Power grid resource dynamic threshold early warning method and system based on multi-source data

Abstract

The invention relates to the technical field of power grid resource management, in particular to a power grid resource dynamic threshold early warning method and system based on multi-source data. The method comprises the following steps: performing dynamic baseline value weighted combination on standard historical multi-source power grid load data to obtain power grid load dynamic baseline value data; carrying out adaptive load early warning threshold processing on the power grid load dynamic baseline value data to obtain dynamic load early warning threshold curve data; constructing a power grid load prediction model; performing short-term load prediction by using the power grid load prediction model to obtain short-term load prediction data; and performing threshold load difference calculation on the short-term load prediction data through the dynamic load early-warning threshold curve data, and performing multi-stage early-warning trigger processing to obtain a power grid multi-stage early-warning trigger strategy. According to the invention, by dynamically evaluating the load change of the power grid, real-time monitoring and self-adaptive early warning are realized, false alarms and missing alarms are effectively reduced, and the safety management level of the power grid is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid resource management, and in particular to a power grid resource dynamic threshold early warning method and system based on multi-source data. Background Art

[0002] The demand for electricity continues to grow, the scale of the power grid continues to expand, and the structure of the power grid is becoming increasingly complex. As a key infrastructure for national economic and social development, the safe and stable operation of the power grid is of vital importance. The operating environment of the power grid is complex and changeable, and various uncertain factors, such as natural disasters, equipment failures, and human errors, may threaten the safe and stable operation of the power grid. Therefore, how to effectively conduct risk assessment and early warning of power grid resources is of great practical significance for ensuring the safe and stable operation of the power grid and improving the reliability of power supply. As an important parameter of power grid operation, the change of power grid load directly affects the safety and economy of the power grid. The power grid load itself has significant dynamic change characteristics. The load size is affected by external factors such as seasons, weather, and holidays. The combined effect of these factors causes the power grid load to present complex time-varying characteristics. For example, the daily load curve shows "peak and valley" changes, the weekly load curve shows differences between working days and non-working days, and the seasonal load curve shows summer and winter peaks. Fixed thresholds cannot capture these dynamic changes and it is difficult to accurately reflect the true state of the load. When the load fluctuates within the normal range, if it exceeds the set fixed threshold, an alarm will be triggered, resulting in frequent false alarms, which not only increases the workload of maintenance personnel, but also may cause misjudgment of the power grid operation status and reduce the credibility of the early warning system. However, traditional power grid load monitoring and early warning methods often use fixed thresholds, that is, a fixed upper or lower load limit is set in advance. When the actual load exceeds these thresholds, the system issues an alarm, which is difficult to adapt to the load characteristics of different regions and different time periods, and is prone to excessive warnings or missed reports. Summary of the invention

[0003] Based on this, the present invention provides a power grid resource dynamic threshold early warning method and system based on multi-source data to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a power grid resource dynamic threshold early warning method based on multi-source data includes the following steps:

[0005] Step S1: Acquire standard historical multi-source power grid load data; classify the standard historical multi-source power grid load data by historical load monitoring date, and decompose the power grid load time-frequency component to generate power grid load time-frequency component data;

[0006] Step S2: setting a grid load rolling window according to the grid load time-frequency component data, and performing weighted combination of dynamic baseline values ​​to obtain grid load dynamic baseline value data;

[0007] Step S3: Screen the effective historical load data of the standard historical multi-source power grid load data, calculate the coefficient of variation, and generate the effective power grid historical load coefficient of variation; analyze the warning source sensitive parameters according to the effective power grid historical load coefficient of variation, and generate the optimal warning sensitivity parameters; perform adaptive load warning threshold processing on the power grid load dynamic baseline value data according to the effective power grid historical load coefficient of variation and the optimal warning sensitivity parameters, and obtain dynamic load warning threshold curve data;

[0008] Step S4: acquiring real-time power grid resource equipment status data; constructing a power grid monitoring resource topology structure based on the real-time power grid resource equipment status data to generate power grid monitoring resource topology data; constructing a power grid load prediction model based on the power grid monitoring resource topology data; performing short-term load prediction using the power grid load prediction model to obtain short-term load prediction data;

[0009] Step S5: Calculate the threshold load difference of the short-term load forecast data through the dynamic load warning threshold curve data, and perform multi-level warning trigger processing to obtain a multi-level warning trigger strategy for the power grid.

[0010] The present invention utilizes multi-source historical power grid load data, decomposes the time-frequency components thereof, and extracts the periodic law and random fluctuation components of load changes. This refined data processing method enables the method to more accurately capture the dynamic characteristics of load changes, avoiding the defect of insufficient understanding of load change laws in traditional methods. By classifying the monitoring date of historical load data, distinguishing the load patterns of different types of dates (such as working days, weekends, holidays, etc.), and constructing corresponding early warning models respectively, the accuracy of early warning is further improved, and the problem that a single model is difficult to adapt to the differences in load patterns on different dates is avoided. The dynamic baseline value is constructed by using a rolling window mechanism and a weighted combination algorithm, which can track the load change trend in real time and adaptively adjust the baseline level according to the changes in historical data. This effectively solves the problem that the fixed threshold method cannot adapt to the periodic fluctuations and long-term trend changes of loads. The introduction of the dynamic baseline value makes the early warning threshold no longer a static value, but a curve that can be dynamically adjusted as the load changes, so as to be more in line with the actual load changes and reduce the occurrence of false alarms and missed alarms. The early warning source sensitive parameter analysis is performed based on the coefficient of variation of the historical load data, and the optimal alarm sensitivity parameter is finally determined. According to the load fluctuation characteristics of different regions and different time periods, the sensitivity of the early warning is adaptively adjusted. For regions or time periods with large load fluctuations, the system will automatically increase the early warning sensitivity and discover potential risks in time; while for regions or time periods with small load fluctuations, the system will reduce the early warning sensitivity accordingly to avoid excessive early warnings and reduce unnecessary maintenance work. The prediction results are compared with the dynamic early warning threshold to achieve the prediction and early warning of future load changes. This prediction-early warning linkage mechanism can discover potential load risks in advance, provide sufficient response time for power grid dispatching and operation personnel, effectively avoid accidents, and ensure the safe and stable operation of the power grid. The application of multi-level early warning triggering strategy further improves the effectiveness and practicality of the early warning, and can issue early warning signals in a graded manner according to the degree of load exceeding the threshold, guide emergency response measures at different levels, and improve the refined management level of power grid operation. Therefore, a dynamic threshold early warning method for power grid resources based on multi-source data of the present invention comprehensively analyzes historical load data and real-time power grid resource equipment status, uses time-frequency component decomposition technology to capture the dynamic changes of power grid load, combines rolling windows and weighted combination algorithms to construct dynamic baseline values, and realizes refined monitoring and adaptive early warning of power grid resource load changes.

[0011] Preferably, step S1 comprises the following steps:

[0012] Step S11: Acquire historical multi-source power grid resource load data;

[0013] Step S12: unifying the multi-source data source formats of historical multi-source power grid resource load data, and performing data preprocessing to generate standard historical multi-source power grid load data;

[0014] Step S13: classifying the historical load monitoring dates according to the standard historical multi-source power grid load data to obtain classified historical multi-source power grid load data;

[0015] Step S14: performing fast Fourier transform processing on the classified historical multi-source power grid load data to generate historical power grid load spectrum data;

[0016] Step S15: Decomposing the time-frequency components of the power grid load according to the historical power grid load spectrum data to generate the time-frequency component data of the power grid load.

[0017] The present invention classifies the historical load monitoring dates. The method can distinguish the load patterns of different types of dates (such as weekdays, weekends, and holidays), and analyze and model different types of dates respectively. This effectively solves the problem that traditional methods are difficult to distinguish the differences in load patterns on different dates, and improves the pertinence and accuracy of early warning. For example, the load curve on weekdays usually shows obvious "peak and valley" characteristics, while the load curve on weekends and holidays is relatively stable. Through date classification, a more accurate load forecasting model can be constructed for different date types, thereby improving the accuracy of early warning. Time-frequency component decomposition further decomposes the load data into components of different frequencies, thereby more finely depicting the dynamic characteristics of load changes and more accurately capturing the laws of load changes. By extracting the periodic components of load changes, future load change trends can be predicted more accurately; and by analyzing the random fluctuation components, the uncertainty of load changes can be better evaluated, thereby improving the reliability of early warnings.

[0018] Preferably, step S15 comprises the following steps:

[0019] Step S151: performing a main load frequency range analysis based on historical power grid load spectrum data to obtain historical load frequency range data;

[0020] Step S152: Perform maximum wavelet decomposition layer number processing according to historical load frequency range data to obtain the maximum wavelet decomposition layer number;

[0021] Step S153: dividing the decomposition level frequency range according to the maximum wavelet decomposition level number, and performing decomposition level mapping on the historical load frequency range data to obtain load frequency-decomposition level mapping data;

[0022] Step S154: determining the decomposition layer coverage range of the load frequency-decomposition layer mapping data, and performing wavelet decomposition layer optimization to generate optimal decomposition layer data;

[0023] Step S155: performing wavelet transform decomposition on the historical power grid load spectrum data based on the optimal decomposition layer number data to generate wavelet decomposition coefficients;

[0024] Step S156: extracting wavelet approximate components according to the wavelet decomposition coefficients to obtain long-term trend data of power grid load;

[0025] Step S157: extracting wavelet detail components according to the wavelet decomposition coefficients, and obtaining power grid load daily fluctuation data, power grid load weekly fluctuation data, and power grid load seasonal fluctuation data;

[0026] Step S158: Combine the long-term trend data of power grid load, the daily fluctuation data of power grid load, the weekly fluctuation data of power grid load and the seasonal fluctuation data of power grid load into the time-frequency component of power grid resource monitoring load to obtain the time-frequency component data of power grid load.

