Risk monitoring and early warning platform for power grid engineering

Abstract

The invention discloses a risk monitoring and early warning platform for power grid engineering, and relates to the technical field of power grids. Building a graph neural network, traversing the power quality indexes to configure the graph neural network, and generating a power quality prediction channel; monitoring a power quality index initial characteristic value and a distributed power supply prediction generating capacity array; analyzing through a power quality prediction channel to obtain a power quality index prediction characteristic value; and when the power quality index threshold interval is not satisfied, generating a distributed power supply grid-connected early warning signal and sending the signal to a user side. The technical problem that the existing power grid risk monitoring and early warning cannot analyze the specific state of each parameter of the power quality by the distributed power supply grid connection, so that the influence of the distributed power supply grid connection on the power quality is difficult to accurately predict in real time, and the power grid risk is uncontrollable is solved. The technical effect of improving the accuracy of power quality monitoring and early warning and the stability of a power grid is achieved.

Description

Technical Field

[0001] The present application relates to the technical field related to power grids, and in particular to a risk monitoring and early warning platform for power grid projects. Background Art

[0002] In today's rapidly developing energy field, power grid engineering is a key link connecting energy production and consumption. Its stability, safety and efficiency are directly related to the normal operation and sustainable development of the social economy. With the widespread application and popularization of distributed power sources (such as solar photovoltaic, wind power generation, etc.), the power grid structure is becoming increasingly complex. Traditional monitoring and early warning methods can no longer meet the needs of modern power grids for efficient and accurate risk management. With the rapid development of the concept of smart grids, power grid systems are gradually developing in the direction of high informatization, automation and interactivity. Distributed power sources, as an important part of smart grids, have significant advantages in flexibility and environmental protection, but at the same time, they also bring many challenges to the power grid, such as power quality problems such as voltage fluctuations, harmonic pollution, and islanding effects, which not only affect the stable operation of the power grid, but may also cause damage to electrical equipment and even cause large-scale power outages. How to effectively monitor the changes in power quality after distributed power sources are connected to the grid and timely warn of potential risks has become a key issue that needs to be urgently solved in the current power grid engineering field.

[0003] Therefore, in the current technologies related to grid risk monitoring and early warning, there is a technical problem that it is impossible to analyze the specific status of various parameters of power quality caused by the grid connection of distributed power sources, which makes it difficult to predict the impact of the grid connection of distributed power sources on power quality in real time and accurately, making the grid risk uncontrollable. Summary of the invention

[0004] This application solves the technical problem that the existing power grid risk monitoring and early warning system is unable to analyze the specific status of various parameters of power quality when distributed power sources are connected to the grid, which makes it difficult to predict the impact of distributed power sources on power quality in real time and accurately, making power grid risks uncontrollable, by providing a risk monitoring and early warning platform for power grid projects. The application achieves the technical effect of improving the accuracy of power quality monitoring and early warning and the stability of the power grid.

[0005] The present application provides a risk monitoring and early warning platform for power grid projects, the platform comprising: a substation grid equivalent topology structure construction module, used to construct a substation grid equivalent topology structure, wherein the substation grid equivalent topology structure includes distributed power source deployment location label information and distributed power source type label information; a power quality prediction channel generation module, used to build a graph neural network based on the substation grid equivalent topology structure, and traverse the power quality index to configure the graph neural network based on the distributed power source deployment location label information and the distributed power source type label information, and generate a power quality prediction channel; a power quality index and power generation array acquisition module, used to monitor the initial characteristic value of the power quality index and the distributed power source predicted power generation array when receiving a distributed power source grid connection signal; a power quality index predicted characteristic value acquisition module, used to parse the initial characteristic value of the power quality index and the distributed power source predicted power generation array through the power quality prediction channel to obtain the power quality index predicted characteristic value; a distributed power source grid connection warning signal generation module, used to generate a distributed power source grid connection warning signal and send it to the user end when the power quality index predicted characteristic value does not meet the power quality index threshold interval.

[0006] In a possible implementation, the power quality prediction channel generation module also performs the following processing: extracting a first power quality indicator based on the power quality indicator; collecting the initial record value of the first power quality indicator, the distributed power generation array record value and the first power quality indicator grid-connected record value based on the equivalent topology structure of the substation power grid, the distributed power source deployment location label information and the distributed power source type label information, and obtaining a first power quality indicator prediction sub-channel construction data set; calling the power quality prediction channel construction data set, configuring the graph neural network, and generating a first power quality indicator prediction sub-channel; adding the first power quality indicator prediction sub-channel to the power quality prediction channel.

[0007] In a possible implementation, the power quality prediction channel generation module also performs the following processing: constructing a first loss function, wherein the first loss function is used to count the proportion of output deviations greater than or equal to an output deviation threshold for a preset number of consecutive trainings; constructing a second loss function, wherein the second loss function is used to count the output deviation variance for a preset number of consecutive trainings; calling the power quality prediction channel construction data set, configuring the graph neural network, and generating the first power quality indicator prediction sub-channel when the first loss function is less than or equal to the first loss threshold, and the second loss function is less than or equal to the second loss threshold.

