transformer settings, switchgear, etc., to achieve the power distribution optimization control within a predetermined window. By controlling according to the optimal power distribution scheme, the power grid resources can be allocated more efficiently, and the operating efficiency of the power grid can be improved.

[0028] Further, step one of the present application includes:

[0029] Collect the power supply characteristic data of the target area. Among them, the power supply characteristic data includes power supply equipment, equipment location coordinates, and power distribution line characteristics. The power distribution line characteristics include line resistivity, line cross-sectional area, and line length; based on graph calculation, with the power supply equipment as the graph nodes and the power distribution lines as the edges of the graph, the regional power supply network topology is generated according to the equipment location coordinates.

[0030] Specifically, collect information on all key equipment in the power supply network of the target area, including thermal power plants and multiple new energy power generation equipment (such as wind power generation, solar power generation, etc.), substations, distribution stations, etc. For each power supply equipment, record its location coordinates through geographic information system data. The power distribution line characteristics are another key part of the power supply characteristic data, including line resistivity, line cross-sectional area, and line length. Among them, the line resistivity is an important parameter, which is determined based on indicators such as the material of the line and the interfacial agent, and affects the loss of electric energy during transmission. Different materials (such as copper, aluminum) and cross-sectional area sizes will have an impact on the resistivity. The line cross-sectional area and line length determine the power transmission capacity of the line. A larger cross-sectional area and a shorter length help to reduce energy loss and improve power transmission efficiency. Collecting these data usually requires the use of advanced data collection technologies and equipment. For example, use sensors carried by unmanned aerial vehicles to collect equipment location coordinates and line characteristic data, or use smart meters and monitoring equipment to monitor the operating state of power supply equipment in real time.

[0031] After collecting power supply characteristic data, a graph calculation method is used to construct the regional power supply network topology. In this topology, power supply devices are regarded as graph nodes, and each power supply device has a unique identifier and occupies a position in the graph. The distribution lines are regarded as the edges of the graph, and each edge connects two power supply devices, indicating that electric energy flows from the source device to the target device. The device position coordinates determine the positions of the nodes in the graph, thus reflecting the structure of the actual power grid. Calculate the shortest paths, minimum spanning trees, etc. between various nodes in the graph to find the optimal path for power transmission. Through such a topological structure, various analyses and optimization operations of the power grid can be carried out. For example, the connectivity of the power grid can be analyzed to identify potential power supply bottlenecks and weak links. Power flow analysis can also be performed based on this topology to predict the operating state of the power grid under different load conditions, thereby providing a basis for the operation optimization of the power grid. Through accurate data collection and advanced graph calculation technology, the accuracy of power grid analysis is improved.

[0032] Further, step two of this application includes:

[0033] Collect the load characteristics of multiple load points in the target area to obtain multiple load characteristic sets. Among them, the load characteristics at least include load access points, load magnitudes, load types, load usage patterns, load volatility, and load persistence; based on the K-means algorithm, cluster the multiple load points according to the multiple load characteristic sets to generate multiple load sub-areas; based on graph calculation, use the load sub-areas as graph nodes and the power consumption lines as the edges of the graph to generate the regional load network topology according to the load position coordinates.

[0034] Specifically, through means such as smart meters, historical power consumption data, and user surveys, collect detailed information on each load point in the target area, including but not limited to load access points, load magnitudes, load types, load usage patterns, load volatility, and load persistence. The load access point refers to the specific location where the load is connected to the power grid; the load magnitude reflects the power demand of the load, which can be fixed or variable, depending on the load type and usage pattern; the load type refers to the type of load, such as residential load, commercial load, industrial load, etc. Different types of loads have different power consumption characteristics and requirements; the load usage pattern describes the usage rules of the load during a day or a week, such as peak hours and off-peak hours; the load volatility refers to the large fluctuations of the load in a short period of time, such as seasonal changes and sudden peaks. High-volatility loads may affect the stability of the power grid and need special attention; the load persistence describes the duration of the load. Some loads are short-term, such as household appliances, while some loads are long-term, such as lighting systems. Through these load characteristics, a detailed load characteristic set is constructed to reflect the power consumption behavior and requirements of the load.

