Power grid fault prediction method and system based on artificial intelligence

Abstract

The invention discloses a power grid fault prediction method and system based on artificial intelligence, and relates to the technical field of power grid fault prediction management. The power grid fault prediction method based on artificial intelligence comprises the following steps: S1, evaluating the accuracy of power grid data; s2, evaluating the real-time performance of the power grid data; and S3, power grid fault evaluation. According to the method, the power grid data accuracy evaluation value is obtained through the power grid accuracy data, whether power grid data preprocessing is carried out or not is judged based on the power grid data accuracy evaluation value, and then the power grid data real-time evaluation value is obtained according to the extraction duration and the power grid real-time data. Whether data real-time optimization is carried out or not is judged based on the power grid data real-time evaluation value, finally, the power grid fault evaluation value is obtained according to the power grid characteristic data, whether fault early warning is carried out or not is judged based on the power grid fault evaluation value, and the effect of improving the power grid asymmetric fault prediction accuracy is achieved. The problem that in the prior art, power grid asymmetric fault prediction is low in accuracy is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid fault prediction and management, and in particular to a power grid fault prediction method and system based on artificial intelligence. Background Art

[0002] As an important infrastructure of modern energy system, the safety, stability and reliability of power grid are directly related to the normal operation of social economy. With the expansion of modern power grid scale and the complexity of structure, asymmetric faults have become an important issue affecting the safe operation of power grid. Asymmetric faults refer to the phenomenon of unbalanced current or voltage in each phase of the power grid. Such faults will not only affect the stable operation of power grid equipment, but also may lead to the decline of power quality and large-scale power outages. Therefore, the rapid prediction and diagnosis of asymmetric faults in power grids are of great significance.

[0003] Existing technologies use historical and real-time data on power grid operation, combined with machine learning, deep learning and other technologies, to model, predict and diagnose asymmetric faults and achieve the prediction of power grid faults.

[0004] For example, the patent application with publication number: CN115545970A discloses a digital twin-based power grid fault analysis method, system, equipment and medium, including: obtaining historical fault data under different fault types during the operation of the power grid, and constructing a power grid digital twin system framework; based on the power grid digital twin system framework, using historical fault data to train the convolutional recurrent neural network model until the model converges, and then generating a fault prediction model; inputting the power grid operation data to be predicted into the fault prediction model, outputting the fault prediction result, and matching different fault handling solutions according to the fault level of the fault prediction result.

[0005] For example, the patent application with publication number: CN114511418A discloses a method for calculating short-circuit current of a distribution network containing an inverter-type distributed power source, which includes: step 1: establishing the output characteristic equation of the distributed power source, the node admittance matrix of the network and the boundary conditions of the fault point; step 2: based on step 1, using the linear iteration method for cyclic calculation; step 3: based on steps 1 and 2, performing second-order convergence iterative calculation based on the Newton-Raphson principle and the positive-sequence augmented network node voltage equation to calculate and obtain the node voltages and branch currents of the positive-sequence network; step 4: based on the calculation results of step 3, combined with the fault point boundary conditions of step 1, calculating the node voltages and branch currents of the negative-sequence and zero-sequence networks, and using the symmetrical component method to synthesize the fault node phase voltages and branch fault currents of the distribution network, and outputting the results.

[0006] However, in the process of implementing the technical solution of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems: In the prior art, when a power grid system fails, the data accuracy may be reduced, and due to the weak characteristics of asymmetric faults, the prediction model may not be able to accurately capture the power grid fault, resulting in long training time, poor data real-time performance, and low accuracy in predicting power grid asymmetric faults. Summary of the invention

[0007] The embodiments of the present application solve the problem of low accuracy in predicting asymmetric power grid faults in the prior art by providing an artificial intelligence-based power grid fault prediction method and system, thereby improving the accuracy in predicting asymmetric power grid faults.

[0008] The embodiment of the present application provides a power grid fault prediction method based on artificial intelligence, comprising the following steps: S1, analyzing the acquired power grid data to obtain power grid data accuracy data, evaluating the accuracy of the power grid data according to the acquired power grid accuracy data to obtain a power grid data accuracy evaluation value, judging whether to perform power grid data preprocessing based on the acquired power grid data accuracy evaluation value, and the power grid data accuracy evaluation value is used to quantitatively evaluate data quality; S2, if data preprocessing is performed, extracting the optimized power grid data to obtain power grid characteristic data and extraction time, evaluating the real-time performance of the power grid data according to the extraction time and the acquired power grid real-time data to obtain a power grid data real-time evaluation value, judging whether to perform data real-time optimization based on the acquired power grid data real-time evaluation value, and the power grid data real-time evaluation value is used to quantitatively evaluate the real-time performance of the data; S3, if data real-time optimization is performed, evaluating the fault of the power grid according to the acquired power grid characteristic data to obtain a power grid fault evaluation value, judging whether to perform a fault warning based on the acquired power grid fault evaluation value, and the power grid fault evaluation value is used to quantitatively evaluate the probability of occurrence of asymmetric power grid faults.

[0009] Furthermore, the power grid data includes current and voltage; the power grid accuracy data includes data missing rate, standardization score, data signal-to-noise ratio and sample balance coefficient; the data missing rate is obtained by statistically analyzing the amount of power grid data in a preset time period and the preset total data amount; the standardization score is obtained by performing a ratio operation on the difference between the power grid data at a preset time point in the preset time period and the average value of the power grid data and the standard deviation of the power grid data; the data signal-to-noise ratio is obtained by processing the signal power and the noise power; the sample balance coefficient is obtained by performing a ratio operation on the number of asymmetric fault events in the historical database and the total number of events; the power grid real-time data includes the voltage waveform change, the current waveform change, the power grid data amount and the power grid data update frequency; the power grid characteristic data includes voltage imbalance, negative sequence current component, zero sequence current component, voltage-current phase difference, voltage waveform distortion and current waveform distortion; the voltage imbalance is obtained by performing a ratio operation on the difference between the maximum voltage value and the minimum voltage value and the voltage average value.