[0008] In a possible implementation, the power quality prediction channel generation module also performs the following processing: divide the power quality prediction channel construction data set into Q groups of power quality prediction channel construction data; perform Q times of sampling with replacement on the Q groups of power quality prediction channel construction data to construct a first power quality prediction channel construction data set; loop execution Z times to obtain a second power quality prediction channel construction data set until the Zth power quality prediction channel construction data set; traverse the first power quality prediction channel construction data set, the second power quality prediction channel construction data set until the Zth power quality prediction channel construction data set, configure the graph neural network, and when the first loss function is less than or equal to the first loss threshold, and the second loss function is less than or equal to the second loss threshold, construct Z basis channels; construct a fully connected layer, use the accuracy of the Z basis channels as weights, fuse the Z basis channels, and generate the first power quality indicator prediction sub-channel.

[0009] In a possible implementation, the power quality prediction channel generation module further performs the following processing: the power quality index includes at least a voltage fluctuation interval index, a current fluctuation interval index and a harmonic distortion index.

[0010] In a possible implementation, the power quality index prediction characteristic value acquisition module also performs the following processing: obtaining the first distributed power supply basic information, wherein the first distributed power supply basic information includes distributed power supply type information, distributed power supply scale information and array position identification; matching the environmental factor factors based on the distributed power supply type information through the environmental factor factor calibration table; performing multi-level sorting and identification of the power generation of the distributed power supply type information according to the environmental factor factors, and generating the first distributed power supply predicted power generation; based on the array position identification, adding the first distributed power supply predicted power generation to the distributed power supply predicted power generation array.

[0011] In a possible implementation, the power quality index prediction characteristic value acquisition module also performs the following processing: configure the retrieval factor item k value, wherein the total number of environmental factor factors ≥ the retrieval factor item k value ≥ 1, and the initial value of the retrieval factor item k value = 1; traverse the environmental factor factors and configure the deviation tolerance interval; extract k items of sorting factor factors from the environmental factor factors according to the retrieval factor item k value; extract k items of deviation tolerance intervals from the deviation tolerance interval based on the k items of sorting factor factors; based on the k items of sorting factor factors, collect power generation data for the distributed power type information, and when the k items of sorting factor factors of the retrieved data all meet the k items of deviation tolerance interval, store the retrieved data, otherwise, delete the retrieved data; determine whether the k value is equal to the total number of environmental factor factors; if equal, perform a centralized trend analysis on the power generation of the stored retrieval data to obtain the predicted power generation of the first distributed power source; if not equal, add one to k.

[0012] In a possible implementation, the power quality index prediction characteristic value acquisition module also performs the following processing: performing grey correlation analysis on the environmental factor based on the distributed power source type information to obtain a grey correlation analysis result; sorting the environmental factor from large to small according to the grey correlation analysis result to obtain an environmental factor sorting result; and sorting k items of factor factors according to the environmental factor sorting result based on the retrieval factor item k value.

[0013] It is proposed to construct an equivalent topology structure of the power grid in the substation area through the risk monitoring and early warning platform for power grid engineering proposed in this application, wherein the equivalent topology structure of the power grid in the substation area includes the label information of the deployment location of the distributed power source and the label information of the type of the distributed power source; a graph neural network is constructed based on the equivalent topology structure of the power grid in the substation area, and the graph neural network is configured by traversing the power quality index based on the label information of the deployment location of the distributed power source and the label information of the type of the distributed power source, so as to generate a power quality prediction channel; when receiving the distributed power grid connection signal, the initial characteristic value of the power quality index and the array of the predicted power generation of the distributed power source are monitored; the predicted characteristic value of the power quality index is obtained through the initial characteristic value of the power quality index and the array of the predicted power generation of the distributed power source through the power quality prediction channel; when the predicted characteristic value of the power quality index does not meet the threshold interval of the power quality index, a distributed power grid connection early warning signal is generated and sent to the user end. The technical problem that the existing power grid risk monitoring and early warning cannot analyze the specific state of the various parameters of the power quality of the distributed power grid connection, which leads to the difficulty of real-time and accurate prediction of the impact of the distributed power grid connection on the power quality, making the power grid risk uncontrollable is solved, and the accuracy of power quality monitoring and early warning and the technical effect of improving the stability of the power grid are achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solution of the embodiment of the present disclosure, the accompanying drawings of the embodiment of the present disclosure are briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the platform according to the embodiment of the present application. It should be understood that the previous or following operations are not necessarily performed precisely in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0015] Figure 1 A schematic diagram of the structure of a risk monitoring and early warning platform for power grid projects provided in an embodiment of the present application; Figure 2 A schematic diagram of the execution flow of a power quality prediction channel generation module in a risk monitoring and early warning platform for power grid projects provided in an embodiment of the present application.

[0016] Explanation of the reference numerals: a module for constructing an equivalent topology structure of an electric power grid in an area 10 , a module for generating a power quality prediction channel 20 , a module for acquiring a power quality index and a power generation array 30 , a module for acquiring a power quality index prediction characteristic value 40 , and a module for generating a distributed power source grid-connected warning signal 50 . DETAILED DESCRIPTION