[0035] Using the K-means algorithm, load points with similar features in multiple load feature sets are divided into a sub-region, generating multiple load sub-regions. The K-means algorithm is a commonly used clustering analysis algorithm, aiming to classify load points with similar features into one category, so as to better understand the load characteristics of the power grid. Through clustering, different load patterns can be identified. Multiple load points are clustered according to the load feature set. Each sub-region contains load points with similar features, and these load points have similar electricity consumption behaviors and demands. For example, data of each load point in a city is collected, including residential areas, shopping malls, factories, etc. The load in residential areas has peaks in the morning and evening, the load in commercial areas has peaks during the day and evening, and the load in industrial areas has peaks during the day. Each load sub-region is regarded as a graph node, and these nodes represent a set of load points with similar features. The power lines are used as the edges connecting these nodes, and the weights of the edges can represent characteristics such as the capacity and resistance of the lines. The positions of the nodes in the graph are determined according to the load location coordinates, reflecting the actual load structure of the power grid, and generating the regional load network topology. By collecting the load characteristics of multiple load points in the target region, performing clustering analysis based on the K-means algorithm to generate multiple load sub-regions, and finally constructing the regional load network topology, the load structure of the power grid can be clearly displayed, and the power demand and supply can be predicted more accurately.

[0036] Further, step three of this application includes:

[0037] Collect the attribute information of multiple power supply devices in the regional power supply network topology to obtain multiple device attribute information. Randomly select the first power supply device and obtain the first device attribute information of the first power supply device; using the first device attribute information as a constraint, perform homologous information retrieval based on power big data to obtain a sample data set, where the sample data set includes a sample power generation factor set and a sample power generation power; divide the sample data set into N equal parts and select N times with replacement to generate a first training set. Iteratively select N times to obtain N training sets; using the sample power generation factor set as the input and the sample power generation power as the supervision, use the N training sets to perform supervised training and cross-validation on the BP neural network to obtain N convergent power generation prediction units, and parallelly build a first power generation prediction model based on the N convergent power generation prediction units. The output of the first power generation prediction model is the mean of the output results of the N convergent power generation prediction units; establish a first mapping relationship between the first power supply device and the first power generation prediction model, and based on the classification decision principle, construct the power generation prediction channel according to the first mapping relationship.

[0038] Specifically, conduct a comprehensive investigation and record of all power supply devices in the regional power supply network topology, such as device type, device specifications, working principle, operating parameters, etc. For different power supply devices, such as new energy devices like photovoltaic power generation devices and wind power generation devices, these attribute information reflects their working characteristics, performance parameters and operating status. For example, the attribute information of photovoltaic power generation devices includes the type of solar panels, conversion efficiency, maximum power point tracking technology, etc.; the attribute information of wind power generation devices includes the type of wind turbines, blade length, wind speed requirements, etc. Use the method of random sampling to select a device as the first power supply device and obtain its specific information. Use the attribute information of the first power supply device as a constraint condition to retrieve homologous information in the power big data. The power big data contains a large amount of power system operation data, such as power generation, load demand, grid parameters, etc. By retrieving these data, information related to the attributes of specific devices can be obtained, thereby constructing a sample data set.

[0039] The sample data set includes a sample generation factor set and a sample generation power. The sample generation factor set is the factors related to power generation, and these factors affect the power generation performance of the device. For photovoltaic power generation devices, these factors include light intensity, light angle, temperature, humidity, etc., which have a direct impact on the power generation efficiency of the photovoltaic panels. For wind power generation devices, it includes wind speed, wind direction, air density, etc. The sample generation power refers to the amount of power generated by the device under specific conditions. Divide the sample data set into N equal parts, each part contains a subset of the sample data set and includes a certain number of samples to ensure that each training set can represent the characteristics and distribution of the entire data set. Conduct sampling with replacement. After each sampling, the sampled sample is put back into the original data set, so that it may be selected again in the next sampling, ensuring that each sampling is independent, thereby increasing the diversity of the training set. Repeat the above process of sampling with replacement, and each sampling will generate a new training set. Through multiple iterative samplings, multiple different training sets can be obtained, and these training sets may vary in data distribution and characteristics.