[0010] Furthermore, the specific method for performing accuracy assessment of power grid data based on the acquired power grid accuracy data to obtain a power grid data accuracy assessment value is as follows: obtaining a data signal-to-noise ratio deviation based on a relative relationship between the data signal-to-noise ratio and a preset data signal-to-noise ratio in a database, wherein the data signal-to-noise ratio deviation is obtained by performing a ratio operation on the data signal-to-noise ratio and the preset data signal-to-noise ratio in the database; obtaining a sample balance coefficient deviation based on a relative relationship between a sample balance coefficient and a preset sample balance coefficient in a database, wherein the sample balance coefficient deviation is obtained by performing a difference operation on the sample balance coefficient and the preset sample balance coefficient in the database; processing the power grid factor, the data missing rate, the data signal-to-noise ratio deviation and the sample balance coefficient deviation to obtain a power grid data accuracy assessment value, wherein the power grid factor represents the degree of influence of the standardized score on the power grid data accuracy assessment value.

[0011] Furthermore, the specific process of judging whether to perform grid data preprocessing based on the acquired grid data accuracy assessment value is as follows: judging whether the grid data accuracy assessment value is less than a preset grid data accuracy threshold: if the grid data accuracy assessment value is less than the preset grid data accuracy threshold, grid data preprocessing is performed; if the grid data accuracy assessment value is not less than the preset grid data accuracy threshold, grid data preprocessing is not performed.

[0012] Furthermore, the specific process of the power grid data preprocessing is as follows; A1, determine whether there are outliers in the power grid data. If there are outliers in the power grid data, remove the outliers, otherwise execute A2; A2, determine whether the data missing rate in the power grid data is greater than 0. If the data missing rate in the power grid data is greater than 0, perform missing value processing, otherwise unify the data format; the power grid data preprocessing includes removing outliers, processing missing values ​​and unifying the data format.

[0013] Furthermore, the specific method for evaluating the real-time performance of power grid data based on the extraction time and the acquired real-time data of the power grid to obtain the real-time performance evaluation value of the power grid data is as follows: obtaining the extraction time deviation based on the relative relationship between the extraction time and the preset extraction time in the database, and the extraction time deviation is obtained by performing a ratio operation on the extraction time and the preset extraction time in the database; obtaining the power grid data update frequency deviation based on the relative relationship between the power grid data update frequency and the preset power grid data update frequency in the database, and the power grid data update frequency deviation is obtained by performing a ratio operation on the power grid data update frequency and the preset power grid data update frequency in the database; obtaining the power grid data volume deviation based on the relative relationship between the power grid data volume and the preset power grid data volume in the database, and the power grid data volume deviation is obtained by performing a ratio operation on the power grid data volume and the preset power grid data volume in the database; processing the optimized power grid data accuracy evaluation value, waveform change factor, extraction time deviation, power grid data update frequency deviation and power grid data volume deviation to obtain the real-time performance evaluation value of the power grid data.

[0014] Furthermore, the specific method for judging whether to perform data real-time optimization based on the acquired power grid data real-time evaluation value is as follows: judging whether the power grid data real-time evaluation value is less than the preset power grid data real-time threshold: if the power grid data real-time evaluation value is less than the preset power grid data real-time threshold, data real-time optimization is performed; if the power grid data real-time evaluation value is not less than the preset power grid data real-time threshold, data real-time optimization is not performed; the data real-time optimization includes feature selection optimization and extraction time optimization; the extraction time optimization includes parallel computing, data compression and incremental learning.

[0015] Furthermore, the specific method for performing grid fault assessment based on the acquired grid characteristic data to obtain a grid fault assessment value is as follows: a voltage imbalance deviation is obtained based on the relative relationship between the voltage imbalance and a preset voltage imbalance in a database, and the voltage imbalance deviation is obtained by performing a ratio operation on the voltage imbalance and the preset voltage imbalance in the database; a negative-sequence current component deviation is obtained based on the relative relationship between the negative-sequence current component and a preset negative-sequence current component in the database, and the negative-sequence current component deviation is obtained by performing a ratio operation on the negative-sequence current component and the preset negative-sequence current component in the database; a zero-sequence current component deviation is obtained based on the relative relationship between the zero-sequence current component and a preset zero-sequence current component in the database, and the zero-sequence current component deviation is obtained The current component deviation is obtained by performing a ratio operation on the zero-sequence current component and the preset zero-sequence current component in the database; the grid phase difference coefficient is obtained according to the relative relationship between the voltage-current phase difference and the preset voltage-current phase difference in the database; the voltage waveform distortion deviation is obtained according to the relative relationship between the voltage waveform distortion and the preset voltage waveform distortion in the database, and the current waveform distortion deviation is obtained according to the relative relationship between the current waveform distortion and the preset current waveform distortion in the database. The voltage waveform distortion deviation and the current waveform distortion deviation are processed to obtain the grid waveform distortion coefficient; the voltage unbalance deviation, the negative-sequence current component deviation, the zero-sequence current component deviation, the grid phase difference coefficient and the grid waveform distortion coefficient are processed to obtain the grid fault assessment value.

[0016] Furthermore, the specific process of determining whether to perform a fault warning based on the acquired power grid fault assessment value is as follows: determining whether the power grid fault assessment value is less than a preset power grid fault threshold: if the power grid fault assessment value is less than the preset power grid fault threshold, no fault warning is performed; if the power grid fault assessment value is not less than the preset power grid fault threshold, a fault warning is performed; the fault warning indicates identifying the type of power grid fault and sending warning information to a preset person.

[0017] The embodiment of the present application provides an artificial intelligence-based power grid fault prediction system, including: a power grid data accuracy evaluation module, a power grid data real-time evaluation module and a power grid fault evaluation module; wherein the power grid data accuracy evaluation module is used to analyze the acquired power grid data to obtain power grid data accuracy data, perform power grid data accuracy evaluation according to the acquired power grid accuracy data to obtain a power grid data accuracy evaluation value, and judge whether to perform power grid data preprocessing based on the acquired power grid data accuracy evaluation value, and the power grid data accuracy evaluation value is used to quantitatively evaluate data quality; the power grid data real-time evaluation module is used to extract the optimized power grid data to obtain power grid characteristic data and extraction time if data preprocessing is performed, perform power grid data real-time evaluation according to the extraction time and the acquired power grid real-time data to obtain a power grid data real-time evaluation value, and judge whether to perform data real-time optimization based on the acquired power grid data real-time evaluation value, and the power grid data real-time evaluation value is used to quantitatively evaluate the real-time of the data; the power grid fault evaluation module is used to perform power grid fault evaluation according to the acquired power grid characteristic data to obtain a power grid fault evaluation value if data real-time optimization is performed, and judge whether to perform fault warning based on the acquired power grid fault evaluation value, and the power grid fault evaluation value is used to quantitatively evaluate the probability of occurrence of power grid asymmetric fault.