[0017] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0018] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0019] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, platform, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or components that are not clearly listed or inherent to these processes, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0020] The present application embodiment provides a risk monitoring and early warning platform for power grid projects, such as Figure 1 As shown, the platform includes: The substation power grid equivalent topology structure construction module 10 is used to construct the substation power grid equivalent topology structure, wherein the substation power grid equivalent topology structure includes distributed power source deployment location label information and distributed power source type label information. Constructing an equivalent topological structure of a power grid in a substation refers to establishing a topological structure model that can reflect the connection relationship and operating characteristics between various electrical equipment and components in the substation power grid (i.e., a collection area of ​​a part of power users powered by one or more transformers) through abstraction and simplification. This topological structure model is an equivalent representation of the substation power grid, which is used to analyze and study the operating status, failure mode, optimization scheme, etc. of the power grid. Specifically, firstly, all electrical equipment and components in the substation power grid, such as transformers, switches, circuit breakers, lines, distributed power sources (such as solar photovoltaic, wind power generation, etc.), are identified, and classified according to their functions, characteristics and roles in the power grid. The connection relationship between these devices and components is analyzed, including the physical connection (such as wire connection) and logical connection (such as control signal transmission) between them, which constitutes the basic skeleton of the substation power grid. Then, the substation power grid is abstracted and simplified, usually including abstracting complex electrical equipment and components into simple nodes and lines, ignoring some detailed factors that have little effect on the analysis results (such as the specific direction of the line, the specific model of the equipment, etc.), and only retaining those that have an important impact on the operating status and performance of the power grid. Key information, and finally, based on this information, an equivalent topological structure of the substation power grid is established to display each node in the power grid (representing electrical equipment and components) and the connection relationship between them (representing power lines and control signal paths). Among them, the equivalent topological structure of the substation power grid includes distributed power deployment location label information and distributed power type label information. The distributed power deployment location label information is used to identify the specific physical location or relative location of the distributed power source in the substation power grid, which can quickly locate the location of the distributed power source and understand its layout in the power grid. It usually includes the substation, distribution line, tower number or specific geographical location coordinates (such as longitude and latitude) to which the distributed power source is connected; the distributed power type label information is used to distinguish and identify different types of distributed power sources. Different types of distributed power sources differ in power generation principles, operating characteristics, and grid-connected requirements. They usually include the type of distributed power source (such as solar photovoltaic, wind power generation, biomass power generation, etc.), capacity, rated voltage, and grid-connected mode (such as grid-connected, off-grid or hybrid), etc., which helps power grid managers understand the basic properties and operating characteristics of distributed power sources, so as to formulate corresponding grid-connected strategies and operation and maintenance plans. Through the equivalent structure of the substation power grid topology, the grid structure can be optimized, the refined management of distributed power sources can be achieved, the stability and reliability of the grid can be improved, and the operation efficiency of the grid can be improved.

[0021] The power quality prediction channel generation module 20 is used to build a graph neural network based on the equivalent topological structure of the substation power grid, and traverse the power quality indicators to configure the graph neural network based on the distributed power source deployment location label information and the distributed power source type label information to generate a power quality prediction channel. The nodes and edges in the equivalent topological structure model of the substation power grid are represented as nodes and edges in a graph data structure. The nodes can represent electrical equipment or components, and the edges can represent the connection relationship or power lines between them. Then, feature information is extracted for each node and edge. The feature information may include the type, capacity, operating status, etc. of the electrical equipment, as well as the length, impedance, voltage level, etc. of the power line. The distributed power source deployment location label information and the distributed power source type label information may also be used as additional feature inputs. A suitable graph neural network model is selected, such as a graph convolutional network (GCN) and a graph attention network (GAT), which can capture complex relationships in graph structure data and use these relationships for reasoning and prediction. The graph neural network model is trained and optimized using historical data or simulation data. By continuously adjusting the parameters and structure of the model, it can accurately predict the operation of the power grid. Status and power quality indicators. By analyzing the abnormal data in the power grid and the connection relationship between nodes, the model can quickly identify the fault point; using the distributed power deployment location label information and distributed power type label information, traverse the power quality indicators, that is, each power quality indicator is used as an input feature or dimension of the graph neural network, and then according to the location and type information of the distributed power supply, as well as the historical data of the power quality indicators, the graph neural network is configured and optimized so that it can learn the relationship between distributed power supply and power quality. After the graph neural network is successfully configured and trained, it is used as a prediction model. By inputting new distributed power supply information (such as the location and type of newly added distributed power supply) and the current power grid operation status (such as load conditions, weather conditions, etc.), the graph neural network can predict future power quality indicators and finally generate a power quality prediction channel.

[0022] The power quality index and power generation array acquisition module 30 is used to monitor the initial characteristic value of the power quality index and the distributed power generation predicted power generation array when receiving the distributed power grid connection signal. When receiving the distributed power grid connection signal, the initial characteristic value of the power quality index and the distributed power generation predicted power generation array are monitored. Specifically, the initial characteristic value of the power quality index refers to the basic data for monitoring and recording various power quality indicators in the power grid before the distributed power is connected to the grid, reflecting the operating status and performance of the power grid before the distributed power is connected to the grid, and is an important basis for evaluating the impact of the distributed power grid connection on the power grid, including voltage quality (such as voltage deviation, voltage fluctuation and flicker, voltage imbalance, etc., reflecting the stability and reliability of the grid voltage), current quality (such as harmonic current content, current waveform distortion rate, etc., reflecting the quality of the current in the grid, the increase of harmonic current will have an adverse effect on the grid equipment and user equipment), frequency quality (such as frequency deviation, frequency stability, etc.) and power supply reliability (including indicators such as power outage time and number of power outages, reflecting the power supply capacity and reliability level of the power grid); the predicted power generation of the distributed power source refers to the historical power generation data of the distributed power source, The predicted power generation value of distributed power sources in the future is calculated by the power generation prediction model based on factors such as weather forecast and equipment status. It is usually presented in the form of a time series, including the predicted power generation at each time point. The predicted power generation array of distributed power sources includes the predicted power generation of multiple distributed power sources. Through the predicted power generation array, the power system can understand the power generation capacity of distributed power sources in advance, reasonably arrange the dispatch plan of the power grid, and ensure the supply and demand balance and stable operation of the power grid. Specifically, the predicted power generation array of distributed power sources can be obtained by collecting historical power generation data, weather forecast data, equipment status data, etc. of distributed power sources, selecting appropriate prediction models (such as time series analysis, machine learning models, etc.) according to the characteristics of the data and prediction needs, using historical data to train and optimize the prediction model to improve the prediction accuracy of the model, and inputting the latest data collected into the trained prediction model to calculate the predicted power generation value of distributed power sources in the future.