[0027] The present invention combines the division of the frequency range of the decomposition level and the determination of the coverage range of the decomposition layer number to intelligently determine the optimal wavelet decomposition layer number, which ensures that the wavelet decomposition can better capture the load change characteristics on different time scales. Decompose the historical power grid load data into fluctuation components of different time scales, such as long-term trends, intraday fluctuations, intraweek fluctuations, and seasonal fluctuations. This multi-scale decomposition method can more comprehensively characterize the dynamic characteristics of load changes, avoiding the limitation of traditional methods that only focus on single-scale fluctuation information. The various extracted time-frequency component data are combined to construct complete power grid load time-frequency component data. This multi-level and multi-angle load data analysis can more accurately reflect the actual situation of load changes, thereby improving the accuracy and reliability of early warnings.

[0028] Preferably, step S2 comprises the following steps:

[0029] Step S21: setting the grid load rolling window according to the grid load time-frequency component data, and obtaining the annual trend baseline rolling window, the daily fluctuation baseline rolling window, the weekly fluctuation baseline rolling window and the seasonal fluctuation baseline rolling window respectively;

[0030] Step S22: Calculate the long-term trend baseline of the power grid load long-term trend data through the annual trend baseline rolling window to generate long-term trend baseline data;

[0031] Step S23: calculating the intraday fluctuation baseline of the power grid load intraday fluctuation data through the intraday fluctuation baseline rolling window to generate the intraday fluctuation baseline data;

[0032] Step S24: Calculate the weekly fluctuation baseline of the power grid load weekly fluctuation data through the weekly fluctuation baseline rolling window to generate weekly fluctuation baseline data;

[0033] Step S25: performing seasonal fluctuation baseline calculation on the seasonal fluctuation data of the power grid load through the seasonal fluctuation baseline rolling window to generate seasonal fluctuation baseline data;

[0034] Step S26: Perform a weighted combination of dynamic baseline values ​​based on the long-term trend baseline data, the intraday fluctuation baseline data, the weekly fluctuation baseline data, and the seasonal fluctuation baseline data to obtain the dynamic baseline value data of the power grid load.

[0035] The present invention sets the annual trend baseline rolling window, the daily fluctuation baseline rolling window, the weekly fluctuation baseline rolling window and the seasonal fluctuation baseline rolling window respectively, and combines the load data of different time scales to calculate the long-term trend baseline, the daily fluctuation baseline, the weekly fluctuation baseline and the seasonal fluctuation baseline respectively, so as to accurately capture the load change trend of different time scales. This divide-and-conquer strategy can more effectively handle the complexity of load changes at different time scales, and avoids the problem that a single baseline is difficult to adapt to fluctuations of multiple time scales at the same time. By weighted combination of long-term trend baseline data, daily fluctuation baseline data, weekly fluctuation baseline data and seasonal fluctuation baseline data, the method constructs the final dynamic baseline value data of power grid load. This weighted combination method can comprehensively consider the impact of load changes at different time scales, and assign different weights according to the importance of fluctuation characteristics at different time scales, so as to make the dynamic baseline value more representative and reliable. For example, during the peak summer period, the weight of the seasonal fluctuation baseline will be higher, while in other time periods, the weight of the daily fluctuation baseline will be higher.

[0036] Preferably, step S3 comprises the following steps:

[0037] Step S31: Obtain real-time power grid monitoring timestamp;

[0038] Step S32: Screening effective historical load data of standard historical multi-source power grid load data through real-time power grid monitoring timestamps, and calculating coefficient of variation to generate effective power grid historical load coefficient of variation;

[0039] Step S33: Analyze the warning source sensitive parameters according to the effective power grid historical load variation coefficient to generate the optimal warning sensitivity parameters;

[0040] Step S34: Calculate the load warning threshold value for the dynamic baseline value data of the power grid load according to the effective power grid historical load variation coefficient and the optimal alarm sensitivity parameter, and perform adaptive threshold constraints to obtain dynamic load warning threshold curve data.

[0041] The present invention obtains the real-time power grid monitoring timestamp and uses it to screen the effective historical load data of the standard historical multi-source power grid load data. The method ensures that the historical data used to calculate the coefficient of variation is relevant to the load characteristics at the current time point. This avoids the warning deviation caused by the use of outdated or irrelevant data, and improves the accuracy and timeliness of the warning. The coefficient of variation is calculated based on the screened effective historical load data, and the degree of load fluctuation can be quantified more accurately. The coefficient of variation reflects the degree of discreteness of the load data and can be used to measure the stability of load changes. A larger coefficient of variation indicates that the load fluctuation is large and a higher warning sensitivity is required; while a smaller coefficient of variation indicates that the load fluctuation is small and the warning sensitivity can be appropriately reduced. By analyzing the warning source sensitive parameters of the effective power grid historical load coefficient of variation and generating the optimal alarm sensitivity parameter, the method realizes the adaptive adjustment of the warning sensitivity. This effectively solves the problem that the traditional fixed threshold warning method cannot adapt to the differences in load fluctuation characteristics in different regions and time periods, and avoids the occurrence of false alarms and missed alarms.

[0042] Preferably, step S32 includes the following steps:

[0043] Step S321: Select a historical effective load window for the real-time power grid monitoring timestamp to obtain effective historical load window data;

[0044] Step S322: Screening the standard historical multi-source power grid load data through the effective historical load window data to obtain the effective historical load data of the power grid;

[0045] Step S323: Calculate the historical load mean value according to the effective historical load data of the power grid to generate effective historical load mean data;

[0046] Step S324: Calculate the historical load standard deviation based on the effective historical load data of the power grid to generate effective historical standard deviation data;

[0047] Step S325: Calculate the coefficient of variation based on the effective historical load mean data and the effective historical standard deviation data to generate the effective power grid historical load coefficient of variation.

[0048] The present invention selects a historical effective load window for real-time power grid monitoring timestamps, and the method can intelligently determine the range of historical data used to calculate the coefficient of variation. This ensures that the selected historical data is highly correlated with the load characteristics at the current time point, avoids the interference of outdated or irrelevant data on the calculation results, and thus improves the accuracy and representativeness of the coefficient of variation. The effective historical load window data is used to screen the standard historical multi-source power grid load data, which ensures that the data used to calculate the coefficient of variation comes from a historical period with similar load characteristics to the current time point, thereby more accurately reflecting the fluctuation of the current load. After screening out the effective historical load data, the method calculates the historical load mean and the historical load standard deviation respectively. The mean reflects the average level of the historical load, and the standard deviation reflects the degree of dispersion of the historical load.

[0049] Preferably, step S33 includes the following steps:

[0050] Step S331: performing a distribution characteristic analysis of the coefficient of variation of the effective power grid historical load, and setting an initial sensitivity parameter to obtain initial sensitivity parameter range data;

[0051] Step S332: Acquire power grid resource load warning source configuration data;

[0052] Step S333: connect the warning source according to the power grid resource load warning source configuration data, and read the historical warning record to obtain the original power grid warning record data;

[0053] Step S334: screening the real alarm records according to the original power grid early warning record data to obtain the real power grid alarm record data;

[0054] Step S335: Screening the false alarm records according to the original power grid early warning record data to obtain the power grid false alarm record data;

[0055] Step S336: performing sensitivity parameter cyclic traversal on the initial sensitivity parameter range data using the preset sensitivity traversal step data to obtain traversal sensitivity parameters;

[0056] Step S337: calculating the alarm recovery rate of the real alarm record data of the power grid and the false alarm record data of the power grid by traversing the sensitivity parameters, and obtaining the alarm recovery rate data of the power grid;

[0057] Step S338: Calculate the alarm precision rate of the real alarm record data of the power grid and the false alarm record data of the power grid by traversing the sensitivity parameters to obtain the alarm precision rate data of the power grid;

[0058] Step S339: Calculate the warning performance index of the ergodic sensitivity parameter according to the power grid alarm precision rate data and the power grid alarm recall rate data, and select the optimal sensitivity parameter to obtain the optimal alarm sensitivity parameter.

[0059] The present invention obtains the real alarm record data of the power grid and the false alarm record data of the power grid respectively by screening the original power grid early warning record data, and uses the preset sensitivity traversal step data to traverse the initial sensitivity parameter range data in a loop, and combines the real alarm record data of the power grid and the false alarm record data of the power grid to calculate the alarm recall rate and the alarm precision rate respectively, so as to comprehensively evaluate the performance of the early warning system under different sensitivity parameters. The recall rate reflects the proportion of real alarms that the early warning system can correctly identify, and the precision rate reflects the proportion of real alarms in the alarms issued by the early warning system. These two indicators together constitute the key indicators for evaluating the performance of the early warning system. According to the power grid alarm precision rate data and the power grid alarm recall rate data, the traversal sensitivity parameter is calculated for the early warning performance index, and the optimal sensitivity parameter is selected. The two indicators of recall rate and precision rate are comprehensively considered, and the sensitivity parameter that achieves the best balance between the two is selected, thereby realizing the optimization of early warning performance.

[0060] Preferably, step S4 comprises the following steps:

[0061] Step S41: Acquire real-time power grid resource equipment status data;

[0062] Step S42: constructing a power grid monitoring resource topology structure according to the real-time power grid resource device status data to generate power grid monitoring resource topology data;

[0063] Step S43: collecting real-time loads of power grid monitoring nodes according to the power grid monitoring resource topology data to obtain real-time load data of power grid resources;

[0064] Step S44: performing Bayesian network node processing according to the power grid monitoring resource topology data, and constructing a power grid load prediction model through a preset convolutional neural network model, thereby obtaining a power grid load prediction model to be trained;

[0065] Step S45: performing model training set processing on the standard historical multi-source power grid load data, and performing model training on the power grid load prediction model to be trained to obtain a power grid load prediction model;

[0066] Step S46: Use the power grid load forecasting model to perform short-term load forecasting on the real-time load data of power grid resources to obtain short-term load forecasting data.