[0040] Use N training sets to conduct supervised training on the BP neural network. The input layer of the network is connected to the sample generation factor set, and the output layer is connected to the sample generation power. Through the backpropagation algorithm, the network will adjust the internal weights to minimize the difference between the actual output and the target output. The BP neural network is a feedforward artificial neural network that can make predictions by learning the relationship between the features in the sample data and the target variables. To ensure the generalization ability of the model and avoid overfitting problems, cross-validation is used for the model. The data set is divided into a training set and a validation set. The training set is used to train the model, and the validation set is used to evaluate the performance of the model. Through multiple trainings and cross-validations, N convergent power generation prediction units are obtained. Each prediction unit is trained on a different training set, so they may have different weights and prediction abilities.

[0041] Combine these N convergent power generation prediction units in parallel to construct a first power generation prediction model. The output of the model is the average of the output results of these prediction units, reducing the prediction error of a single model and improving the overall prediction accuracy. Match the first power supply device with the first power generation prediction model to establish a first mapping relationship, which can be determined by the identifier or location information of the device. For example, a specific photovoltaic power generation device can correspond to a specific power generation prediction model. Based on the classification decision principle, according to the established mapping relationship, input the real-time data of the power supply device into the corresponding power generation prediction model to obtain the predicted power generation. The classification decision principle is a commonly used machine learning method that assigns an observation value (or a set of observation values) to different pre-defined categories to achieve prediction and decision-making on data. For example, in a photovoltaic power generation system, first establish the mapping relationship between each photovoltaic power generation device and the corresponding power generation prediction model. Then, based on the classification decision principle, input the real-time data of each photovoltaic power generation device into the corresponding power generation prediction model to obtain the predicted power generation. Constructing a power generation prediction channel can achieve real-time prediction and optimization of the power grid operation, improving the operation efficiency and power supply quality of the power grid.

[0042] Furthermore, as Figure 2 shown, step four of this application includes:

[0043] Based on the regional load network topology, randomly select a first load sub-region; with the preset window and the first load sub-region as constraints, call the load database to retrieve and obtain a sample starting load set and a sample load set, where the sample starting load and the sample load correspond one by one and are marked with time series characteristics; configure a gradually decreasing weight based on the time series characteristics, where the weight is negatively correlated with the time series characteristics; perform weighted fusion on the sample starting load set and the sample load set respectively according to the gradually decreasing weight to obtain a standard starting load and a standard load; generate a first predicted load based on the standard starting load and the standard load and add it to the multiple predicted loads.

[0044] Specifically, the regional load network topology diagram reflects the connection relationships among various load sub-regions in the power grid. In this topology diagram, each load sub-region is regarded as a node, and the lines connecting these regions are regarded as edges. Randomly select one from all the load sub-regions as the research object to ensure that the selected load sub-region is representative, so as to better reflect the characteristics of different load sub-regions in the power grid. For example, in an urban area, through a comprehensive survey and record of all load sub-regions in the power grid, load characteristics such as load type, load size, and load volatility are obtained. Then, a load sub-region in a residential area is randomly selected, and its load characteristics are analyzed in depth. Through the analysis, it is found that the load in this region is mainly caused by household appliances and has a large load volatility.

[0045] Determine a preset window, which is used to determine the time range for prediction and analysis. The preset window refers to a specific time period in the future, such as the next few hours or days. Based on the first load sub-region, call the load database, which contains historical and real-time load data and reflects the power demand situation of the load sub-region at different time points. By calling the load database, load data related to the first load sub-region can be obtained. Retrieve and obtain the historical starting point load set and the historical load set, calculate the average value of the load data at each time point to determine the historical starting point average load and the historical average load. Compare the load data at each time point with the corresponding historical starting point average load or historical average load, and calculate the difference between them. According to a predetermined deviation threshold, eliminate the unreliable data greater than the threshold, and use the remaining data as sample data, that is, the sample starting point load set and the sample load set. The sample starting point load set refers to the starting load data at the same historical moment within the preset window, that is, the load data at the historical moment corresponding to the current prediction moment, while the sample load set includes the average load data at each time point within the historical preset window. These sample load data correspond one by one to the sample starting point load data and identify specific time points, that is, they have a time series feature. The time series feature refers to the time interval between the sample starting point load and the sample load. The shorter the interval, the closer the load situation at the current moment is to the historical data, so the higher the credibility and the greater the weight. On the contrary, the longer the time interval, the greater the difference between the load situation at the current moment and the historical data, so its impact on the prediction result is smaller and the weight is lower. For example, if the load at the current moment is very close to the load at the same time point in the past hour, then the weight of this sample starting point load will be very high.