[0023] The power quality index prediction characteristic value acquisition module 40 is used to obtain the power quality index prediction characteristic value through the power quality index initial characteristic value and the distributed power source predicted power generation array through the power quality prediction channel for analysis. The power quality index initial characteristic value and the distributed power source predicted power generation array are used as input data and input into the power quality prediction channel for analysis. The graph neural network model in the power quality prediction channel will perform complex calculations and analyses based on the input data, combined with the topological structure of the power grid, the characteristics of the distributed power source and the operation law of the power grid, and at the same time consider the influence of the distributed power source on the parameters such as the voltage, current and frequency of the power grid after the grid connection, as well as the regulation ability and stability of the power grid itself. After the calculation and analysis of the model, the power quality prediction channel will output the power quality index prediction characteristic value, such as voltage fluctuation, frequency deviation, harmonic content, etc., reflecting the power quality problems that the power grid may face in the future after the distributed power source is connected to the grid. The power system can take corresponding measures in advance based on these power quality index prediction characteristic values, such as adjusting the grid connection strategy of the distributed power source, optimizing the operation mode of the power grid, strengthening the monitoring and maintenance of the power grid, etc., to ensure the safe, stable and efficient operation of the power grid.

[0024] The distributed power grid connection warning signal generation module 50 is used to generate a distributed power grid connection warning signal and send it to the user end when the power quality index prediction characteristic value does not meet the power quality index threshold interval. The obtained predicted characteristic value is compared with the pre-set power quality index threshold interval, wherein the power quality index threshold interval refers to the critical value range of the power quality index set according to the historical data of the power grid in the substation area. If the power quality index prediction characteristic value does not meet the power quality index threshold interval, that is, it exceeds the allowed range, it will be automatically judged that there may be a problem with the power quality, and then a distributed power grid connection warning signal is generated, which usually includes key information such as warning level, warning content, and warning time, so as to take corresponding countermeasures in time, and send these warning signals to the user end, such as power grid dispatchers, distributed power operators, etc., through SMS, email, etc. By monitoring the early warning process, power quality problems can be discovered and responded to in a timely manner to ensure the safe and stable operation of the power grid and the normal power demand of users.

[0025] The risk monitoring and early warning platform for power grid projects according to the embodiment of the present invention is used to solve the technical problem that the existing power grid risk monitoring and early warning cannot analyze the specific status of various parameters of power quality caused by the grid connection of distributed power sources, which leads to the difficulty in real-time and accurate prediction of the impact of the grid connection of distributed power sources on power quality, making the grid risk uncontrollable, and achieves the technical effect of improving the accuracy of power quality monitoring and early warning and the stability of the power grid. The risk monitoring and early warning platform for power grid projects includes: a grid equivalent topology structure construction module 10, a power quality prediction channel generation module 20, a power quality index and power generation array acquisition module 30, a power quality index prediction characteristic value acquisition module 40, and a distributed power grid connection early warning signal generation module 50.

[0026] The specific configuration of the power quality prediction channel generation module 20 will be described in detail below. Figure 2 As shown, the power quality prediction channel generation module 20 may further include: extracting a first power quality indicator according to the power quality indicator. Randomly obtaining a power quality indicator from the power quality indicators as the first power quality indicator.

[0027] The power quality prediction channel generation module 20 also includes, based on the equivalent topological structure of the substation power grid, the distributed power source deployment location label information and the distributed power source type label information, collecting the initial record value of the first power quality indicator, the distributed power source generation array record value and the first power quality indicator grid connection record value, and obtaining the first power quality indicator prediction sub-channel construction data set. According to the equivalent topological structure of the substation power grid, the distributed power source deployment location label information and the distributed power source type label information, the initial record value of the first power quality indicator, the distributed power source generation array record value and the first power quality indicator grid connection record value are collected. Specifically, the initial record value of the first power quality indicator is the power quality indicator value when there is no distributed power source connected or the power grid is in a certain reference state; the distributed power source generation array record value records the power generation data of the distributed power source in the past period of time, reflecting the impact of the distributed power source on the power grid; the first power quality indicator grid connection record value refers to the actual value of the first power quality indicator recorded after the distributed power source is connected to the power grid, which is used to evaluate the impact of the distributed power source grid connection on the power quality; the above-mentioned collected data are integrated into a data set, that is, the first power quality indicator prediction sub-channel construction data set.