[0067] The present invention collects real-time loads of power grid monitoring nodes based on the topological data of power grid monitoring resources, obtains the latest load data, and provides real-time data input for short-term load forecasting. This ensures the timeliness of the prediction results and can reflect the load changes of the current power grid in a more timely manner. A power grid load prediction model is constructed using Bayesian network node processing and a preset convolutional neural network model, and model training is performed using standard historical multi-source power grid load data, so that the model can learn the changing laws of historical load data and apply it to future load forecasting. Bayesian networks can effectively handle uncertainty and complex relationships, while convolutional neural networks are good at capturing spatial and temporal features in time series data. The combination of the two can improve the accuracy of load forecasting. The trained power grid load prediction model is used to perform short-term load forecasting on the real-time load data of power grid resources to obtain predicted short-term load data.

[0068] Preferably, step S5 comprises the following steps:

[0069] Step S51: extracting the forecast time point of the short-term load forecast data to obtain the forecast time point data of the power grid resources;

[0070] Step S52: using the power grid resource prediction time point data to identify the real-time load warning threshold interval of the dynamic load warning threshold curve data, and performing threshold load difference calculation on the short-term load prediction data to obtain predicted load difference data;

[0071] Step S53: Calculate the load over-limit range according to the predicted load difference data to obtain the power grid predicted load over-limit range data;

[0072] Step S54: divide the warning levels according to the power grid predicted load over-limit amplitude data, and perform multi-level warning trigger processing through a preset warning level execution strategy to obtain a power grid multi-level warning trigger strategy.

[0073] The present invention can accurately determine the warning threshold corresponding to each prediction time point by extracting the prediction time point of short-term load prediction data and identifying the real-time load warning threshold interval in combination with the dynamic load warning threshold curve data. The dynamic threshold curve can automatically adjust the threshold according to the load characteristics of different time periods, so as to more effectively identify abnormal load changes. By calculating the threshold load difference of the short-term load prediction data, the predicted load difference data is obtained, which reflects the gap between the predicted load and the warning threshold. The load overlimit amplitude is calculated according to the predicted load difference data, and the degree of load overlimit can be quantified more finely. The larger the overlimit amplitude, the higher the risk of load overlimit, and more urgent countermeasures need to be taken. According to the different degrees of load overlimit, different levels of warnings are triggered, and corresponding warning strategies are executed, thereby improving the effectiveness and practicality of the warning. For example, for a slight load overlimit, the system will only issue a reminder message; for a serious load overlimit, the system will trigger a higher level of warning and take corresponding control measures to ensure the safe and stable operation of the power grid.

[0074] Preferably, the present invention further provides a power grid resource dynamic threshold warning system based on multi-source data, which executes the power grid resource dynamic threshold warning method based on multi-source data as described above, and the power grid resource dynamic threshold warning system based on multi-source data includes:

[0075] The historical load time-frequency decomposition module is used to obtain standard historical multi-source power grid load data; classify the standard historical multi-source power grid load data by historical load monitoring date, and decompose the power grid load time-frequency components to generate power grid load time-frequency component data;

[0076] A load dynamic baseline construction module is used to set a grid load rolling window according to the grid load time-frequency component data, and to perform weighted combination of dynamic baseline values ​​to obtain grid load dynamic baseline value data;

[0077] The adaptive threshold generation module is used to screen the effective historical load data of the standard historical multi-source power grid load data, calculate the coefficient of variation, and generate the effective power grid historical load coefficient of variation; analyze the warning source sensitive parameters according to the effective power grid historical load coefficient of variation, and generate the optimal warning sensitivity parameters; perform adaptive load warning threshold processing on the dynamic baseline value data of the power grid load according to the effective power grid historical load coefficient of variation and the optimal warning sensitivity parameters, and obtain the dynamic load warning threshold curve data;

[0078] The short-term load forecasting module is used to obtain real-time power grid resource equipment status data; construct the power grid monitoring resource topology structure according to the real-time power grid resource equipment status data to generate power grid monitoring resource topology data; construct a power grid load forecasting model according to the power grid monitoring resource topology data; use the power grid load forecasting model to perform short-term load forecasting and obtain short-term load forecasting data;

[0079] The multi-level warning trigger module is used to calculate the threshold load difference of short-term load forecast data through dynamic load warning threshold curve data, and perform multi-level warning trigger processing to obtain a multi-level warning trigger strategy for the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 A schematic diagram of the steps of the power grid resource dynamic threshold early warning method based on multi-source data of the present invention;

[0081] Figure 2 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0082] Figure 3 for Figure 1 Detailed implementation steps of step S5;

[0083] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0084] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0085] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0086] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0087] To achieve this, please refer to Figures 1 to 3 The present invention provides a power grid resource dynamic threshold early warning method based on multi-source data, comprising the following steps:

[0088] Step S1: Acquire standard historical multi-source power grid load data; classify the standard historical multi-source power grid load data by historical load monitoring date, and decompose the power grid load time-frequency component to generate power grid load time-frequency component data;

[0089] Step S2: setting a grid load rolling window according to the grid load time-frequency component data, and performing weighted combination of dynamic baseline values ​​to obtain grid load dynamic baseline value data;

[0090] Step S3: Screen the effective historical load data of the standard historical multi-source power grid load data, calculate the coefficient of variation, and generate the effective power grid historical load coefficient of variation; analyze the warning source sensitive parameters according to the effective power grid historical load coefficient of variation, and generate the optimal warning sensitivity parameters; perform adaptive load warning threshold processing on the power grid load dynamic baseline value data according to the effective power grid historical load coefficient of variation and the optimal warning sensitivity parameters, and obtain dynamic load warning threshold curve data;

[0091] Step S4: acquiring real-time power grid resource equipment status data; constructing a power grid monitoring resource topology structure according to the real-time power grid resource equipment status data to generate power grid monitoring resource topology data; constructing a power grid load prediction model according to the power grid monitoring resource topology data; performing short-term load prediction using the power grid load prediction model to obtain short-term load prediction data;

[0092] Step S5: Calculate the threshold load difference of the short-term load forecast data through the dynamic load warning threshold curve data, and perform multi-level warning trigger processing to obtain a multi-level warning trigger strategy for the power grid.

[0093] In the embodiment of the present invention, reference Figure 1 As shown, it is a schematic diagram of the steps of the power grid resource dynamic threshold early warning method based on multi-source data of the present invention. In this embodiment, the power grid resource dynamic threshold early warning method based on multi-source data includes the following steps:

[0094] Step S1: Acquire standard historical multi-source power grid load data; classify the standard historical multi-source power grid load data by historical load monitoring date, and decompose the power grid load time-frequency component to generate power grid load time-frequency component data;

[0095] In an embodiment of the present invention, historical load data is obtained from multiple sources including a power grid dispatching and control center, an intelligent substation, a distribution automation system, and an advanced measurement system (AMI). These data include power load information of different voltage levels and different regions, and the time granularity can be 15 minutes, 30 minutes, or 1 hour. The data format is unified as a tabular data consisting of fields such as timestamp and load value. The historical load data is classified according to the date attribute, for example, it can be divided according to categories such as working days, weekends, holidays, etc., and seasonal factors can also be considered, such as summer, winter, etc. The classified historical load data is decomposed into time-frequency components using methods such as wavelet transform or empirical mode decomposition (EMD). The original load data is decomposed into fluctuation components of different frequencies, for example, it can be decomposed into daily cycle components, weekly cycle components, and random fluctuation components. The decomposed time-frequency component data can more clearly reflect the law of load changes, such as baseline components, daily cycle components, weekly cycle components, and seasonal fluctuation components. For example, using db4 wavelet to perform 4-layer decomposition, we can obtain the wavelet coefficients of each layer, and extract the approximate component and detail component, which represent the long-term trend, intraday fluctuation, weekly fluctuation and seasonal fluctuation respectively.

[0096] Step S2: setting a grid load rolling window according to the grid load time-frequency component data, and performing weighted combination of dynamic baseline values ​​to obtain grid load dynamic baseline value data;

[0097] In the embodiment of the present invention, different rolling window sizes are set for different time-frequency components (long-term trend, intraday fluctuation, intraweek fluctuation, seasonal fluctuation) decomposed in step S1. For example, a rolling window of 365 days is used for long-term trend, a rolling window of 24 hours is used for intraday fluctuation, a rolling window of 168 hours is used for intraweek fluctuation, and a rolling window of 90 days is used for seasonal fluctuation. Then, in the corresponding rolling window, the baseline value of each component is calculated. For example, for the long-term trend component, the load average value of the corresponding date of the past 365 days is calculated as the baseline value; for the intraday fluctuation component, the load average value at the same time in the past 24 hours is calculated as the baseline value. Finally, the baseline values ​​of each component are weighted and combined to obtain the final dynamic baseline value. The weight coefficient can be set according to the contribution of different components to the load change, for example, the long-term trend weight is 0.5, the intraday fluctuation weight is 0.2, the weekly fluctuation weight is 0.2, and the seasonal fluctuation weight is 0.1.