[0046] Gradually decreasing weights are configured based on temporal characteristics. As time goes by, the contribution of sample load data to the load prediction at the current moment gradually decreases. For example, if the time interval between the current moment and the historical load data in the past hour is the shortest, then the weight of this sample load is the highest; if the time interval with the historical load data in the past six hours is the longest, then the weight of this sample load is the lowest. The sample starting load set and the sample load set are weighted and fused. Weighted fusion means that the historical load data is weighted and averaged according to the configured weights to obtain a more accurate predicted value. In this process, the load data at each time point is weighted according to its corresponding weight, so as to obtain the standard starting load and the standard load. The standard starting load reflects the load situation at the start time of the preset window, while the standard load is the average load at each time point within the preset window.

[0047] The first real-time starting load of the first load sub-region is obtained through the non-intrusive power load monitoring method, and the deviation from the standard starting load is calculated, that is, the first starting load deviation. The first fluctuation coefficient is analyzed and determined. If the first starting load deviation is large, it means that the load volatility is strong, and the first fluctuation coefficient is large. The standard load is compensated according to the first fluctuation coefficient to obtain the first predicted load. The first predicted load is positively correlated with the first fluctuation coefficient, that is, the larger the first fluctuation coefficient, the larger the first predicted load. The first predicted load is added to multiple predicted loads. Multiple predicted loads refer to the predicted load data at each time point within the preset window. By adding the first predicted load to these predicted loads, a more comprehensive load prediction result can be obtained. This result contains the information of the first predicted load and the information of other predicted loads that have been generated before. By configuring gradually decreasing weights and performing weighted fusion, the load demand at each time point within the preset window can be predicted more accurately, which helps to formulate more effective power supply plans and resource allocation strategies.

[0048] Furthermore, the present application further includes the following steps:

[0049] Retrieve and obtain the historical starting load set and the historical load set, calculate the mean values of the historical starting load set and the historical load set respectively to determine the historical starting mean load and the historical mean load; based on the historical starting mean load and the historical mean load, calculate the deviations of the historical starting load set and the historical load set respectively to obtain the historical starting load deviation set and the historical load deviation set; screen the historical starting load deviation set and the historical load deviation set according to a predetermined deviation threshold, and map to obtain the sample starting load set and the sample load set.

[0050] Specifically, the historical starting point load set refers to the starting load data at each time point within a past period, while the historical load set includes the load data at each time point within a past period. These data are usually stored in a load database and can be obtained by calling the load database. Mean calculation refers to calculating the average value of the load data at each time point to obtain the average load level over a period of time. The historical starting point mean load refers to the average value of the starting load data at each time point within a past period, while the historical mean load refers to the average value of the load data at each time point within a past period. Deviation calculation refers to comparing the load data at each time point with the corresponding historical starting point mean load or historical mean load and calculating the difference between them. For the historical starting point load deviation set, calculate the difference between the load data at each time point and the corresponding historical starting point mean load; for the historical load deviation set, calculate the difference between the load data at each time point and the corresponding historical mean load.

[0051] According to the actual situation and requirements, a deviation threshold is preset in advance to determine whether the load data deviation is within an acceptable range. According to the predetermined deviation threshold, the historical starting point load deviation set and the historical load deviation set are screened. Compare the load deviation at each time point with the predetermined deviation threshold to determine whether it is within the acceptable range. If the absolute value of the load deviation is greater than the predetermined deviation threshold, it is considered that the influence of this data point on the current load prediction is small and can be screened out; if the absolute value of the load deviation is less than or equal to the predetermined deviation threshold, it is considered that the influence of this data point on the current load prediction is large and should be retained. The screened data sets are called the sample starting point load set and the sample load set, which contain the load data with deviations within the predetermined threshold and can be used for subsequent load analysis and prediction. Based on the screened load data, power production and distribution can be planned more reasonably to cope with load fluctuations.

[0052] Furthermore, the present application further includes the following steps:

[0053] Obtain the first real-time starting point load of the first load sub-region through a non-intrusive power load monitoring method; calculate the first starting point load deviation between the first real-time starting point load and the standard starting point load, and determine the first fluctuation coefficient based on the analysis of the first starting point load deviation; compensate the standard load according to the first fluctuation coefficient to obtain the first predicted load, where the first predicted load is positively correlated with the first fluctuation coefficient.