[0028] The power quality prediction channel generation module 20 also includes calling the power quality prediction channel construction data set, configuring the graph neural network, and generating a first power quality indicator prediction sub-channel. Retrieving or loading the existing power quality prediction channel construction data set from the storage system usually contains rich historical data, including but not limited to the operating status of the power grid, the measured values ​​of various power quality indicators, the operating data of distributed power sources (such as power generation, grid connection status, etc.), and external factors that may affect power quality (such as weather conditions, load changes, etc.), configuring the graph neural network according to the first power quality indicator prediction sub-channel construction data set, defining the learning goal of the network, that is, letting the network learn to predict future power quality indicator values ​​based on the input graph structure and related features, and then using the power quality prediction channel construction data set to train the graph neural network to generate a first power quality indicator prediction sub-channel, which is used to predict the first power quality indicator based on the input initial power data.

[0029] The power quality prediction channel generation module 20 also includes adding the first power quality indicator prediction subchannel into the power quality prediction channel. The prediction subchannel for predicting the first power quality indicator is integrated into a comprehensive power quality prediction channel, which may include multiple different prediction subchannels, each of which focuses on predicting different types of power quality indicators, and can more comprehensively cover different types of power quality indicators, improve the accuracy and reliability of prediction, and thus ensure stable operation of the power grid.

[0030] The specific configuration of the power quality prediction channel generation module 20 will be described in detail below. The power quality prediction channel generation module 20 may further include: constructing a first loss function, wherein the first loss function is used to count the proportion of output deviations greater than or equal to the output deviation threshold for a preset number of consecutive trainings. The first loss function is used to count the proportion of output deviations greater than or equal to the output deviation threshold for a preset number of consecutive trainings, where the output deviation refers to the difference between the model prediction value and the true value; the output deviation threshold refers to a set standard value. When the prediction error exceeds this threshold, the output deviation is considered to be large. The first loss function is as follows: ; in, The preset number of consecutive trainings; For the The actual output of the training; For the The model prediction output of the training; is the output deviation threshold; is an indicator function, which takes the value 1 if the condition is true, otherwise it is 0.

[0031] The power quality prediction channel generation module 20 also includes constructing a second loss function, wherein the second loss function is used to count the output deviation variance of a preset number of consecutive trainings. The second loss function is used to count the output deviation variance of a preset number of consecutive trainings, and the output deviation variance is used to measure the volatility of the model prediction error. The smaller the variance, the better the model prediction stability. The second loss function is as follows: ; in, is the mean of the output deviations, .

[0032] The power quality prediction channel generation module 20 also includes: calling the power quality prediction channel construction data set, configuring the graph neural network, and generating the first power quality indicator prediction sub-channel when the first loss function is less than or equal to the first loss threshold, and the second loss function is less than or equal to the second loss threshold. Retrieve the power quality prediction channel construction data set for training the graph neural network from the storage system, such as the operating status of the power grid, the measured value of the power quality indicator, the operating data of the distributed power source, etc., and then configure the graph neural network, define the network structure (such as the number of layers, the node representation dimension, the activation function, etc.), and set the necessary hyperparameters (such as the learning rate, the batch size, the optimization algorithm, etc.), and use the power quality prediction channel construction data set to train the graph neural network. If at a certain training point, the values ​​of the two loss functions meet the threshold conditions (that is, the first loss function is less than or equal to the first loss threshold, and the second loss function is less than or equal to the second loss threshold), it means that the graph neural network has reached a sufficient performance level, and its current state (including network weights, biases and other parameters) is saved to generate the first power quality indicator prediction sub-channel.

[0033] The specific configuration of the power quality prediction channel generation module 20 will be described in detail below. The power quality prediction channel generation module 20 may further include: dividing the power quality prediction channel construction data set into Q groups of power quality prediction channel construction data. Dividing the original power quality prediction channel construction data set into Q groups of power quality prediction channel construction data facilitates multiple independent and random resampling in the future.

[0034] The power quality prediction channel generation module 20 also includes performing Q times of sampling with replacement on the Q groups of power quality prediction channel construction data to construct a first power quality prediction channel construction data set. Performing Q times of sampling with replacement (or Bootstrap sampling) on ​​the Q groups of data means that in each sampling, each sample has the same probability of being selected, and the selected sample can still be selected in the next sampling, and each sampling will obtain a new data set that may contain repeated samples, that is, the first power quality prediction channel construction data set.

[0035] The power quality prediction channel generation module 20 also includes looping Z times to obtain a second power quality prediction channel construction data set until the Zth power quality prediction channel construction data set. The resampling and training process is looped Z times, and each time a graph neural network model trained based on different sample sets is obtained. During the training process, the first loss function and the second loss function are calculated respectively, and whether the performance requirements are met is determined based on whether the threshold conditions are met. In each loop, a new power quality prediction channel construction data set (from the second to the Zth power quality prediction channel construction data set) is obtained.

[0036] The power quality prediction channel generation module 20 also includes traversing the first power quality prediction channel construction data set, the second power quality prediction channel construction data set until the Zth power quality prediction channel construction data set, configuring the graph neural network, and constructing Z basis channels when the first loss function is less than or equal to the first loss threshold, and the second loss function is less than or equal to the second loss threshold. Traverse all data sets obtained by resampling and training (i.e., the first to Zth power quality prediction channel construction data sets), and configure and train the graph neural network. When a model meets the conditions during the training process (i.e., the first loss function is less than or equal to the first loss threshold, and the second loss function is less than or equal to the second loss threshold), it is regarded as a valid basis channel, and Z basis channels will eventually be obtained (although the actual number may be less than Z, because not every training can meet the conditions).