[0098] Step S3: Screen the effective historical load data of the standard historical multi-source power grid load data, calculate the coefficient of variation, and generate the effective power grid historical load coefficient of variation; analyze the warning source sensitive parameters according to the effective power grid historical load coefficient of variation, and generate the optimal warning sensitivity parameters; perform adaptive load warning threshold processing on the power grid load dynamic baseline value data according to the effective power grid historical load coefficient of variation and the optimal warning sensitivity parameters, and obtain dynamic load warning threshold curve data;

[0099] In an embodiment of the present invention, the current real-time power grid monitoring timestamp is obtained. According to the timestamp, the historical load data of the same date and time in the past several years (for example, 5 years) are selected as the effective historical load data. The mean and standard deviation of the effective historical load data are calculated, and then the coefficient of variation (standard deviation / mean) is calculated. Next, the distribution characteristics of the coefficient of variation are analyzed, the initial sensitivity parameter range is set, for example, [1.5, 3.0], and the traversal step size is set, for example, 0.1. The warning source database is connected, the historical warning records are read, and the real alarm and false alarm records are screened out. Using different sensitivity parameters, the recall rate and precision rate are calculated, and performance indicators such as the F1 value are calculated. The sensitivity parameter with the highest F1 value is selected as the optimal alarm sensitivity parameter. Finally, according to the optimal sensitivity parameter and the coefficient of variation, the upper and lower limits of the dynamic load warning threshold curve are calculated. Upper limit = dynamic baseline value + (sensitivity parameter × coefficient of variation), lower limit = dynamic baseline value - (sensitivity parameter × coefficient of variation).

[0100] Step S4: acquiring real-time power grid resource equipment status data; constructing a power grid monitoring resource topology structure according to the real-time power grid resource equipment status data to generate power grid monitoring resource topology data; constructing a power grid load prediction model according to the power grid monitoring resource topology data; performing short-term load prediction using the power grid load prediction model to obtain short-term load prediction data;

[0101] In an embodiment of the present invention, power grid resource equipment status data is collected in real time, for example, the voltage, current, power and other data of the substation are obtained from the power grid dispatching and control center system. The topological structure of the power grid monitoring resources is constructed based on the real-time data, for example, a graph data structure is used to represent the connection relationship between the substation, the line and the load. Based on the topological structure, a load forecasting model is constructed, for example, a graph convolutional neural network (GCN) is used in combination with a Bayesian network to integrate the topological structure information into the model. The model is trained using historical load data, weather data, etc. as training data. After the training is completed, the real-time load data is used as input to perform short-term load forecasting, for example, predicting the load value for the next 24 hours.

[0102] Step S5: Calculate the threshold load difference of the short-term load forecast data through the dynamic load warning threshold curve data, and perform multi-level warning trigger processing to obtain a multi-level warning trigger strategy for the power grid.

[0103] In the embodiment of the present invention, the short-term load forecast data obtained in step S4 is compared with the dynamic load warning threshold curve data obtained in step S3 to calculate the threshold load difference. For example, the predicted hourly load value for the next 24 hours is compared with the upper and lower limits of the corresponding dynamic warning threshold curve to calculate the difference. Secondly, according to the size of the threshold load difference, multi-level warning trigger processing is performed. For example, three warning levels can be set: first-level warning, second-level warning and third-level warning. When the predicted load exceeds the upper limit of the warning threshold curve by a certain amplitude, the first-level warning is triggered; when the predicted load exceeds the upper limit of the warning threshold curve but does not reach the first-level warning triggering condition, the second-level warning is triggered; when the predicted load is lower than the lower limit of the warning threshold curve by a certain amplitude, the third-level warning is triggered. For example, when the predicted load exceeds the upper limit by 10%, the first-level warning is triggered; when the predicted load exceeds the upper limit by 5% but is lower than 10%, the second-level warning is triggered; when the predicted load is lower than the lower limit by 5%, the third-level warning is triggered. For example, after the first-level warning is triggered, the emergency plan is activated and power-limiting measures are taken; after the second-level warning is triggered, the relevant departments are notified to be prepared; after the third-level warning is triggered, the load changes are closely monitored.

[0104] Preferably, step S1 comprises the following steps:

[0105] Step S11: Acquire historical multi-source power grid resource load data;

[0106] Step S12: unifying the multi-source data source formats of historical multi-source power grid resource load data, and performing data preprocessing to generate standard historical multi-source power grid load data;

[0107] Step S13: classifying the historical load monitoring dates according to the standard historical multi-source power grid load data to obtain classified historical multi-source power grid load data;

[0108] Step S14: performing fast Fourier transform processing on the classified historical multi-source power grid load data to generate historical power grid load spectrum data;

[0109] Step S15: Decomposing the time-frequency components of the power grid load according to the historical power grid load spectrum data to generate the time-frequency component data of the power grid load.

[0110] In an embodiment of the present invention, historical power grid resource load data is obtained from different data sources. These data sources may include a power grid dispatching control center database, an intelligent substation data acquisition system, a distribution automation system, and an advanced measurement system (AMI). The data types obtained include power load information of different voltage levels and different regions, for example, active power and reactive power of 110kV substations, 35kV substations, and 10kV feeders. The time granularity of the data may be 15 minutes, 30 minutes, or 1 hour. For example, the hourly active power and reactive power data covering all 110kV substations, 35kV substations, and some 10kV feeders in a city in the past two years are obtained. All data are converted into a unified CSV file format, including fields such as timestamp, substation / feeder ID, voltage level, active power, and reactive power. After the format is unified, the data is preprocessed, including missing value processing and outlier processing. For missing values, linear interpolation, spline interpolation, or mean filling can be used to fill them. For outliers, the Laida criterion or the 3σ criterion can be used to identify and eliminate them, or replace them with reasonable values. Classify the standard historical multi-source power grid load data according to the date attribute. The classification criteria may include working days, weekends, holidays, and seasons. For example, the dates of the year are divided into working days, weekends, and statutory holidays such as the Spring Festival and National Day. The data can also be divided into spring, summer, autumn, and winter according to the month or temperature. The classified historical multi-source power grid load data obtained in step S13 is processed by fast Fourier transform (FFT). FFT can convert time domain signals into frequency domain signals, thereby analyzing the amplitude and phase information of different frequency components in the load data. For example, the classified working day load data is subjected to FFT processing to obtain the amplitudes of different frequency components, for example, the amplitudes of frequency components such as daily cycles (24 hours) and weekly cycles (168 hours) can be obtained. Time-frequency component decomposition can decompose the original load data into fluctuation components of different frequencies, for example, it can be decomposed into baseline components, daily cycle components, weekly cycle components, and random fluctuation components. For example, the workday load data is decomposed into a baseline component, a daily cycle component, and a random fluctuation component using wavelet transform.

[0111] Preferably, step S15 comprises the following steps:

[0112] Step S151: performing a main load frequency range analysis based on historical power grid load spectrum data to obtain historical load frequency range data;

[0113] Step S152: Perform maximum wavelet decomposition layer number processing according to historical load frequency range data to obtain the maximum wavelet decomposition layer number;

[0114] Step S153: dividing the decomposition level frequency range according to the maximum wavelet decomposition level number, and performing decomposition level mapping on the historical load frequency range data to obtain load frequency-decomposition level mapping data;

[0115] Step S154: determining the decomposition layer coverage range of the load frequency-decomposition layer mapping data, and performing wavelet decomposition layer optimization to generate optimal decomposition layer data;

[0116] Step S155: performing wavelet transform decomposition on the historical power grid load spectrum data based on the optimal decomposition layer number data to generate wavelet decomposition coefficients;

[0117] Step S156: extracting wavelet approximate components according to the wavelet decomposition coefficients to obtain long-term trend data of power grid load;

[0118] Step S157: extracting wavelet detail components according to the wavelet decomposition coefficients, and obtaining power grid load daily fluctuation data, power grid load weekly fluctuation data, and power grid load seasonal fluctuation data;

[0119] Step S158: Combine the long-term trend data of power grid load, the daily fluctuation data of power grid load, the weekly fluctuation data of power grid load and the seasonal fluctuation data of power grid load into the time-frequency component of power grid resource monitoring load to obtain the time-frequency component data of power grid load.