[0018] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. The grid data accuracy evaluation value obtained through the grid accuracy data is used to determine whether to preprocess the grid data. Then, the grid data real-time evaluation value obtained based on the extraction time and the grid real-time data is used to determine whether to optimize the data real-time performance. Finally, the grid fault evaluation value obtained based on the grid characteristic data is used to determine whether to perform fault warning. This improves the quality of the grid data, thereby achieving an improvement in the accuracy of grid asymmetric fault prediction, and effectively solves the problem of low grid asymmetric fault prediction accuracy in the prior art.

[0019] 2. The data signal-to-noise ratio deviation is obtained through the relative relationship between the data signal-to-noise ratio and the preset data signal-to-noise ratio in the database, and then the sample balance coefficient deviation is obtained according to the relative relationship between the sample balance coefficient and the preset sample balance coefficient in the database. Finally, the power grid factor, data missing rate, data signal-to-noise ratio deviation and sample balance coefficient deviation are processed to obtain the power grid data accuracy assessment value, thereby quantitatively evaluating the quality of the power grid data and further improving the quality of the power grid data in power grid faults.

[0020] 3. The extraction time deviation is obtained by comparing the extraction time with the preset extraction time, and then the grid data update frequency deviation is obtained according to the grid data update frequency and the preset grid data update frequency. Then, the grid data volume deviation is obtained according to the grid data volume and the preset grid data volume. Finally, the optimized grid data accuracy assessment value, waveform change factor, extraction time deviation, grid data update frequency deviation and grid data volume deviation are processed to obtain the grid data real-time assessment value, thereby quantitatively evaluating the real-time of the grid data, and then realizing the improvement of the real-time of grid fault warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A flowchart of an artificial intelligence-based power grid fault prediction method provided in an embodiment of the present application.

[0022] Figure 2 A schematic diagram of how the power grid data accuracy assessment value varies with the data missing rate provided in an embodiment of the present application.

[0023] Figure 3 A schematic diagram of how the accuracy assessment value of power grid data changes with the data signal-to-noise ratio provided in an embodiment of the present application.

[0024] Figure 4 A schematic diagram of the change of the grid data accuracy assessment value with the sample balance coefficient provided in an embodiment of the present application.

[0025] Figure 5 A schematic diagram of the structure of an artificial intelligence-based power grid fault prediction system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] The embodiments of the present application solve the problem of low accuracy in predicting asymmetric power grid faults in the prior art by providing an artificial intelligence-based power grid fault prediction method and system. A power grid data accuracy evaluation value is obtained through power grid accuracy data, and whether to perform power grid data preprocessing is determined based on the power grid data accuracy evaluation value. Then, a power grid data real-time evaluation value is obtained based on the extraction time and power grid real-time data, and whether to perform data real-time optimization is determined based on the power grid data real-time evaluation value. Finally, a power grid fault evaluation value is obtained based on power grid characteristic data, and whether to perform fault warning is determined based on the power grid fault evaluation value, thereby improving the accuracy of power grid asymmetric fault prediction.

[0027] The technical solution in the embodiment of the present application is to solve the problem of low accuracy in predicting asymmetric faults in the above-mentioned power grid, and the overall idea is as follows: The grid data accuracy assessment value obtained through the grid accuracy data is used to determine whether to perform grid data preprocessing, and then the grid data real-time assessment value obtained based on the extraction time and grid real-time data is used to determine whether to perform data real-time optimization. Finally, the grid fault assessment value obtained based on the grid characteristic data is used to determine whether to perform fault warning, thereby achieving the effect of improving the accuracy of grid asymmetric fault prediction.

[0028] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0029] like Figure 1 As shown, it is a flow chart of the power grid fault prediction method based on artificial intelligence provided by an embodiment of the present application, and the method includes the following steps: S1, power grid data accuracy assessment: analyzing the acquired power grid data to obtain power grid data accuracy data, performing power grid data accuracy assessment according to the acquired power grid accuracy data to obtain a power grid data accuracy assessment value, judging whether to perform power grid data preprocessing based on the acquired power grid data accuracy assessment value, and the power grid data accuracy assessment value is used to quantitatively evaluate data quality; S2, power grid data real-time assessment: if data preprocessing is performed, then the optimized power grid data is extracted to obtain power grid characteristic data and extraction time, and the real-time evaluation of the power grid data is performed according to the extraction time and the acquired power grid real-time data to obtain a power grid data real-time assessment value, and judging whether to perform data real-time optimization based on the acquired power grid data real-time assessment value, and the power grid data real-time assessment value is used to quantitatively evaluate the real-time of the data; S3, power grid fault assessment: if data real-time optimization is performed, then the power grid fault assessment is performed according to the acquired power grid characteristic data to obtain a power grid fault assessment value, and judging whether to perform fault warning based on the acquired power grid fault assessment value, and the power grid fault assessment value is used to quantitatively evaluate the probability of occurrence of power grid asymmetric fault.

[0030] It should be added that the power grid data includes current and voltage; the power grid accuracy data includes data missing rate, standardized score, data signal-to-noise ratio and sample balance coefficient.

[0031] The specific method of obtaining the power grid accuracy data is as follows: the data missing rate is obtained by statistically analyzing the data volume of the power grid data within a preset time period and the preset total data volume; the standardized score is obtained by performing a ratio operation on the difference between the power grid data at a preset time point within the preset time period and the average value of the power grid data and the standard deviation of the power grid data; the data signal-to-noise ratio is obtained by processing the signal power and the noise power; the sample balance coefficient is obtained by performing a ratio operation on the number of asymmetric fault events in the historical database and the total number of events; the preset time period and the preset time point are set according to the real-time application requirements, for example, if the power grid status needs to be monitored in real time, the preset time period is set to 1 minute and the preset time point is set to 10 seconds; the preset total data volume is set according to, for example, if the acquisition frequency is 1 data point / second and the acquisition time is 1 minute, the preset total data volume is 60 data points, that is, the preset total data volume is the product of the acquisition frequency and the acquisition time; the total number of events includes the number of fault events and the number of normal events; the number of fault events includes the number of asymmetric fault events and the number of symmetric fault events.

[0032] The real-time data of the power grid includes the change in voltage waveform, the change in current waveform, the amount of power grid data and the update frequency of power grid data.