[0037] The power quality prediction channel generation module 20 also includes constructing a fully connected layer, taking the accuracy of the Z base channels as weights, fusing the Z base channels, and generating the first power quality index prediction sub-channel. It is necessary to evaluate the accuracy of each base channel on the validation set, which represents the proportion of correct predictions of the model. According to the accuracy of each base channel, their respective weights are calculated. The base channels with higher accuracy should have greater weights in the fusion process, and then a fully connected layer is constructed as a fusion layer. The fully connected layer is a neural network layer in which each neuron is connected to all neurons in the previous layer. Specifically, the input of the fully connected layer is the output of the Z base channels, and the output is the weighted sum of these outputs. The calculated weights are then used to perform weighted fusion on the outputs of the Z base channels. The fusion result is the first power quality index prediction sub-channel, thereby providing a more accurate and stable power quality prediction.

[0038] Below, the specific configuration of the power quality prediction channel generation module 20 will continue to be described in detail. The power quality prediction channel generation module 20 may further include: the power quality index includes at least a voltage fluctuation range index, a current fluctuation range index and a harmonic distortion index. The voltage fluctuation range index mainly measures the range of change of the voltage value in the power grid within a certain period of time. The stability of the voltage directly affects the normal operation and life of the electrical equipment, including the voltage fluctuation range and voltage sag, swell and flicker. The voltage fluctuation range usually refers to the amplitude of the voltage fluctuation between the maximum and minimum values; voltage sag, swell and flicker reflect the instantaneous mutation of the voltage, such as short-term sag, swell and flicker; the current fluctuation range index measures the change of the current value in the power grid within a certain period of time, including the difference between the maximum current and the minimum current, the rate or frequency of current change, etc., wherein the current deviation refers to the difference between the actual current value and the rated current value. The current fluctuation rate refers to the degree of fluctuation of the current value within a certain period of time, which can be measured by standard deviation, variance or the difference between the maximum and minimum values; the harmonic distortion index is used to measure the degree of distortion of the voltage and current waveforms in the power system. Harmonics are generated by nonlinear loads (such as power electronic equipment). They will cause the voltage and current waveforms in the power grid to deviate from the ideal sinusoidal waveform, including total harmonic distortion rate and harmonic content rate. Among them, the total harmonic distortion rate represents the ratio of the effective value of the harmonic component to the effective value of the fundamental component. The larger the value, the more serious the waveform distortion; the harmonic content rate refers to the percentage of the root mean square value (effective value) of the nth harmonic component in the periodic alternating current to the root mean square value of the fundamental component.

[0039] The specific configuration of the power quality index prediction characteristic value acquisition module 40 will be described in detail below. The power quality index prediction characteristic value acquisition module 40 may further include: obtaining the first distributed power basic information, wherein the first distributed power basic information includes distributed power type information, distributed power scale information and array position identification. Obtaining the basic information of a distributed power source includes distributed power type information, distributed power scale information and array position identification. Specifically, the distributed power type information refers to the type of distributed power source, such as solar photovoltaic panels, wind turbines, small hydroelectric generators, biomass generators, etc.; the distributed power scale information describes the scale or capacity of the distributed power source. For example, for solar photovoltaic panels, the scale may refer to its total installed capacity (i.e., the total power generation capacity of all photovoltaic panels), and for wind turbines, it may refer to its rated power or the area swept by the blades; the array position identification refers to a system containing multiple distributed power sources, in which a unique position identification is assigned to each distributed power source, which is used to quickly correspond the basic information, power generation forecast, actual power generation and other data of a distributed power source to it.

[0040] The power quality index prediction characteristic value acquisition module 40 also includes, based on the distributed power type information, matching environmental factor factors through the environmental factor calibration table. The environmental factor calibration table is a table containing the relationship between different distributed power types and corresponding environmental factor factors. According to the type of distributed power, the environmental factor factors are matched in the environmental factor calibration table, that is, the environmental factors that have a significant impact on its power generation performance are found, for example, solar photovoltaic corresponds to sunshine intensity, temperature, cloud thickness, humidity, etc.; wind power generation corresponds to wind speed, wind direction, temperature, air pressure, etc.; biomass power generation corresponds to raw material supply stability, fermentation / combustion efficiency, ambient temperature, etc.

[0041] The power quality index prediction characteristic value acquisition module 40 also includes, according to the environmental factor, performing multi-level sorting and identification of the power generation of the distributed power type information, and generating the first distributed power predicted power generation. By comprehensively considering the type characteristics of the distributed power source and the environmental factor factors in which it is located, a prediction model is used to predict the power generation of the distributed power source under different environmental conditions, and a multi-level sorting and identification is performed on it. Specifically, the prediction model is constructed and trained based on various factors such as historical data, meteorological forecasts, and technical parameters of the distributed power source. The environmental factor factor is used as an input variable, and the corresponding predicted power generation of the distributed power source is calculated through the model, and then a multi-level sorting and identification is performed on it. The predicted power generation is divided into different levels or intervals according to the standard, such as high, medium, and low levels or specific power generation ranges. Through the multi-level sorting identification, the power generation potential and stability of the distributed power source under different environmental conditions can be intuitively understood.