[0120] In the embodiment of the present invention, the spectrum data corresponding to each date type is analyzed. For example, for the spectrum data of working days, its frequency distribution characteristics can be observed by drawing a spectrum graph. In the spectrum graph, the horizontal axis represents the frequency, and the vertical axis represents the energy or amplitude of the corresponding frequency. By observing the spectrum graph, the frequency bands where the energy is concentrated can be found. These frequency bands where the energy is concentrated usually correspond to the main frequency components of the load change. Assume that the analysis finds that the energy of the workday load is mainly concentrated in three frequency bands: the first frequency band is the low frequency band, corresponding to a cycle of about 24 hours (i.e., daily load changes); the second frequency band is the medium frequency band, corresponding to a cycle of about 6-8 hours (i.e., peak power consumption); the third frequency band is the high frequency band, corresponding to a cycle of less than 1 hour (i.e., some sudden power consumption events). Then, the frequency ranges of these main frequency bands are recorded, for example, the frequency range of the low frequency band is 0-0.0417Hz (corresponding to a 24-hour cycle), the frequency range of the medium frequency band is 0.125-0.1667Hz (corresponding to a 6-8 hour cycle), and the frequency range of the high frequency band is above 1Hz. A similar analysis is performed on the spectrum data of weekends and holidays to obtain their respective main frequency ranges. The lowest frequency value in the historical load frequency range data is analyzed. This lowest frequency usually corresponds to the long-term trend of the load signal, for example, the daily load cycle. In step S151, the lowest frequency on weekdays is 0.0417 Hz (corresponding to 24 hours), and the lowest frequency on weekends is 0.035 Hz. Then, a suitable wavelet basis function is selected, for example, Daubechies wavelet (dbN) can be used. Each layer of the wavelet transform will halve the frequency range, so it is necessary to select a suitable decomposition layer number so that the lowest frequency after decomposition can cover the lowest frequency in the historical load frequency range data. The following formula can be used to calculate the maximum number of decomposition layers: number of layers = ceil (log2 (Fs / f_min)). Among them, Fs represents the sampling frequency, f_min represents the lowest frequency value, and the ceil function represents rounding up. The frequency range is divided according to the maximum number of wavelet decomposition layers. Assuming that the maximum number of wavelet decomposition layers is 6, each layer of wavelet decomposition will halve the frequency range. Starting from the maximum frequency (half of the sampling frequency, i.e., the Nyquist frequency), decompose downward layer by layer. For a sampling frequency of once every 15 minutes (0.0011 Hz), the corresponding Nyquist frequency is 0.00055 Hz. Therefore, the frequency range of the first layer can be set to 0.00055 Hz / 2 to 0.00055 Hz, the frequency range of the second layer is 0.00055 Hz / 4 to 0.00055 Hz / 2, and so on. The frequency range of the sixth layer is 0.00055 Hz / 64 to 0.00055 Hz / 32. Then, the historical load frequency range data obtained in step S151 needs to be mapped to these decomposition levels.For each historical load frequency range (for example, the low frequency band of 0-0.0417Hz, the medium frequency band of 0.125-0.1667Hz, and the high frequency band of 1Hz or above on weekdays), it is necessary to determine which decomposition level or levels of frequency range it falls within. Analyze the load frequency-decomposition level mapping data obtained in step S153 to determine the number of layers that actually need to be decomposed by wavelet. In order to avoid the computational burden and noise interference caused by excessive decomposition, the number of decomposition layers needs to be optimized. The minimum number of decomposition layers covering the main load frequency range can be selected as the optimal number of decomposition layers. For example, if the main load frequency range is mapped to the 5th to 8th layers, 4 can be selected as the optimal number of decomposition layers, and only decomposition to the 4th layer can cover the target frequency range. Select a suitable wavelet basis function, for example, Daubechies wavelet (dbN) can be used, where N can be selected according to actual conditions. Then, for each type of load data on a date, such as weekdays, weekends, and holidays, wavelet transform decomposition is performed separately. For example, for the load data of working days, first determine the number of layers to be decomposed according to the optimal decomposition layer number data obtained in step S154, assuming that it is the 1st layer, the 3rd layer and the 5th layer. Then, use the selected wavelet basis function to perform wavelet decomposition on the load data. After wavelet decomposition, two parts of coefficients will be generated, one part is the approximate coefficient (cA), and the other part is the detail coefficient (cD). The decomposition of the 1st layer will obtain cA1 and cD1, the decomposition of the 3rd layer will be carried out on the basis of cA1 obtained in the 1st layer to obtain cA3 and cD3, and the decomposition of the 5th layer will be carried out on the basis of cA3 obtained in the 3rd layer to obtain cA5 and cD5. According to the decomposition layer number determination logic of step S154, only cD1, cD3 and cD5 are retained as decomposition coefficients of different frequency bands. For each date type (for example, working days, weekends, holidays), it is necessary to extract the approximate component from its wavelet decomposition coefficients. In step S155, assuming that the decomposition uses the 1st, 3rd and 5th layers, cA1, cA3 and cA5, as well as the corresponding detail coefficients cD1, cD3 and cD5 will be obtained. Among them, cA5 is the approximate coefficient obtained after 5 layers of decomposition, which contains lower frequency information and can be considered as the long-term trend of the load. Therefore, cA5 needs to be extracted as the long-term trend data of the load. For different load types, due to the different levels of decomposition, it is necessary to use the corresponding number of decomposition layers and approximate coefficients for reconstruction to obtain the corresponding wavelet approximate components. Wavelet reconstruction can be achieved by performing an inverse wavelet transform on the approximate coefficients, converting the frequency domain signal back to the time domain signal to obtain the actual load value. According to different frequency ranges, the corresponding detail coefficients are extracted. In step S155, assuming that the 1st, 3rd and 5th layers are used for wavelet decomposition, cD1, cD3 and cD5 are obtained.Among them, cD1 reflects the highest frequency fluctuation, such as the impact of some emergencies, and can therefore be considered to correspond to intraday fluctuations; cD3 reflects the medium frequency fluctuation, such as the load difference between weekdays and weekends, and can therefore be considered to correspond to intraweek fluctuations; cD5 reflects the low frequency fluctuation, for example, the load changes in different seasons, and can therefore be considered to correspond to seasonal fluctuations. The various component data extracted in steps S156 and S157 are combined to form the final grid load time-frequency component data. These component data include long-term trend data, intraday fluctuation data, intraweek fluctuation data, and seasonal fluctuation data. These component data can be stored in a data structure, such as a multidimensional array or a data frame containing multiple time series. For example, long-term trend data, intraday fluctuation data, intraweek fluctuation data, and seasonal fluctuation data can be stored as different columns, with timestamps used as indexes, to form a complete data table.

[0121] Preferably, step S2 comprises the following steps:

[0122] Step S21: setting the grid load rolling window according to the grid load time-frequency component data, and obtaining the annual trend baseline rolling window, the daily fluctuation baseline rolling window, the weekly fluctuation baseline rolling window and the seasonal fluctuation baseline rolling window respectively;

[0123] Step S22: Calculate the long-term trend baseline of the power grid load long-term trend data through the annual trend baseline rolling window to generate long-term trend baseline data;

[0124] Step S23: calculating the intraday fluctuation baseline of the power grid load intraday fluctuation data through the intraday fluctuation baseline rolling window to generate the intraday fluctuation baseline data;

[0125] Step S24: Calculate the weekly fluctuation baseline of the power grid load weekly fluctuation data through the weekly fluctuation baseline rolling window to generate weekly fluctuation baseline data;

[0126] Step S25: performing seasonal fluctuation baseline calculation on the seasonal fluctuation data of the power grid load through the seasonal fluctuation baseline rolling window to generate seasonal fluctuation baseline data;

[0127] Step S26: Perform a weighted combination of dynamic baseline values ​​based on the long-term trend baseline data, the intraday fluctuation baseline data, the weekly fluctuation baseline data, and the seasonal fluctuation baseline data to obtain the dynamic baseline value data of the power grid load.

[0128] In the embodiment of the present invention, different rolling windows are set according to the characteristics of the time-frequency component data of the power grid load, and are used to calculate the annual trend baseline, the intraday fluctuation baseline, the weekly fluctuation baseline and the seasonal fluctuation baseline. The size of the rolling window needs to be adjusted according to the periodic characteristics of different components. For example, for the annual trend baseline, a larger rolling window, such as 365 days, can be set; for the intraday fluctuation baseline, a 24-hour rolling window can be set; for the weekly fluctuation baseline, a 168-hour rolling window can be set; for the seasonal fluctuation baseline, a 90-day rolling window can be set, representing the four seasons of spring, summer, autumn and winter. These rolling windows can be adjusted according to actual conditions, for example, they can be fine-tuned according to the change law of historical data and the sensitivity requirements of early warning. The annual trend baseline rolling window determined in step S21 is used to calculate the baseline of the long-term trend data of the power grid load. For example, a rolling window of 365 days is used to calculate the long-term trend baseline value of each day. The calculation method of the baseline value can adopt the moving average method, the moving median method, etc. For example, the average value of the load data of the past 365 days can be used as the long-term trend baseline value at the current moment. The intraday fluctuation baseline rolling window determined in step S21 is used to perform a baseline calculation on the intraday fluctuation data of the power grid load. For example, a 24-hour rolling window is used to calculate the intraday fluctuation baseline value every hour. The calculation method of the baseline value can be selected according to the characteristics of the intraday fluctuation. For example, the average value of the load data at the same time in the past 24 hours can be calculated as the intraday fluctuation baseline value at the current time, so as to capture the periodic change law of the intraday load. The weekly fluctuation baseline rolling window determined in step S21 is used to perform a baseline calculation on the weekly fluctuation data of the power grid load. For example, a 168-hour rolling window is used to calculate the weekly fluctuation baseline value every hour. The average value of the load data at the same day of the week and the same time in the past week can be calculated as the weekly fluctuation baseline value at the current time, so as to capture the difference between weekday and weekend loads. The seasonal fluctuation baseline rolling window determined in step S21 is used to perform a baseline calculation on the seasonal fluctuation data of the power grid load. For example, a 90-day rolling window is used to calculate the seasonal fluctuation baseline value for each day. The average value of the load on the same date in the past 90 days can be calculated as the seasonal fluctuation baseline value of the current date to capture the change law of loads in different seasons. The various baseline data calculated in steps S22 to S25 are weighted and combined to obtain the final dynamic baseline value data of the power grid load. The weight coefficient can be set according to the contribution of different components to the load change, and can be adjusted according to the actual situation. For example, based on historical data analysis, the weights of the long-term trend baseline, intraday fluctuation baseline, weekly fluctuation baseline and seasonal fluctuation baseline can be determined to be 0.5, 0.2, 0.2, and 0.1 respectively. The baseline value of each component is multiplied by its corresponding weight and then summed to obtain the final dynamic baseline value.

[0129] Preferably, step S3 comprises the following steps:

[0130] Step S31: Obtain real-time power grid monitoring timestamp;

[0131] Step S32: Screening effective historical load data of standard historical multi-source power grid load data through real-time power grid monitoring timestamps, and calculating coefficient of variation to generate effective power grid historical load coefficient of variation;

[0132] Step S33: Analyze the warning source sensitive parameters according to the effective power grid historical load variation coefficient to generate the optimal warning sensitivity parameters;

[0133] Step S34: Calculate the load warning threshold value for the dynamic baseline value data of the power grid load according to the effective power grid historical load variation coefficient and the optimal alarm sensitivity parameter, and perform adaptive threshold constraints to obtain dynamic load warning threshold curve data.