[0033] The specific method of acquiring real-time data of the power grid is as follows: the voltage waveform change represents the change of the voltage waveform within a preset time period; the current waveform change represents the change of the current waveform within a preset time period; the power grid data volume is represented by the real-time statistics of the power grid data; the power grid update frequency is represented by the collection frequency of the power grid data.

[0034] The grid characteristic data include voltage unbalance, negative sequence current component, zero sequence current component, voltage-current phase difference, voltage waveform distortion and current waveform distortion.

[0035] The specific method of acquiring the grid characteristic data is as follows: the voltage unbalance is obtained by calculating the ratio of the difference between the maximum voltage value and the minimum voltage value to the voltage average value; the negative sequence current component is obtained by performing phasor calculation on the amplitude of the three-phase current; the zero sequence current component is obtained by performing average calculation on the sum of the three-phase currents; the voltage-current phase difference is obtained by performing phase calculation on the time domain signals of the voltage and current using the cross-correlation function; the voltage waveform distortion is obtained by extracting the various frequency components in the voltage waveform through Fourier transform and calculating the total harmonic distortion; the current waveform distortion is obtained by extracting the various frequency components in the current waveform through Fourier transform and calculating the total harmonic distortion.

[0036] In this embodiment, through the grid data accuracy evaluation and preprocessing steps, errors and anomalies in the data can be identified and corrected, thereby improving the accuracy and reliability of the data; through the grid data real-time evaluation and optimization steps, the timeliness of the data can be improved; thereby achieving an improvement in the accuracy of grid asymmetric fault prediction, effectively solving the problem of low grid asymmetric fault prediction accuracy in the prior art. Further, the specific method of evaluating the accuracy of the power grid data based on the acquired power grid accuracy data to obtain the power grid data accuracy evaluation value is as follows: the data signal-to-noise ratio deviation (i.e. ), the data signal-to-noise ratio deviation is obtained by performing a ratio operation on the data signal-to-noise ratio and the preset data signal-to-noise ratio in the database; the sample balance coefficient deviation (i.e. ), the sample balance coefficient deviation is obtained by performing a difference operation on the sample balance coefficient and the preset sample balance coefficient in the database; the power grid factor, data missing rate, data signal-to-noise ratio deviation and sample balance coefficient deviation are processed to obtain the power grid data accuracy assessment value, and the power grid factor represents the influence of the standardized score on the power grid data accuracy assessment value.

[0037] Among them, the specific method for obtaining the accuracy evaluation value of power grid data is: ; In the formula, i represents the number of the preset time period, , N represents the total number of preset time periods, represents the accuracy evaluation value of the power grid data in the i-th preset time period, represents the data missing rate of the i-th preset time period, BZH represents the power grid factor, represents the data signal-to-noise ratio of the i-th preset time period, represents the sample balance coefficient of the i-th preset time period, Indicates the preset data signal-to-noise ratio, Indicates the preset sample balance coefficient.

[0038] In this embodiment, the preset data signal-to-noise ratio is represented by the average value of the data signal-to-noise ratio in the historical time period; the preset sample balance coefficient is set according to the application requirements. For example, if the ratio of fault events to normal events is required to be 1:1, the preset sample balance coefficient is .

[0039] The grid factor is obtained from the database and represents the degree of influence of the standardized score on the grid data accuracy assessment value. For example, based on the relationship between the historical standardized scores (voltage standardized scores and current standardized scores) and the grid accuracy data (such as data missing rate, data signal-to-noise ratio), a mapping set of standardized scores and their corresponding influencing factors is constructed, and the real-time standardized scores are input into the mapping set to obtain the corresponding grid factor.

[0040] The power grid data accuracy assessment value in this algorithm involves processing multiple independent variables (data missing rate, data signal-to-noise ratio and sample balance coefficient), and there is a mutual influence relationship between these independent variables; a high data missing rate may lead to a lower signal-to-noise ratio; a higher signal-to-noise ratio can improve data quality, thereby improving the accuracy of the standardization process; when the data signal-to-noise ratio is lower, it is more difficult to extract valid signals from the power grid data. At this time, if the sample balance coefficient is lower, it is more difficult to extract features from fault events; in the power grid monitoring system, if the data missing rate of the fault event is greater, the missing proportion of the fault event will be higher, resulting in a smaller sample balance coefficient.

[0041] In this algorithm, the grid data accuracy assessment value is negatively correlated with the data missing rate, and is positively correlated with the data signal-to-noise ratio. The closer the sample balance coefficient is to the preset sample balance coefficient, the greater the grid data accuracy assessment value.

[0042] Take the grid factor as 1, the preset data signal-to-noise ratio as 1dB, and the preset sample balance coefficient as 1 as an example. Figure 2 As shown, it is a schematic diagram of the change of the power grid data accuracy evaluation value provided in an embodiment of the present application with the change of the data missing rate. When the data signal-to-noise ratio is 1dB and the sample balance coefficient is 1, the power grid data accuracy evaluation value decreases with the increase of the data missing rate.

[0043] like Figure 3 As shown, it is a schematic diagram of the change of the power grid data accuracy evaluation value provided in an embodiment of the present application with the change of the data signal-to-noise ratio. When the data missing rate is 0.5 and the sample balance coefficient is 1, the power grid data accuracy evaluation value increases with the increase of the data signal-to-noise ratio.

[0044] like Figure 4 The figure shows a schematic diagram of the change of the power grid data accuracy evaluation value provided in an embodiment of the present application with the change of the sample balance coefficient. When the data missing rate is 0.5, the data signal-to-noise ratio is 1dB, and the sample balance coefficient is less than 1, the power grid data accuracy evaluation value increases with the increase of the sample balance coefficient. When the data missing rate is 0.5, the data signal-to-noise ratio is 1, and the sample balance coefficient is not less than 1, the power grid data accuracy evaluation value decreases with the increase of the sample balance coefficient.

[0045] Through the above steps, the quality of the power grid data is quantitatively evaluated, thereby achieving improvement in the quality of the power grid data during power grid failures.

[0046] Furthermore, the specific process of determining whether to perform grid data preprocessing based on the acquired grid data accuracy assessment value is as follows: Determine whether the grid data accuracy assessment value is less than a preset grid data accuracy threshold: If the grid data accuracy assessment value is less than the preset grid data accuracy threshold, grid data preprocessing is performed; If the grid data accuracy assessment value is not less than the preset grid data accuracy threshold, grid data preprocessing is not performed.