[0042] The power quality index prediction characteristic value acquisition module 40 also includes, based on the array position identifier, adding the first distributed power supply predicted power generation into the distributed power supply predicted power generation array. According to the unique position identifier of each distributed power supply, the first distributed power supply predicted power generation is added into the distributed power supply predicted power generation array, and each position in the distributed power supply predicted power generation array stores the predicted power generation data of the corresponding distributed power supply.

[0043] The specific configuration of the power quality index prediction characteristic value acquisition module 40 will be described in detail below. The power quality index prediction characteristic value acquisition module 40 may further include: configuring the retrieval factor item k value, wherein the total number of environmental factor factors ≥ the retrieval factor item k value ≥ 1, and the retrieval factor item k value initial value = 1; traversing the environmental factor factors and configuring the deviation tolerance interval; extracting k items of sorting factor factors from the environmental factor factors according to the retrieval factor item k value; extracting k items of deviation tolerance intervals from the deviation tolerance interval based on the k items of sorting factor factors; collecting power generation data for the distributed power type information based on the k items of sorting factor factors, and storing the retrieved data when the k items of sorting factor factors of the retrieved data all meet the k items of deviation tolerance intervals, otherwise, deleting the retrieved data; judging whether the k value is equal to the total number of environmental factor factors; if equal, performing a centralized trend analysis on the stored retrieval data power generation to obtain the predicted power generation of the first distributed power source; if not equal, adding one to k. During initialization, a search factor item k is set, and its range is between 1 and the total number of environmental factors (including 1 and the total number). The initial value of the search factor item k is 1, which means that only one environmental factor is considered at the beginning. Then the environmental factors are traversed and the deviation tolerance interval is configured. That is, for each environmental factor, a deviation tolerance interval is set, which defines the range within which the environmental factor is considered normal. Then, from all the environmental factors, k sorting factors are selected according to the current k value. For each selected sorting factor, the corresponding interval is extracted from the previously configured deviation tolerance interval to determine whether the collected power generation data meets the requirements. Then, the power generation data of distributed power sources under different environmental conditions are collected. For each collected data, check whether the corresponding k sorting factor factors fall within the corresponding deviation tolerance interval. If all data items meet the conditions, store them, that is, retain the data collected this time. Otherwise, delete the data. Then determine whether the k value is equal to the total number of environmental factor factors, that is, check whether the current k value has reached the total number of environmental factor factors. If the k value is equal to the total number of environmental factor factors, perform a central trend analysis on all retained power generation data (such as calculating the average, median, etc.) to obtain the predicted power generation of the first distributed power source. If the k value has not reached the total number of environmental factor factors, increase the k value by 1 and repeat the power generation data collection until the k value is equal to the total number of environmental factor factors.

[0044] The specific configuration of the power quality index prediction characteristic value acquisition module 40 will be described in detail below. The power quality index prediction characteristic value acquisition module 40 may further include: performing a grey correlation analysis on the environmental factor based on the distributed power type information to obtain a grey correlation analysis result. By performing a grey correlation analysis on the distributed power type information and the environmental factor, it is possible to determine which environmental factors have the greatest impact on the distributed power supply, and sort and select them. Specifically, the distributed power type information and the environmental factor data are standardized to eliminate the impact of different dimensions; based on the standardized data, the correlation coefficient between each environmental factor and the distributed power type information is calculated, and the correlation coefficient is weighted averaged to obtain the correlation of each environmental factor, and then the grey correlation analysis result is obtained. The correlation calculation formula is as follows: ; in, Indicates The environmental factor and the distributed power type information are in the The correlation coefficient at each moment, is the number of data points.

[0045] The power quality index prediction characteristic value acquisition module 40 also includes, according to the grey correlation analysis result, sorting the environmental factor factors from large to small to obtain the environmental factor factor sorting result. According to the grey correlation degree of each environmental factor, sorting is performed from large to small to obtain the environmental factor factor sorting result. It also includes, according to the retrieval factor item k value, sorting k items of sorting factor factors according to the environmental factor factor sorting result. According to the preset retrieval factor item k value, the most important k factors are selected from the sorted environmental factor factors as sorting factor factors. Specifically, according to the sorting result of the environmental factor factors, the first k factors are selected from the top of the list as sorting factor factors to obtain a set of sorting factor factors, which includes the most important k environmental factor factors.

[0046] Although the present application makes various references to certain components in the platform according to the embodiments of the present application, any number of different components may be used and run on the user terminal and / or server, and the various units and components included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be implemented; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0047] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A risk monitoring and early warning platform for power grid projects, characterized by: include: A substation power grid equivalent topology structure construction module, used to construct an equivalent topology structure of the substation power grid, wherein the equivalent topology structure of the substation power grid includes distributed power source deployment location label information and distributed power source type label information; A power quality prediction channel generation module is used to build a graph neural network based on the equivalent topological structure of the substation power grid, and to configure the graph neural network by traversing the power quality indicators based on the distributed power source deployment location label information and the distributed power source type label information to generate a power quality prediction channel; The power quality index and power generation array acquisition module is used to monitor the initial characteristic value of the power quality index and the distributed power generation predicted power generation array when receiving the distributed power generation grid connection signal; A power quality index prediction characteristic value acquisition module, used to obtain a power quality index prediction characteristic value by parsing the power quality index initial characteristic value and the distributed power supply predicted power generation array through the power quality prediction channel; The distributed power grid connection warning signal generation module is used to generate a distributed power grid connection warning signal and send it to the user end when the power quality index prediction characteristic value does not meet the power quality index threshold range.