[0134] In an embodiment of the present invention, a real-time timestamp of the current power grid monitoring is obtained. The timestamp is used to filter out the valid historical load data corresponding to the current time from the historical data. According to the real-time power grid monitoring timestamp, the time range to be filtered in the historical data is determined. In order to ensure the validity of the data, a time period similar to the current time is usually selected. For example, the data of a period of time before and after the same time of each day in the past three years is selected as the valid historical load data. During the screening process, abnormal data must also be excluded. For example, data points that exceed the historical data average plus or minus 3 times the standard deviation are removed. For the filtered effective historical load data, the coefficient of variation needs to be calculated. The coefficient of variation reflects the degree of discreteness of the data, and the calculation formula is: coefficient of variation = standard deviation / average value. For each set of effective historical load data, the mean and standard deviation of this set of data are calculated, and then the coefficient of variation is calculated to form the effective power grid historical load coefficient of variation. Define the range and step size of the sensitivity parameter. For example, the sensitivity parameter can be defined as between 1.0 and 2.0, with a step size of 0.1. Then, different sensitivity parameters need to be set for different coefficient of variation intervals. For example, the coefficient of variation can be divided into three intervals: low coefficient of variation (for example, less than 0.1), medium coefficient of variation (for example, between 0.1 and 0.3) and high coefficient of variation (for example, greater than 0.3). A sensitivity coefficient k can be set and multiplied by the coefficient of variation to obtain the fluctuation range of the warning threshold. The optimal alarm sensitivity parameter is selected by analyzing the warning effect under different sensitivity parameters, such as the false alarm rate and the missed alarm rate. For example, by simulating different values ​​such as k=1.5, 2.0, and 2.5, and combining the known abnormal conditions in the historical data for retrospective testing, the k value with low false alarm rate and missed alarm rate is selected as the optimal sensitivity parameter. According to the effective power grid historical load variation coefficient calculated in step S32 and the optimal alarm sensitivity parameter k obtained in step S33, the load warning threshold is calculated for the dynamic baseline value data of the power grid load obtained in step S26. The upper and lower limits of the dynamic load warning threshold curve are obtained by adding and subtracting the product of the sensitivity parameter k and the coefficient of variation from the dynamic baseline value. For example, add the dynamic baseline value to k multiplied by the value of the coefficient of variation to obtain the upper limit of the warning threshold curve; subtract the dynamic baseline value from k multiplied by the value of the coefficient of variation to obtain the lower limit of the warning threshold curve. In order to avoid the threshold being too large or too small, an adaptive threshold constraint can be set. For example, the maximum and minimum values ​​of the threshold can be set to limit the calculated threshold to a reasonable range. For example, the maximum and minimum values ​​of the threshold can be set according to the extreme values ​​or empirical values ​​of historical data. Finally, the dynamic load warning threshold curve data is obtained, which includes two curves, the upper limit and the lower limit.

[0135] Preferably, step S32 includes the following steps:

[0136] Step S321: Select a historical effective load window for the real-time power grid monitoring timestamp to obtain effective historical load window data;

[0137] Step S322: Screening the standard historical multi-source power grid load data through the effective historical load window data to obtain the effective historical load data of the power grid;

[0138] Step S323: Calculate the historical load mean value according to the effective historical load data of the power grid to generate effective historical load mean data;

[0139] Step S324: Calculate the historical load standard deviation based on the effective historical load data of the power grid to generate effective historical standard deviation data;

[0140] Step S325: Calculate the coefficient of variation based on the effective historical load mean data and the effective historical standard deviation data to generate the effective power grid historical load coefficient of variation.

[0141] In an embodiment of the present invention, the real-time power grid monitoring timestamp is analyzed to extract the date and time information. Then, the width of a historical effective load window is defined, for example, it can be set to a period of time near the same time point of the previous day or the previous few days. In order to ensure the representativeness of the data, data near the same date and time in the past few years are usually selected. Assume that the selected window width is 1 hour before and after, and the data of the same day in the past three years are selected. For example, if the load fluctuates quickly, a narrower window can be selected; if the load fluctuates slowly, a wider window can be selected. Traverse the standard historical multi-source power grid load data to check whether the timestamp of each load data falls within any of the above time windows. If the timestamp of a load data falls within a certain time window, the load data is filtered out and added to the effective historical load data set. In the screening process, it is also necessary to exclude outliers, for example, data that exceeds the mean value plus or minus 3 times the standard deviation, to avoid outliers affecting the subsequent coefficient of variation calculation. After excluding outliers, all filtered out effective historical load data are stored, for example, in a list, each element represents an effective historical load data. The average value of all load data in the three time windows is calculated for the load data in each time window. A mean value is calculated for each window. Then, the three mean values ​​can be used to calculate an overall mean value as the final effective historical load mean, or the three mean values ​​can be directly retained. In order to improve the stability of the data, the calculated mean value can be smoothed, for example, using a moving average method. For example, the average value of the mean values ​​of each time window and the two time windows before and after it is calculated as the final mean value. Similar to step S323, the standard deviation is calculated for the load data in each time window. The standard deviation of all load data in the three time windows is calculated respectively. A standard deviation is calculated for each window. Then, the three standard deviations can be used to calculate an overall standard deviation as the final effective historical standard deviation, or the three standard deviations can be directly retained. In order to improve the stability of the data, the calculated standard deviation can be smoothed, for example, using a moving average method. For example, the average value of the standard deviation of each time window and the two time windows before and after it is calculated as the final standard deviation. For each time window, or the overall mean and standard deviation, the coefficient of variation is calculated using a coefficient of variation calculation formula. The coefficient of variation calculation formula is: coefficient of variation = standard deviation / mean value. For example, for each time window, the standard deviation of the time window is divided by the mean of the time window to obtain the coefficient of variation corresponding to each time window. Alternatively, the overall standard deviation is divided by the overall mean to obtain a coefficient of variation.

[0142] Preferably, step S33 includes the following steps:

[0143] Step S331: performing a distribution characteristic analysis of the coefficient of variation of the effective power grid historical load, and setting an initial sensitivity parameter to obtain initial sensitivity parameter range data;

[0144] Step S332: Acquire power grid resource load warning source configuration data;

[0145] Step S333: connect the warning source according to the power grid resource load warning source configuration data, and read the historical warning record to obtain the original power grid warning record data;

[0146] Step S334: screening the real alarm records according to the original power grid early warning record data to obtain the real power grid alarm record data;

[0147] Step S335: Screening the false alarm records according to the original power grid early warning record data to obtain the power grid false alarm record data;

[0148] Step S336: performing sensitivity parameter cyclic traversal on the initial sensitivity parameter range data using the preset sensitivity traversal step data to obtain traversal sensitivity parameters;

[0149] Step S337: calculating the alarm recovery rate of the real alarm record data of the power grid and the false alarm record data of the power grid by traversing the sensitivity parameters, and obtaining the alarm recovery rate data of the power grid;

[0150] Step S338: Calculate the alarm precision rate of the real alarm record data of the power grid and the false alarm record data of the power grid by traversing the sensitivity parameters to obtain the alarm precision rate data of the power grid;

[0151] Step S339: Calculate the warning performance index of the ergodic sensitivity parameter according to the power grid alarm precision rate data and the power grid alarm recall rate data, and select the optimal sensitivity parameter to obtain the optimal alarm sensitivity parameter.

[0152] In an embodiment of the present invention, the distribution characteristics of these coefficients of variation are analyzed, for example, the minimum value, maximum value, mean value, median, standard deviation and other statistics of the coefficient of variation are calculated, and a histogram or box plot of the coefficient of variation is drawn to observe the distribution of the coefficient of variation. For example, it is assumed that the analysis finds that the minimum value of the coefficient of variation is close to 0, the maximum value is close to 0.5, the mean value is close to 0.2, and most of the coefficients of variation are distributed between 0.1 and 0.3. Then, according to the analysis results, the initial sensitivity parameter range is set. The sensitivity parameter is usually used to adjust the warning threshold. The larger the coefficient of variation, the smaller the sensitivity parameter should be, and vice versa. For example, the initial sensitivity parameter range can be set to between 1.0 and 2.0, or this range can be adjusted according to actual conditions. The sensitivity parameter range can be divided into multiple intervals according to the distribution of the coefficient of variation. The configuration data of the power grid resource load warning source is obtained. These configuration data include the connection information of the warning source, such as the database address, user name, password, etc., and the relevant parameters of the early warning system, such as the classification standard of the warning level, the sending method of the early warning information, etc. The connection mode can be a connection based on an API interface or a connection based on a database connection, depending on the type of early warning system. Assuming that the early warning system provides an API interface based on HTTP, Python's requests library can be used for connection and data request. Then, read the historical early warning record data. These data contain historical early warning event information, such as early warning time, early warning level, early warning reason, etc. For example, the early warning records of the past year are read from the database, including the early warning time, early warning level (level 1, level 2 or level 3), and early warning reason (for example, too high or too low load). Filter out the real alarm records from the original power grid early warning record data obtained in step S333. The real alarm refers to the early warning event of the actual power grid abnormality. This needs to be judged in combination with historical power grid operation data and expert experience. For example, it can be determined which early warning records are related to the real power grid abnormality based on the power outage events, equipment failures and other information recorded in the historical records. Filter out the false alarm records from the original power grid early warning record data obtained in step S333. The false alarm refers to the early warning system issuing an early warning, but no power grid abnormality actually occurred. This also needs to be judged in combination with historical power grid operation data and expert experience. For example, if after an early warning event occurs, the power grid operates normally and no equipment failure or abnormality is found, it can be judged as a false alarm. Using the preset sensitivity traversal step, the initial sensitivity parameter range obtained in step S331 is traversed in a loop to generate a series of traversal sensitivity parameters. For example, assuming that the initial sensitivity parameter range is [1.5, 3.0] and the preset traversal step is 0.1, the generated traversal sensitivity parameter sequence is 1.5, 1.6, 1.7, ..., 2.9, 3.0.Using the traversal sensitivity parameters generated in step S336, the alarm recall rate is calculated for the real alarm record data of the power grid and the false alarm record data of the power grid. The alarm recall rate refers to the proportion of real alarm events that are successfully detected by the early warning system. For each traversal sensitivity parameter, the historical real alarm events are used as a benchmark, the number of real alarm events that the early warning system can successfully detect under the sensitivity parameter is calculated, and then divided by the total number of real alarm events to obtain the alarm recall rate corresponding to the sensitivity parameter. Using the traversal sensitivity parameters generated in step S336, the alarm precision rate is calculated for the real alarm record data of the power grid and the false alarm record data of the power grid. The alarm precision rate refers to the proportion of real alarms among all the warnings issued by the early warning system. For each traversal sensitivity parameter, the number of all warnings issued by the early warning system under the sensitivity parameter, as well as the number of real alarms therein, is calculated, and then the number of real alarms is divided by the number of all warnings to obtain the alarm precision rate corresponding to the sensitivity parameter. Define one or more early warning performance indicators. Commonly used early warning performance indicators include F1 value, F2 value, or other customized indicators. F1 value is the harmonic mean of recall and precision, and the calculation formula is: F1 = 2 × (recall × precision) / (recall + precision). F2 value is the harmonic mean of recall weighted, and the calculation formula is F2 = 5 × (recall × precision) / (4 × precision + recall). According to the actual situation, you can choose F1 value, or customize a weighted combination to meet different needs. For example, if you pay more attention to the coverage of real alarms, you can choose a higher recall weight, and if you pay more attention to the accuracy of the warning, you can choose a higher precision weight. Then, traverse all sensitivity parameters and calculate the early warning performance indicator value corresponding to each sensitivity parameter, for example, calculate the F1 value or customized indicator value corresponding to each sensitivity parameter. For example, suppose that the calculated F1 value of sensitivity parameter 1.5 is 0.6, 1.6 is 0.7, 1.7 is 0.8, 1.8 is 0.85, 1.9 is 0.9, and 2.0 is 0.8. Compare the warning performance index values ​​corresponding to all sensitivity parameters, and select the sensitivity parameter with the largest warning performance index value as the optimal alarm sensitivity parameter.