[0054] Specifically, non-intrusive power load monitoring is a monitoring technology that does not require direct contact with the load or modification of the load, and can be used to monitor the operating status of the load in the power grid without affecting the load. This method is usually implemented using technologies such as smart meters, power line communication technology (PLC), and wireless sensor networks, which are directly installed in the power grid and do not require direct access to the load device. Randomly select a first load sub-region in the regional load network topology. The first real-time starting load refers to the total load of all load points in the first load sub-region at a specific moment. The standard starting load refers to the starting load data at the same historical moment at the start time of the preset window. Compare the first real-time starting load with the standard starting load and calculate the deviation between them, that is, the first starting load deviation. Determine the first fluctuation coefficient based on the analysis of the first starting load deviation. The first fluctuation coefficient is a measure of the amplitude and frequency of load fluctuations, reflecting the volatility and change trend of the load. By analyzing the first starting load deviation, the first fluctuation coefficient can be determined. For example, if the first starting load deviation is large, it indicates strong load volatility and a large first fluctuation coefficient; if the first starting load deviation is small, it indicates weak load volatility and a small first fluctuation coefficient. To more accurately determine the first fluctuation coefficient, a decision tree algorithm can be used. A decision tree is a commonly used machine learning algorithm that can classify and predict based on input data (the first starting load deviation). Specifically, a deviation-coefficient matching table can be constructed based on multiple sample data, which contains the first fluctuation coefficients corresponding to different deviation values. Through the decision tree algorithm, the first fluctuation coefficient can be obtained by matching the first starting load deviation.

[0055] Compensate the standard load according to the first fluctuation coefficient by increasing or decreasing the value of the standard load to compensate for the load volatility. Compensation means adjusting the standard load according to the first fluctuation coefficient to reflect the load volatility. For example, if the first fluctuation coefficient is large, indicating strong load volatility, the value of the standard load can be increased to compensate for this volatility; if the first fluctuation coefficient is small, indicating weak load volatility, the value of the standard load can be decreased to compensate for this volatility. Obtain the first predicted load based on the compensated standard load, taking into account the load volatility. The first predicted load is positively correlated with the first fluctuation coefficient, that is, the larger the first fluctuation coefficient, the larger the first predicted load; the smaller the first fluctuation coefficient, the smaller the first predicted load. By calculating the first starting load deviation through the first real-time starting load, analyzing and determining the first fluctuation coefficient, and finally compensating the standard load according to the first fluctuation coefficient to obtain the first predicted load, the volatility and change trend of the power grid load can be understood more deeply.

[0056] Further, step five of this application includes:

[0057] Within the power distribution optimization model, based on the regional power supply network topology and the regional load network topology, with the satisfaction of the multiple predicted loads as a constraint, power distribution scheme simulation is carried out according to the multiple predicted powers to generate multiple initial power distribution schemes; according to the power grid loss evaluation function, loss evaluations are respectively carried out on the multiple initial power distribution schemes, and the initial power distribution scheme corresponding to the minimum loss coefficient is set as the optimal power distribution scheme.

[0058] Specifically, an FID device is used to monitor the power grid frequency change to understand the operating state of the power grid in real time. Based on the regional power supply network topology and the regional load network topology, a structure and load distribution model of the power grid is established. With the satisfaction of multiple predicted load demands as a constraint, power distribution scheme simulation is carried out according to multiple predicted powers, and multiple power distribution schemes are randomly generated. The structure information of the power grid and the predicted load and power data are used to simulate different power distribution schemes. For example, in an urban area, different power supply schemes can be simulated according to the topology of the power supply network and the predicted load demands, including different power supply paths and equipment operating states. Different power supply equipment operating states and load distribution strategies are considered, thus generating multiple possible power distribution schemes. For example, different equipment operating modes, load distribution strategies, etc. can be considered to generate multiple initial power distribution schemes.

[0059] The constructed power grid loss evaluation function is used to evaluate the performance of each initial power distribution scheme. Each initial power distribution scheme may lead to different combinations of current distribution, resistivity, cross-sectional area, and length, thus affecting the total loss of the power grid. For each initial power distribution scheme, loss calculations are carried out for each load point, and the losses of all load points are summed up to calculate the loss coefficient of each initial power distribution scheme. The loss coefficients of all initial power distribution schemes are compared, and the power distribution scheme with the minimum loss coefficient is selected as the optimal scheme. This scheme will theoretically provide the lowest power grid loss, thus improving the operating efficiency and power supply quality of the power grid. In this way, one or more initial power distribution schemes can be found that have the lowest power grid loss while satisfying the predicted loads.