2. The risk monitoring and early warning platform for power grid engineering according to claim 1, characterized in that: Building a graph neural network based on the equivalent topological structure of the substation power grid, configuring the graph neural network by traversing the power quality indicators based on the distributed power source deployment location label information and the distributed power source type label information, and generating a power quality prediction channel, including: Extracting a first power quality indicator according to the power quality indicator; Taking the equivalent topological structure of the substation power grid, the distributed power source deployment location label information and the distributed power source type label information as constraints, collecting the initial record value of the first power quality indicator, the distributed power source generation array record value and the first power quality indicator grid connection record value, and obtaining the first power quality indicator prediction sub-channel construction data set; Retrieving the power quality prediction channel to construct a data set, configuring the graph neural network, and generating a first power quality indicator prediction subchannel; The first power quality indicator prediction sub-channel is added to the power quality prediction channel.

3. The risk monitoring and early warning platform for power grid engineering according to claim 2, characterized in that: Retrieving the power quality prediction channel to construct a data set, configuring the graph neural network, and generating a first power quality indicator prediction sub-channel, including: Constructing a first loss function, wherein the first loss function is used to count the proportion of output deviations greater than or equal to an output deviation threshold for a preset number of consecutive trainings; Constructing a second loss function, wherein the second loss function is used to count the output deviation variance of a preset number of consecutive trainings; The power quality prediction channel is retrieved to construct a data set, and the graph neural network is configured to generate the first power quality indicator prediction sub-channel when the first loss function is less than or equal to the first loss threshold and the second loss function is less than or equal to the second loss threshold.

4. The risk monitoring and early warning platform for power grid engineering according to claim 3, characterized in that: Retrieving the power quality prediction channel to construct a data set, configuring the graph neural network, and generating the first power quality indicator prediction subchannel when the first loss function is less than or equal to a first loss threshold and the second loss function is less than or equal to a second loss threshold, including: The power quality prediction channel construction data set is equally divided into Q groups of power quality prediction channel construction data; Performing Q sampling with replacement on the Q groups of power quality prediction channel construction data to construct a first power quality prediction channel construction data set; The loop is executed Z times to obtain a second power quality prediction channel construction data set until a Zth power quality prediction channel construction data set; Traversing the first power quality prediction channel construction data set, the second power quality prediction channel construction data set until the Zth power quality prediction channel construction data set, configuring the graph neural network, and constructing Z basis channels when the first loss function is less than or equal to a first loss threshold and the second loss function is less than or equal to a second loss threshold; A fully connected layer is constructed, and the accuracy of the Z basis channels is used as a weight to fuse the Z basis channels to generate the first power quality indicator prediction sub-channel.

5. The risk monitoring and early warning platform for power grid engineering according to claim 2, characterized in that: The power quality index at least includes a voltage fluctuation range index, a current fluctuation range index and a harmonic distortion index.

6. The risk monitoring and early warning platform for power grid engineering according to claim 1, characterized in that: The power quality indicator prediction characteristic value is obtained by analyzing the power quality indicator initial characteristic value and the distributed power supply predicted power generation array through the power quality prediction channel, which includes: Obtaining first distributed power basic information, wherein the first distributed power basic information includes distributed power type information, distributed power scale information, and array position identifier; Based on the distributed power type information, matching the environmental factor factors through the environmental factor factor calibration table; According to the environmental factor, the distributed power source type information is sorted and labeled for power generation in multiple levels to generate a predicted power generation of the first distributed power source; Based on the array position identifier, the predicted power generation of the first distributed power source is added to the predicted power generation array of the distributed power sources.

7. The risk monitoring and early warning platform for power grid engineering according to claim 6, characterized in that: According to the environmental factor, the distributed power source type information is sorted and labeled for power generation in multiple levels to generate a first distributed power source predicted power generation, including: Configure the retrieval factor item k value, wherein the total number of environmental factor factors ≥ the retrieval factor item k value ≥ 1, and the initial value of the retrieval factor item k value = 1; Traversing the environmental factors and configuring the deviation tolerance interval; Extracting k sorting factor factors from the environmental factor factors according to the retrieval factor item k value; Extracting k deviation tolerance intervals from the deviation tolerance interval based on the k sorting factor factors; Based on the k sorting factor factors, the power generation data of the distributed power type information is collected, and when the k sorting factor factors of the retrieved data all meet the k deviation tolerance intervals, the retrieved data is stored, otherwise, the retrieved data is deleted; Determine whether the k value is equal to the total number of environmental factor factors; If they are equal, performing a centralized trend analysis on the stored retrieved data power generation to obtain the predicted power generation of the first distributed power source; If not equal, k increases by one.

8. The risk monitoring and early warning platform for power grid engineering according to claim 7, characterized in that: Extracting k items of sorting factor factors from the environmental factor factors according to the retrieval factor item k value includes: Performing grey correlation analysis on the environmental factor based on the distributed power type information to obtain a grey correlation analysis result; According to the grey relational analysis result, the environmental factor factors are sorted from large to small to obtain the environmental factor factor sorting result; According to the k value of the search factor item, k items of sorting factor factors are sorted according to the ranking result of the environmental factor factors.

Images