[0153] As an example of the present invention, refer to Figure 2 As shown, Figure 1 Detailed implementation steps of step S4 in the embodiment are shown in the flowchart. In this embodiment, step S4 includes:

[0154] Step S41: Acquire real-time power grid resource equipment status data;

[0155] In an embodiment of the present invention, power grid resource equipment status data is collected in real time from various data sources. These data sources include but are not limited to SCADA systems, PMUs (phasor measurement units), smart meters, and other sensor devices. The data types collected cover various key parameters reflecting the operating status of the power grid, such as transformer load rate, line current, voltage, frequency, active power, reactive power, etc. The frequency of data collection can be adjusted according to actual needs, for example, data can be collected once per second, every minute, or every hour. For example, real-time active power, reactive power, voltage, and current data of all 110kV substations can be collected from the SCADA system once a minute; at the same time, voltage and current phasor data are collected once a second from the PMUs deployed on key lines. In addition, real-time electricity consumption data of users can also be collected from smart meters. The collected data needs to be quality-checked, for example, to check whether the data is complete, whether there are outliers, etc. For missing or erroneous data, corresponding processing is required, such as using interpolation methods to fill missing values, or using filtering algorithms to remove outliers.

[0156] Step S42: constructing a power grid monitoring resource topology structure according to the real-time power grid resource device status data to generate power grid monitoring resource topology data;

[0157] In an embodiment of the present invention, a power grid monitoring resource topology is constructed based on the real-time power grid resource device status data obtained in step S41. The topology describes the connection relationship between various devices in the power grid, such as the connection relationship between substations, lines, and loads. The construction of the topology can use a graphical modeling tool, or it can be represented by a data structure, such as an adjacency matrix or an adjacency list. For example, a topology graph containing all 110kV substations, 35kV substations, 10kV feeders, and user loads can be constructed. In the topology graph, nodes represent substations, feeders, and users, and edges represent line connection relationships. Each node and edge contains its corresponding attribute information, for example, a substation node contains its name, voltage level, geographic location, and other information; a line edge contains its length, impedance, and other information.

[0158] Step S43: collecting real-time loads of power grid monitoring nodes according to the power grid monitoring resource topology data to obtain real-time load data of power grid resources;

[0159] In an embodiment of the present invention, according to the power grid monitoring resource topology data constructed in step S42, real-time load data is collected from relevant monitoring nodes. The monitoring nodes may be substations, distribution lines, etc. For example, based on the topological structure data, it can be determined which substations and distribution lines need to collect load data. Then, the real-time load data of these nodes, such as active power, reactive power, etc., are read from data sources such as SCADA systems and smart meters. The collected data needs to be associated with the topological structure data, for example, the collected load data is associated with the corresponding substation or distribution line node. In this way, the load data can be combined with the power grid topology for analysis, for example, the total load of each substation, the load flow of each line, etc. can be calculated.

[0160] Step S44: performing Bayesian network node processing according to the power grid monitoring resource topology data, and constructing a power grid load prediction model through a preset convolutional neural network model, thereby obtaining a power grid load prediction model to be trained;

[0161] In an embodiment of the present invention, the topological data of power grid monitoring resources is processed by Bayesian network nodes. Bayesian network is a probabilistic graph model that can be used to represent the dependency between variables. Nodes in the power grid topology (such as substations, transformers, etc.) can be regarded as nodes in the Bayesian network, and the connection relationship between nodes can be regarded as edges in the Bayesian network. For each node, a conditional probability distribution can be constructed based on its historical load data and other relevant factors (such as temperature, humidity, etc.), reflecting that the load of the node is affected by other nodes and environmental factors. For example, a Bayesian network can be established in which the load of the transformer node is affected by its upstream node and factors such as temperature and humidity. The parameters in the Bayesian network can be learned through historical data. The output result of the Bayesian network is the load prediction value and corresponding confidence of each node. Secondly, a convolutional neural network (CNN) model needs to be preset for load prediction. The CNN model is good at processing data with spatial structure and is very suitable for analyzing power grid topological structure data. The input of the CNN model can be the historical load information of each node output by the Bayesian network, as well as the real-time topological structure data, and the output is the load prediction value of each node in the future period of time. The structure of the CNN model can include convolutional layers, pooling layers, fully connected layers, etc. For example, a CNN model with multiple convolutional layers and pooling layers can be designed to extract features in the topological structure, and then use the fully connected layer for load forecasting. For example, it can be set that: the first layer is a convolutional layer, using 32 convolution kernels, the convolution kernel size is 3x3, the activation function uses the ReLU function, the second layer is a pooling layer, using maximum pooling, the pooling window size is 2x2, the third layer is a convolutional layer, using 64 convolution kernels, the convolution kernel size is 3x3, the activation function uses the ReLU function, the fourth layer is a pooling layer, using maximum pooling, the pooling window size is 2x2, the fifth layer is a fully connected layer, and the number of output layers is the number of predicted loads. Finally, the output result of the Bayesian network is used as the input of the CNN model to construct a power grid load forecasting model to be trained.

[0162] Step S45: performing model training set processing on the standard historical multi-source power grid load data, and performing model training on the power grid load prediction model to be trained to obtain a power grid load prediction model;

[0163] In an embodiment of the present invention, standard historical multi-source power grid load data is processed into a model training set. This includes steps such as data cleaning, feature engineering, and data normalization. For example, the historical load data can be arranged in chronological order, and features related to load forecasting, such as date type (weekdays, weekends, holidays), weather data (temperature, humidity), historical load values, etc., can be extracted. Then, these features are normalized, for example, the feature values ​​are scaled to the range of [0,1] to improve the efficiency and stability of model training. The processed training set is input into the power grid load forecasting model to be trained constructed in step S44 for training. During the training process, the model will learn the law of load changes based on the training data and adjust the model parameters to minimize the prediction error. A variety of optimization algorithms can be used to train the model, such as stochastic gradient descent (SGD), Adam optimizer, etc. After the training is completed, a trained power grid load forecasting model is obtained, which can be used for short-term load forecasting.

[0164] Step S46: Use the power grid load forecasting model to perform short-term load forecasting on the real-time load data of power grid resources to obtain short-term load forecasting data.

[0165] In an embodiment of the present invention, the real-time load data of the power grid resources obtained in step S43 is used as input and input into the power grid load prediction model trained in step S45 for short-term load prediction. For example, the load value for the next 24 hours can be predicted. The prediction results are output in the form of a time series, for example, the predicted load value for each hour in the next 24 hours can be output. These prediction data will be used for subsequent warning threshold comparison and warning triggering. The duration of the prediction can be adjusted according to actual needs, for example, the load value for the next 12 hours, 48 ​​hours or longer can be predicted.

[0166] As an example of the present invention, refer to Figure 3 As shown, Figure 1 Detailed implementation steps of step S5 in the flowchart, in this example, step S5 includes:

[0167] Step S51: extracting the forecast time point of the short-term load forecast data to obtain the forecast time point data of the power grid resources;

[0168] In the embodiment of the present invention, the short-term load forecast data is usually stored in the form of a time series, including a forecast time point and a corresponding forecast load value. For example, if the load for the next 24 hours is forecasted, each hour is a forecast time point. The extracted power grid resource forecast time point data is a timestamp sequence.