[0060] Furthermore, the present application further includes the following steps:

[0061] The expression of the power grid loss evaluation function is: where S is the loss coefficient, M is the number of predicted loads, I m is the load current of the m-th predicted load, P m is the resistivity of the power consumption line of the m-th predicted load, A m is the cross-sectional area of the power consumption line of the m-th predicted load, L m is the length of the power consumption line of the m-th predicted load.

[0062] Specifically, the expression of the power grid loss evaluation function is: Among them, S is the loss coefficient, representing the overall level of power grid losses; M is the predicted number of loads, indicating the total number of load points to be considered; I m is the load current of the m-th predicted load; P m is the resistivity of the power consumption line of the m-th predicted load, which determines the loss degree during the power transmission process; A m is the cross-sectional area of the power consumption line of the m-th predicted load, which affects the power transmission efficiency; L m is the length of the power consumption line of the m-th predicted load, which directly affects the losses during the power transmission process. By comprehensively considering the current, resistivity, cross-sectional area, and length of each load point, and through the method of product and summation, the total losses of the entire power grid are calculated. The loss coefficient S is an adjustment of the overall loss level to make it meet specific evaluation criteria or ranges.

[0063] In summary, a method for optimizing the operation of a distribution network based on FID provided by this application has the following technical effects:

[0064] By collecting the power supply characteristic data of the target area, a regional power supply network topology is built; based on the load characteristics, multiple load points in the target area are clustered to obtain multiple load sub-areas, and a regional load network topology is constructed according to the multiple load sub-areas; based on the regional power supply network topology, new energy power generation in a predetermined window is predicted through a power generation prediction channel to obtain multiple predicted powers; based on the regional load network topology, load prediction within a preset window is performed through load monitoring data and a load database to determine multiple predicted loads; a distribution optimization model is constructed by fusing an FID device, the regional power supply network topology, and the regional load network topology. In the distribution optimization model, with minimizing the power grid losses as the optimization index, a distribution plan is optimized according to the multiple predicted powers and the multiple predicted loads, and the optimal distribution plan is output; through the FID device, distribution optimization control within a predetermined window is performed according to the optimal distribution plan. That is to say, by accurately matching based on the regional power consumption load and the new energy power generation status, the scientificity and rationality of the regional distribution plan setting can be improved, thereby achieving the technical effects of reducing the regional power grid losses and improving the distribution operation efficiency.

[0065] The above description of the disclosed embodiments enables those skilled in the art to implement or use this application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0066] Obviously, those skilled in the art can make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of this application and its equivalent technologies, this application also intends to include these changes and modifications.

Claims

1. A distribution network operation optimization method based on FID, characterized in that: include: Collect power supply characteristic data of the target area and build the regional power supply network topology; Clustering multiple load points in the target area based on load characteristics to obtain multiple load sub-areas, and constructing a regional load network topology according to the multiple load sub-areas; Based on the regional power supply network topology, a new energy power generation forecast is performed within a predetermined window through a power generation forecast channel to obtain multiple predicted powers; Based on the regional load network topology, load forecasting within a predetermined window is performed through load monitoring data and a load database to determine a plurality of forecast loads; A power distribution optimization model is constructed by integrating the FID device, the regional power supply network topology and the regional load network topology. In the power distribution optimization model, minimizing power grid loss is used as an optimization index, and a power distribution plan is optimized according to the multiple predicted powers and the multiple predicted loads to output an optimal power distribution plan. The FID device is a device for monitoring power grid frequency changes. The FID device installed in the power grid monitors the power grid frequency changes in real time to understand the operating status and load conditions of the power grid. By means of the FID device, optimizing power distribution control within the predetermined window is performed according to the optimal power distribution plan; Among them, building a regional power supply network topology includes: Collecting power supply characteristic data of the target area, wherein the power supply characteristic data includes power supply equipment, equipment location coordinates and distribution line characteristics, and the distribution line characteristics include line resistivity, line cross-sectional area and line length; Based on graph computing, the power supply equipment is used as the graph node, the distribution line is used as the graph edge, and the regional power supply network topology is generated according to the equipment location coordinates; Among them, building a regional load network topology includes: Collecting load characteristics of multiple load points in the target area to obtain multiple load characteristic sets, wherein the load characteristics at least include load access points, load size, load type, load usage mode, load volatility and load continuity; Based on the K-means algorithm, clustering the multiple load points according to the multiple load feature sets to generate multiple load sub-areas; Based on graph calculation, the load sub-area is used as the graph node, the power lines are used as the graph edge, and the regional load network topology is generated according to the load position coordinates; Among them, the optimal power distribution solution is output, including: In the power distribution optimization model, based on the regional power supply network topology and the regional load network topology, with the plurality of predicted loads as a constraint, a power distribution scheme simulation is performed according to the plurality of predicted powers to generate a plurality of initial power distribution schemes; According to the power grid loss evaluation function, respectively perform loss evaluation on the multiple initial power distribution plans, and set the initial power distribution plan corresponding to the minimum loss coefficient as the optimal power distribution plan; Construct a power grid loss evaluation function, including: The expression of the power grid loss evaluation function is: ； Where S is the loss coefficient, To predict the load quantity, is the load current of the mth predicted load, is the power line resistivity of the mth predicted load, is the cross-sectional area of ​​the power line of the mth predicted load, is the length of the power line for the mth predicted load.