[0047] It should be added that the specific process of power grid data preprocessing is as follows; A1, determine whether there are outliers in the power grid data. If there are outliers in the power grid data, remove the outliers, otherwise execute A2, the outliers represent data with a standardized score greater than the preset standardized score; A2, determine whether the data missing rate in the power grid data is greater than 0. If the data missing rate in the power grid data is greater than 0, perform missing value processing, otherwise unify the data format; power grid data preprocessing includes removing outliers, processing missing values ​​and unifying data formats.

[0048] In this embodiment, the preset grid data accuracy threshold is represented by the minimum value of the grid data accuracy assessment value within the historical time period; missing value processing: interpolation method is used to fill in missing values; unified data format: data in different units are converted into the same unit, and the consistency of the data format is ensured, for example, the voltage unit is unified as volts, the current unit is unified as amperes, and the timestamp format is unified as YYYY-MM-DD HH:MM:SS.

[0049] Through the above steps, the quality of power grid data is improved.

[0050] Further, the specific method of evaluating the real-time performance of the power grid data according to the extraction time and the obtained real-time performance data of the power grid to obtain the real-time performance evaluation value of the power grid data is as follows: the extraction time deviation (i.e. ), the extraction time deviation is obtained by performing a ratio operation on the extraction time and the preset extraction time in the database; the grid data update frequency deviation (i.e. ), the power grid data update frequency deviation is obtained by performing a ratio operation on the power grid data update frequency and the preset power grid data update frequency in the database; the power grid data volume deviation (i.e. ), the grid data volume deviation is obtained by performing a ratio operation on the grid data volume and the preset grid data volume in the database; the optimized grid data accuracy evaluation value, waveform change factor, extraction time deviation, grid data update frequency deviation and grid data volume deviation are processed to obtain the grid data real-time evaluation value.

[0051] Among them, the specific method for obtaining the real-time evaluation value of power grid data is: ; In the formula, i represents the number of the preset time period, , N represents the total number of preset time periods, represents the real-time evaluation value of the power grid data in the i-th preset time period, represents the optimized grid data accuracy evaluation value of the i-th preset time period, represents the extraction duration of the i-th preset time period, represents the grid data update frequency of the i-th preset time period, BXB represents the waveform change factor, represents the amount of power grid data in the i-th preset time period, Indicates the preset grid data update frequency. Indicates the preset power grid data volume, Indicates the preset extraction duration.

[0052] In this embodiment, the preset power grid data update frequency is represented by the average value of the power grid data update frequency in the historical time period; the preset power grid data volume is represented by the average value of the power grid data volume in the historical time period; and the preset extraction duration is represented by the average value of the extraction duration in the historical time period.

[0053] The waveform change factor is obtained from the database and represents the degree of influence of the waveform change on the real-time evaluation value of the power grid data. For example, based on the relationship between the historical waveform change (voltage waveform change and current waveform change) and the real-time evaluation data (such as extraction time and power grid data update frequency), a mapping set of waveform changes and their corresponding influencing factors is constructed, and the real-time waveform change is input into the mapping set to obtain the corresponding waveform change factor.

[0054] In this algorithm, the real-time evaluation value of power grid data involves processing multiple independent variables (extraction time, power grid data volume and power grid data update frequency), and there is a mutual influence relationship between these independent variables; when the amount of power grid data increases, the extraction time may be extended because more data needs to be processed; if the extraction time is longer, the speed of data update will be limited, which may lead to a decrease in the update frequency of power grid data; when the amount of power grid data increases, in order to maintain the real-time nature of the data, the power grid data may need to increase the update frequency; in summary, the real-time evaluation value of power grid data is negatively correlated with the extraction time and the amount of power grid data, and the real-time evaluation value of power grid data is positively correlated with the update frequency of power grid data.

[0055] Taking the optimized grid data accuracy evaluation value and waveform change factor as 1, the preset grid data update frequency as 1 second, the preset grid data volume as 100MB, and the preset extraction time as 1 second as an example, the change statistics of the real-time evaluation value of the grid data are shown in Table 1: Table 1 Statistics of changes in real-time evaluation values ​​of power grid data

[0056] It can be seen from the first and second groups of data in the table that when the extraction time decreases, the real-time evaluation value of the power grid data increases accordingly; it can be seen from the second and third groups of data that when the amount of power grid data decreases, the real-time evaluation value of the power grid data increases accordingly; it can be seen from the third and fourth groups of data that when the update frequency of power grid data increases, the real-time evaluation value of the power grid data increases accordingly.

[0057] Through the above steps, the real-time performance of power grid data is quantitatively evaluated, thereby improving the real-time performance of power grid fault warning.

[0058] Furthermore, the specific method for judging whether to perform data real-time optimization based on the acquired real-time evaluation value of power grid data is as follows: judging whether the real-time evaluation value of power grid data is less than the preset real-time threshold of power grid data: if the real-time evaluation value of power grid data is less than the preset real-time threshold of power grid data, data real-time optimization is performed; if the real-time evaluation value of power grid data is not less than the preset real-time threshold of power grid data, data real-time optimization is not performed; data real-time optimization includes feature selection optimization and extraction time optimization; feature selection optimization means reducing redundant power grid feature data through principal component analysis method; extraction time optimization includes parallel computing, data compression and incremental learning.

[0059] In this embodiment, the preset real-time threshold of the power grid data is represented by the minimum value of the real-time evaluation value of the power grid data in the historical time period.

[0060] Feature selection optimization: First, the power grid data set is standardized; then, the covariance matrix of the standardized data is calculated to evaluate the correlation between features; the covariance matrix is ​​eigendecomposed to obtain eigenvectors and eigenvalues; the size of the eigenvalue reflects the importance of the corresponding eigenvector in the data set; according to the size of the eigenvalue, the eigenvectors corresponding to the first few largest eigenvalues ​​are selected as the principal components; the original data is projected into the new feature space to obtain the reduced-dimensional data set.

[0061] Parallel computing: Parallel computing technology is used to accelerate the feature extraction process. For example, the power grid data set is divided into multiple subsets and processed simultaneously on multiple processors or computers, thereby increasing the speed of feature extraction. Data compression: Before extracting features, the power grid data is compressed to reduce the amount of data. For example, data compression algorithms (such as principal component analysis technology) are used to reduce the dimension of the data while retaining most of the information, thereby reducing the amount of calculation for subsequent feature extraction. Incremental learning: For example, when new power grid data arrives, the incremental learning algorithm is used to update the parameters to adapt to the new data distribution and characteristics, thereby saving computing time and resources.

[0062] Through the above steps, the power grid data processing time is reduced, thereby achieving the improvement of the real-time performance of the power grid characteristic data.

[0063] Further, the specific method of performing grid fault assessment based on the acquired grid characteristic data to obtain a grid fault assessment value is as follows: the voltage imbalance deviation (i.e. ), the voltage unbalance deviation is obtained by performing a ratio operation on the voltage unbalance and the preset voltage unbalance in the database; the negative sequence current component deviation (i.e. ), the negative-sequence current component deviation is obtained by performing a ratio operation on the negative-sequence current component and the preset negative-sequence current component in the database; the zero-sequence current component deviation is obtained according to the relative relationship between the zero-sequence current component and the preset zero-sequence current component in the database (i.e. ), the zero-sequence current component deviation is obtained by performing a ratio operation on the zero-sequence current component and the preset zero-sequence current component in the database; the grid phase difference coefficient (i.e. ); According to the relative relationship between the voltage waveform distortion and the preset voltage waveform distortion in the database, the voltage waveform distortion deviation (i.e. ), the current waveform distortion deviation (i.e. ), the voltage waveform distortion deviation and the current waveform distortion deviation are processed to obtain the grid waveform distortion coefficient (i.e. ); the voltage unbalance deviation, negative sequence current component deviation, zero sequence current component deviation, grid phase difference coefficient and grid waveform distortion coefficient are processed to obtain the grid fault assessment value.

[0064] Among them, the specific method for obtaining the power grid fault assessment value is: ; ; ; In the formula, i represents the number of the preset time period, , N represents the total number of preset time periods, represents the grid fault assessment value of the i-th preset time period, SZ represents the fault assessment factor, represents the voltage imbalance degree in the i-th preset time period, represents the negative sequence current component of the i-th preset time period, represents the zero-sequence current component of the i-th preset time period, represents the grid phase difference coefficient of the i-th preset time period, represents the grid waveform distortion coefficient of the i-th preset time period, represents the voltage-current phase difference in the i-th preset time period, represents the voltage waveform distortion in the i-th preset time period, represents the current waveform distortion in the i-th preset time period, Indicates the preset voltage imbalance, Represents the preset negative sequence current component, Represents the preset zero-sequence current component, Indicates the preset maximum voltage-current phase difference, Indicates that the preset voltage waveform is distorted. Indicates the preset current waveform distortion.

[0065] In this embodiment, the fault assessment factor is obtained from a database, which represents the degree of influence of the real-time assessment value of the power grid data and the accuracy assessment value of the power grid data on the power grid fault assessment value. For example, a mapping set of the channel power grid data real-time assessment value and the accuracy assessment value of the power grid data and their corresponding influencing factors is constructed based on the relationship between the historical power grid data real-time assessment value and the accuracy assessment value of the power grid data and the power grid characteristic data, and the real-time power grid data real-time assessment value and the accuracy assessment value of the power grid data are input into the mapping set to obtain the corresponding fault assessment factor.

[0066] The preset voltage imbalance is usually set according to the standards and regulations of the power grid. For example, the preset value of the voltage imbalance of a 35 kV bus is set to not more than 5%; the preset negative-sequence current component is set according to the stable operation requirements of the power grid and the tolerance of the equipment. For example, the preset value of the negative-sequence current component is set to not more than 10% of the rated current; the preset zero-sequence current component is set according to the grounding system and fault detection requirements of the power grid. For example, the preset zero-sequence current component is set to not more than a certain threshold (such as 100A); the preset maximum voltage-current phase difference is set according to the stable operation requirements of the power grid and the phase synchronization requirements of the equipment, for example, it is set to not more than a certain angle (such as 5°); the preset voltage waveform distortion and the preset current waveform distortion are set according to the power quality standards of the power grid and the tolerance of the equipment. For example, the preset voltage waveform distortion and the preset current waveform distortion are set to not more than a certain percentage (such as 5%).

[0067] The grid fault assessment value in this algorithm involves processing multiple independent variables (voltage unbalance, negative sequence current component, zero sequence current component, voltage-current phase difference, voltage waveform distortion and current waveform distortion), and there is a mutual influence relationship between these independent variables; when the voltage unbalance is large, a stronger negative sequence current component will be generated, because the negative sequence current is the direct consequence of voltage unbalance, therefore, an increase in voltage unbalance usually means that the negative sequence current will also increase; greater waveform distortion may lead to an increase in harmonics in the grid, which may also cause an increase in voltage unbalance; zero sequence current is usually related to ground faults (such as single-phase grounding), and a larger zero sequence current component may lead to an increase in voltage unbalance; when the voltage-current phase difference increases, it may lead to higher reactive power and harmonics in the grid, which may cause an increase in waveform distortion; a larger waveform distortion may in turn lead to an increase in the voltage-current phase difference; in summary, the grid fault assessment value is positively correlated with voltage unbalance, negative sequence current component, zero sequence current component, voltage-current phase difference, voltage waveform distortion and current waveform distortion.

[0068] Through the above steps, the probability of occurrence of asymmetric power grid faults is quantitatively evaluated, thereby achieving accurate prediction of asymmetric power grid faults.

[0069] Furthermore, the specific process of determining whether to perform a fault warning based on the acquired power grid fault assessment value is as follows: Determine whether the power grid fault assessment value is less than a preset power grid fault threshold: if the power grid fault assessment value is less than the preset power grid fault threshold, no fault warning is performed; if the power grid fault assessment value is not less than the preset power grid fault threshold, a fault warning is performed; the fault warning indicates identifying the type of power grid fault and sending a warning message to the preset personnel.

[0070] In this embodiment, the preset grid fault threshold is represented by the maximum value of the grid fault assessment value in the historical time period; the fault warning means analyzing the waveform signals of current and voltage through a convolutional neural network, identifying the fault type, and giving warning information during real-time monitoring. For example, when a ground fault occurs, the negative sequence current will increase significantly. The negative sequence current is monitored to see whether it is greater than the preset negative sequence current. If so, it indicates that a ground fault exists, and a warning message is sent to the preset personnel (quickly disconnecting the faulty line to avoid large-scale power outages).

[0071] Through the above steps, it is possible to identify potential power grid asymmetric fault types before a power grid fault occurs, thereby alleviating the impact of the power grid asymmetric fault.