2. According to the FID-based distribution network operation optimization method of claim 1, it is characterized in that: Construct a power generation prediction channel, including: Collecting attribute information of multiple power supply devices in the regional power supply network topology to obtain multiple device attribute information, randomly selecting a first power supply device, and obtaining first device attribute information of the first power supply device; Taking the first device attribute information as a constraint, performing homologous information retrieval based on power big data to obtain a sample data set, wherein the sample data set includes a sample power generation factor set and a sample power generation power; Divide the sample data set into N equal parts, select N parts with replacement to generate a first training set, and iterate and select N parts N times to obtain N training sets; Taking the sample power generation factor set as input and the sample power generation as supervision, the BP neural network is supervised and cross-validated using the N training sets to obtain N convergent power generation prediction units, and a first power generation prediction model is built in parallel based on the N convergent power generation prediction units, wherein the output of the first power generation prediction model is the average of the output results of the N convergent power generation prediction units; A first mapping relationship between the first power supply device and the first power generation prediction model is established, and based on a classification decision principle, the power generation prediction channel is constructed according to the first mapping relationship.

3. According to the FID-based distribution network operation optimization method of claim 1, it is characterized in that: Determine multiple forecast loads, including: Based on the regional load network topology, randomly selecting a first load sub-region; Taking the predetermined window and the first load sub-area as constraints, calling the load database, retrieving and acquiring a sample starting load set and a sample load set, wherein the sample starting load and the sample load correspond one to one and are marked with a time series feature; Configuring a gradually decreasing weight based on the time series feature, wherein the weight is negatively correlated with the time series feature; According to the step-by-step decreasing weights, weighted fusion is performed on the sample starting point load set and the sample load set to obtain a standard starting point load and a standard load; A first predicted load is generated based on the standard starting load and the standard load, and is added to the plurality of predicted loads.

4. The method for optimizing the operation of a distribution network based on FID according to claim 3, characterized in that: Get a sample starting load set and a sample load set, including: Retrieve and obtain a historical starting point load set and a historical load set, perform mean calculations on the historical starting point load set and the historical load set respectively, and determine a historical starting point mean load and a historical mean load; Taking the historical starting point mean load and the historical mean load as references, respectively performing deviation calculations on the historical starting point load set and the historical load set to obtain a historical starting point load deviation set and a historical load deviation set; The historical starting point load deviation set and the historical load deviation set are screened according to a predetermined deviation threshold, and a sample starting point load set and a sample load set are obtained by mapping.

5. The method for optimizing distribution network operation based on FID according to claim 3, characterized in that: Generating a first predicted load based on the standard starting load and the standard load includes: Acquiring a first real-time starting point load of the first load sub-area by a non-intrusive power load monitoring method; Calculating a first starting point load deviation between the first real-time starting point load and the standard starting point load, and determining a first fluctuation coefficient based on the first starting point load deviation analysis; The standard load is compensated according to the first fluctuation coefficient to obtain the first predicted load, wherein the first predicted load is positively correlated with the first fluctuation coefficient.

Images