[0072] like Figure 5 As shown, it is a structural schematic diagram of the power grid fault prediction system based on artificial intelligence provided in the embodiment of the present application. The power grid fault prediction system based on artificial intelligence provided in the embodiment of the present application includes: a power grid data accuracy evaluation module, a power grid data real-time evaluation module and a power grid fault evaluation module; wherein the power grid data accuracy evaluation module is used to analyze the acquired power grid data to obtain power grid data accuracy data, to evaluate the accuracy of the power grid data according to the acquired power grid accuracy data to obtain a power grid data accuracy evaluation value, to determine whether to perform power grid data preprocessing based on the acquired power grid data accuracy evaluation value, and the power grid data accuracy evaluation value is used to quantitatively evaluate the data quality; the power grid data real-time evaluation module If data preprocessing is performed, the optimized power grid data is extracted to obtain power grid characteristic data and extraction time, the real-time evaluation of the power grid data is evaluated based on the extraction time and the obtained power grid real-time data to obtain a power grid data real-time evaluation value, and whether data real-time optimization is performed is determined based on the obtained power grid data real-time evaluation value. The power grid data real-time evaluation value is used to quantitatively evaluate the real-time of the data; the power grid fault evaluation module is used to perform power grid fault evaluation based on the obtained power grid characteristic data to obtain a power grid fault evaluation value if data real-time optimization is performed, and whether fault warning is performed based on the obtained power grid fault evaluation value. The power grid fault evaluation value is used to quantitatively evaluate the probability of occurrence of asymmetric power grid faults.

[0073] In this embodiment, the principal component analysis method in artificial intelligence is used to evaluate the correlation between features, and the most important feature vector is selected as the principal component to achieve data dimensionality reduction processing; parallel computing technology is used to accelerate the feature extraction process, the power grid data is compressed by the data compression algorithm, and the power grid feature data is updated by the incremental learning algorithm, thereby improving the data processing speed and real-time performance; the convolutional neural network in artificial intelligence technology is used to analyze the waveform signals of current and voltage, and identify the fault type. In addition, artificial intelligence can also be used to automatically send early warning information. Once the convolutional neural network identifies the fault characteristics in the current and voltage waveform signals, the system can immediately trigger the early warning mechanism and generate early warning information to ensure that the preset personnel can receive the early warning in time; the application of artificial intelligence technology improves the efficiency and accuracy of power grid data processing, thereby achieving an improvement in the accuracy of power grid asymmetric fault prediction, and providing strong support for the safe and stable operation of the power grid.

[0074] To summarize, the embodiment of the present application determines whether to perform grid data preprocessing based on the grid data accuracy evaluation value obtained through grid accuracy data, and then determines whether to perform data real-time optimization based on the grid data real-time evaluation value obtained based on the extraction time and grid real-time data. Finally, it determines whether to perform fault warning based on the grid fault evaluation value obtained based on the grid characteristic data, thereby improving the quality of grid data, and further achieving an improvement in the accuracy of grid asymmetric fault prediction, effectively solving the problem of low accuracy of grid asymmetric fault prediction in the prior art.

[0075] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0076] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0077] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0078] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0079] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0080] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A power grid fault prediction method based on artificial intelligence, characterized in that: The following steps are involved: S1, analyzing the acquired power grid data to obtain power grid data accuracy data, evaluating the accuracy of the power grid data according to the acquired power grid accuracy data to obtain a power grid data accuracy evaluation value, and judging whether to perform power grid data preprocessing based on the acquired power grid data accuracy evaluation value, wherein the power grid data accuracy evaluation value is used to quantitatively evaluate data quality; S2, if data preprocessing is performed, the optimized power grid data is extracted to obtain power grid characteristic data and extraction time, the real-time performance of the power grid data is evaluated according to the extraction time and the obtained power grid real-time performance data to obtain a power grid data real-time performance evaluation value, and whether to perform data real-time performance optimization is determined based on the obtained power grid data real-time performance evaluation value, wherein the power grid data real-time performance evaluation value is used to quantitatively evaluate the real-time performance of the data; S3: If data real-time optimization is performed, a grid fault assessment is performed based on the acquired grid characteristic data to obtain a grid fault assessment value, and a fault warning is determined based on the acquired grid fault assessment value. The grid fault assessment value is used to quantitatively assess the probability of occurrence of asymmetric grid faults.

2. The power grid fault prediction method based on artificial intelligence as claimed in claim 1, characterized in that: The power grid data includes current and voltage; The power grid accuracy data includes data missing rate, standardized score, data signal-to-noise ratio and sample balance coefficient; The data missing rate is obtained by statistically analyzing the amount of power grid data within a preset time period and the preset total data amount; The standardized score is obtained by performing a ratio operation on the difference between the power grid data at a preset time point within a preset time period and the average value of the power grid data and the standard deviation of the power grid data; The data signal-to-noise ratio is obtained by processing the signal power and the noise power; The sample balance coefficient is obtained by calculating the ratio of the number of asymmetric fault events to the total number of events in the historical database; The real-time data of the power grid includes the voltage waveform change, the current waveform change, the power grid data volume and the power grid data update frequency; The grid characteristic data includes voltage unbalance, negative sequence current component, zero sequence current component, voltage-current phase difference, voltage waveform distortion and current waveform distortion; The voltage imbalance degree is obtained by performing a ratio operation on the difference between the maximum voltage value and the minimum voltage value and the voltage average value.

3. The power grid fault prediction method based on artificial intelligence as claimed in claim 2, characterized in that: The specific method of evaluating the accuracy of the power grid data according to the acquired power grid accuracy data to obtain the power grid data accuracy evaluation value is as follows: Obtaining a data signal-to-noise ratio deviation according to a relative relationship between the data signal-to-noise ratio and a preset data signal-to-noise ratio in a database, wherein the data signal-to-noise ratio deviation is obtained by performing a ratio operation on the data signal-to-noise ratio and the preset data signal-to-noise ratio in the database; Obtaining a sample balance coefficient deviation according to a relative relationship between the sample balance coefficient and a preset sample balance coefficient in a database, wherein the sample balance coefficient deviation is obtained by performing a difference operation on the sample balance coefficient and the preset sample balance coefficient in the database; The power grid factor, data missing rate, data signal-to-noise ratio deviation and sample balance coefficient deviation are processed to obtain a power grid data accuracy assessment value, wherein the power grid factor represents the degree of influence of the standardized score on the power grid data accuracy assessment value.

4. The power grid fault prediction method based on artificial intelligence as claimed in claim 1, characterized in that: The specific process of determining whether to perform grid data preprocessing based on the acquired grid data accuracy evaluation value is as follows: Determine whether the grid data accuracy assessment value is less than the preset grid data accuracy threshold: If the grid data accuracy assessment value is less than the preset grid data accuracy threshold, grid data preprocessing is performed; If the grid data accuracy assessment value is not less than the preset grid data accuracy threshold, grid data preprocessing is not performed.

5. The power grid fault prediction method based on artificial intelligence as claimed in claim 4, characterized in that: The specific process of the power grid data preprocessing is as follows; A1, determine whether there are outliers in the power grid data. If there are outliers in the power grid data, remove the outliers, otherwise execute A2; A2, determine whether the data missing rate in the power grid data is greater than 0. If the data missing rate in the power grid data is greater than 0, the missing value processing is performed, otherwise the data format is unified; The power grid data preprocessing includes removing outliers, processing missing values ​​and unifying data formats.

6. The power grid fault prediction method based on artificial intelligence as claimed in claim 2, characterized in that: The specific method of performing real-time evaluation of power grid data according to the extraction time and the acquired real-time data of the power grid to obtain the real-time evaluation value of the power grid data is as follows: Obtaining an extraction time deviation according to a relative relationship between the extraction time and a preset extraction time in a database, wherein the extraction time deviation is obtained by performing a ratio operation on the extraction time and the preset extraction time in the database; Obtaining a power grid data update frequency deviation according to a relative relationship between the power grid data update frequency and a preset power grid data update frequency in a database, wherein the power grid data update frequency deviation is obtained by performing a ratio operation on the power grid data update frequency and the preset power grid data update frequency in the database; Obtaining a power grid data amount deviation according to a relative relationship between the power grid data amount and a preset power grid data amount in a database, wherein the power grid data amount deviation is obtained by performing a ratio operation on the power grid data amount and the preset power grid data amount in the database; The optimized power grid data accuracy assessment value, waveform change factor, extraction time deviation, power grid data update frequency deviation and power grid data volume deviation are processed to obtain the power grid data real-time assessment value.

7. The power grid fault prediction method based on artificial intelligence according to claim 1, characterized in that: The specific method for determining whether to perform data real-time optimization based on the acquired power grid data real-time evaluation value is as follows: Determine whether the real-time evaluation value of the power grid data is less than the preset real-time threshold of the power grid data: If the real-time evaluation value of the power grid data is less than the preset real-time threshold of the power grid data, the real-time optimization of the data is performed; If the real-time evaluation value of the power grid data is not less than the preset real-time threshold of the power grid data, the data real-time optimization will not be performed; The data real-time optimization includes feature selection optimization and extraction time optimization; The extraction time optimization includes parallel computing, data compression and incremental learning.

8. The power grid fault prediction method based on artificial intelligence as claimed in claim 2, characterized in that: The specific method of performing grid fault assessment based on the acquired grid characteristic data to obtain a grid fault assessment value is as follows: Obtaining a voltage imbalance deviation according to a relative relationship between the voltage imbalance and a preset voltage imbalance in a database, wherein the voltage imbalance deviation is obtained by performing a ratio operation on the voltage imbalance and the preset voltage imbalance in the database; Obtaining a negative-sequence current component deviation according to a relative relationship between the negative-sequence current component and a preset negative-sequence current component in a database, wherein the negative-sequence current component deviation is obtained by performing a ratio operation on the negative-sequence current component and the preset negative-sequence current component in the database; Obtaining a zero-sequence current component deviation according to a relative relationship between the zero-sequence current component and a preset zero-sequence current component in a database, wherein the zero-sequence current component deviation is obtained by performing a ratio operation on the zero-sequence current component and the preset zero-sequence current component in the database; Obtaining a grid phase difference coefficient according to a relative relationship between a voltage-current phase difference and a preset voltage-current phase difference in a database; The voltage waveform distortion deviation is obtained according to the relative relationship between the voltage waveform distortion and the preset voltage waveform distortion in the database, the current waveform distortion deviation is obtained according to the relative relationship between the current waveform distortion and the preset current waveform distortion in the database, and the voltage waveform distortion deviation and the current waveform distortion deviation are processed to obtain the grid waveform distortion coefficient; The voltage unbalance deviation, negative sequence current component deviation, zero sequence current component deviation, grid phase difference coefficient and grid waveform distortion coefficient are processed to obtain a grid fault assessment value.

9. The power grid fault prediction method based on artificial intelligence according to claim 1, characterized in that: The specific process of determining whether to perform a fault warning based on the acquired power grid fault assessment value is as follows: Determine whether the power grid fault assessment value is less than the preset power grid fault threshold: If the grid fault assessment value is less than the preset grid fault threshold, no fault warning will be issued; If the grid fault assessment value is not less than the preset grid fault threshold, a fault warning is issued; The fault warning means identifying the type of power grid fault and sending warning information to a preset person.

10. The power grid fault prediction system based on artificial intelligence is characterized by: The artificial intelligence-based power grid fault prediction system includes: a power grid data accuracy assessment module, a power grid data real-time assessment module and a power grid fault assessment module; The power grid data accuracy assessment module is used to analyze the acquired power grid data to obtain power grid data accuracy data, to assess the accuracy of the power grid data based on the acquired power grid accuracy data to obtain a power grid data accuracy assessment value, and to determine whether to perform power grid data preprocessing based on the acquired power grid data accuracy assessment value, wherein the power grid data accuracy assessment value is used to quantitatively assess data quality; The power grid data real-time evaluation module is used for extracting the optimized power grid data to obtain power grid characteristic data and extraction time if data preprocessing is performed, and performing real-time evaluation of the power grid data according to the extraction time and the obtained power grid real-time data to obtain a power grid data real-time evaluation value, and judging whether to perform data real-time optimization based on the obtained power grid data real-time evaluation value, wherein the power grid data real-time evaluation value is used to quantitatively evaluate the real-time performance of the data; The power grid fault assessment module is used to perform power grid fault assessment based on the acquired power grid characteristic data to obtain a power grid fault assessment value if data real-time optimization is performed, and to determine whether to perform a fault warning based on the acquired power grid fault assessment value. The power grid fault assessment value is used to quantitatively assess the probability of occurrence of asymmetric power grid faults.

Images