[0169] Step S52: using the power grid resource prediction time point data to identify the real-time load warning threshold interval of the dynamic load warning threshold curve data, and performing threshold load difference calculation on the short-term load prediction data to obtain predicted load difference data;

[0170] In an embodiment of the present invention, the prediction time point data is traversed. For each prediction time point, it is necessary to find the warning threshold corresponding to the time point in the dynamic load warning threshold curve data. If the threshold curve data does not exactly match the prediction time point, the closest warning threshold data can be selected, or the corresponding warning threshold can be calculated using an interpolation method. Then, based on the warning threshold, the difference between the short-term load forecast data at the time point and the corresponding warning threshold is calculated. The predicted load difference for each time point and each monitoring node is stored, for example, stored as a dictionary with the key being the time point and the monitoring node, and the value being the corresponding load difference.

[0171] Step S53: Calculate the load over-limit range according to the predicted load difference data to obtain the power grid predicted load over-limit range data;

[0172] In an embodiment of the present invention, the load excess amplitude is calculated based on the predicted load difference data obtained in step S52. The load excess amplitude refers to the percentage of the predicted load value exceeding the warning threshold interval. For example, if the predicted load difference at a time point is 100MW and the corresponding upper threshold limit is 1000MW, the load excess amplitude at that time point is (100 / 1000)×100%=10%. If the predicted load value is lower than the lower threshold limit, the load excess amplitude is a negative value. For example, if the predicted load value is 700MW and the lower threshold limit is 800MW, the load excess amplitude is (-100 / 800)×100%=-12.5%.

[0173] Step S54: divide the warning levels according to the power grid predicted load over-limit amplitude data, and perform multi-level warning trigger processing through a preset warning level execution strategy to obtain a power grid multi-level warning trigger strategy.

[0174] In an embodiment of the present invention, the warning level is divided according to the power grid predicted load over-limit amplitude data calculated in step S53. The warning level is usually divided into multiple levels, such as a first-level warning, a second-level warning, a third-level warning, etc. The classification standard of the warning level can be set according to the magnitude of the load over-limit amplitude. For example, the following warning rules can be set: when the load over-limit amplitude is greater than 10%, a first-level warning is triggered; when the load over-limit amplitude is between 5% and 10%, a second-level warning is triggered; when the load over-limit amplitude is less than -5%, a third-level warning is triggered. According to the preset warning level execution strategy, multi-level warning triggering processing is performed. For example, after the first-level warning is triggered, the emergency plan is activated and power-limiting measures are taken; after the second-level warning is triggered, the relevant departments are notified to prepare; after the third-level warning is triggered, the load changes are closely monitored. The warning level execution strategy can be adjusted according to the actual situation. For example, different emergency measures can be set according to the severity of different warning levels.

[0175] Preferably, the present invention further provides a power grid resource dynamic threshold warning system based on multi-source data, which executes the power grid resource dynamic threshold warning method based on multi-source data as described above, and the power grid resource dynamic threshold warning system based on multi-source data includes:

[0176] The historical load time-frequency decomposition module is used to obtain standard historical multi-source power grid load data; classify the standard historical multi-source power grid load data by historical load monitoring date, and decompose the power grid load time-frequency components to generate power grid load time-frequency component data;

[0177] A load dynamic baseline construction module is used to set a grid load rolling window according to the grid load time-frequency component data, and to perform weighted combination of dynamic baseline values ​​to obtain grid load dynamic baseline value data;

[0178] The adaptive threshold generation module is used to screen the effective historical load data of the standard historical multi-source power grid load data, calculate the coefficient of variation, and generate the effective power grid historical load coefficient of variation; analyze the warning source sensitive parameters according to the effective power grid historical load coefficient of variation, and generate the optimal warning sensitivity parameters; perform adaptive load warning threshold processing on the dynamic baseline value data of the power grid load according to the effective power grid historical load coefficient of variation and the optimal warning sensitivity parameters, and obtain the dynamic load warning threshold curve data;

[0179] The short-term load forecasting module is used to obtain real-time power grid resource equipment status data; construct the power grid monitoring resource topology structure according to the real-time power grid resource equipment status data to generate power grid monitoring resource topology data; construct a power grid load forecasting model according to the power grid monitoring resource topology data; use the power grid load forecasting model to perform short-term load forecasting and obtain short-term load forecasting data;

[0180] The multi-level warning trigger module is used to calculate the threshold load difference of short-term load forecast data through dynamic load warning threshold curve data, and perform multi-level warning trigger processing to obtain a multi-level warning trigger strategy for the power grid.

[0181] The present application is to extract the periodic law and random fluctuation components of load changes by decomposing the time-frequency components of historical multi-source power grid load data, so that the early warning method can more accurately capture the dynamic characteristics of load changes. By classifying the monitoring date of historical load data, distinguishing the load patterns of different types of dates, and constructing corresponding early warning models respectively, the accuracy of early warning is further improved, and the problem that a single model is difficult to adapt to the differences in load patterns on different dates is avoided. The dynamic baseline value is constructed by using a rolling window mechanism and a weighted combination algorithm, which can track the load change trend in real time and adjust the baseline level adaptively according to the changes in historical data. This effectively solves the problem that the fixed threshold method cannot adapt to the periodic fluctuations and long-term trend changes of the load, so that the early warning threshold is no longer a static value, but a curve that can be dynamically adjusted with the load change, so as to be more in line with the actual load change situation and reduce the occurrence of false alarms and missed reports. The early warning source sensitive parameter analysis is performed based on the coefficient of variation of historical load data, and the optimal alarm sensitivity parameter is finally determined. According to the load fluctuation characteristics in different regions and different time periods, the sensitivity of the early warning is adaptively adjusted. For areas or time periods with large load fluctuations, the system will automatically increase the warning sensitivity to detect potential risks in a timely manner; for areas or time periods with small load fluctuations, the system will reduce the warning sensitivity accordingly to avoid excessive warnings and reduce unnecessary maintenance work.

[0182] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0183] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A power grid resource dynamic threshold early warning method based on multi-source data, characterized in that: The following steps are involved: Step S1: Obtain standard historical multi-source power grid load data; Classify the historical load monitoring date of standard historical multi-source power grid load data, decompose the power grid load time-frequency components, and generate power grid load time-frequency component data; Step S2: setting a grid load rolling window according to the grid load time-frequency component data, and performing weighted combination of dynamic baseline values ​​to obtain grid load dynamic baseline value data; Step S3: Screen the effective historical load data of the standard historical multi-source power grid load data, calculate the coefficient of variation, and generate the effective power grid historical load coefficient of variation; analyze the warning source sensitive parameters according to the effective power grid historical load coefficient of variation, and generate the optimal warning sensitivity parameters; perform adaptive load warning threshold processing on the power grid load dynamic baseline value data according to the effective power grid historical load coefficient of variation and the optimal warning sensitivity parameters, and obtain dynamic load warning threshold curve data; Step S4: Acquire real-time power grid resource equipment status data; Construct the topological structure of power grid monitoring resources according to the real-time power grid resource equipment status data to generate power grid monitoring resource topological data; construct a power grid load prediction model according to the power grid monitoring resource topological data; use the power grid load prediction model to perform short-term load prediction and obtain short-term load prediction data; Step S5: Calculate the threshold load difference of the short-term load forecast data through the dynamic load warning threshold curve data, and perform multi-level warning trigger processing to obtain a multi-level warning trigger strategy for the power grid.

2. The power grid resource dynamic threshold early warning method based on multi-source data according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: Acquire historical multi-source power grid resource load data; Step S12: unifying the multi-source data source formats of historical multi-source power grid resource load data, and performing data preprocessing to generate standard historical multi-source power grid load data; Step S13: classifying the historical load monitoring dates according to the standard historical multi-source power grid load data to obtain classified historical multi-source power grid load data; Step S14: performing fast Fourier transform processing on the classified historical multi-source power grid load data to generate historical power grid load spectrum data; Step S15: Decomposing the time-frequency components of the power grid load according to the historical power grid load spectrum data to generate the time-frequency component data of the power grid load.

3. The method for early warning of dynamic threshold value of power grid resources based on multi-source data according to claim 2 is characterized in that: Step S15 includes the following steps: Step S151: performing a main load frequency range analysis based on historical power grid load spectrum data to obtain historical load frequency range data; Step S152: Perform maximum wavelet decomposition layer number processing according to historical load frequency range data to obtain the maximum wavelet decomposition layer number; Step S153: dividing the decomposition level frequency range according to the maximum wavelet decomposition level number, and performing decomposition level mapping on the historical load frequency range data to obtain load frequency-decomposition level mapping data; Step S154: determining the decomposition layer coverage range of the load frequency-decomposition layer mapping data, and performing wavelet decomposition layer optimization to generate optimal decomposition layer data; Step S155: performing wavelet transform decomposition on the historical power grid load spectrum data based on the optimal decomposition layer number data to generate wavelet decomposition coefficients; Step S156: extracting wavelet approximate components according to the wavelet decomposition coefficients to obtain long-term trend data of power grid load; Step S157: extracting wavelet detail components according to the wavelet decomposition coefficients, and obtaining power grid load daily fluctuation data, power grid load weekly fluctuation data, and power grid load seasonal fluctuation data; Step S158: Combine the long-term trend data of power grid load, the daily fluctuation data of power grid load, the weekly fluctuation data of power grid load and the seasonal fluctuation data of power grid load into the time-frequency component of power grid resource monitoring load to obtain the time-frequency component data of power grid load.

4. The power grid resource dynamic threshold early warning method based on multi-source data according to claim 3 is characterized in that: Step S2 includes the following steps: Step S21: setting the grid load rolling window according to the grid load time-frequency component data, and obtaining the annual trend baseline rolling window, the daily fluctuation baseline rolling window, the weekly fluctuation baseline rolling window and the seasonal fluctuation baseline rolling window respectively; Step S22: Calculate the long-term trend baseline of the power grid load long-term trend data through the annual trend baseline rolling window to generate long-term trend baseline data; Step S23: calculating the intraday fluctuation baseline of the power grid load intraday fluctuation data through the intraday fluctuation baseline rolling window to generate the intraday fluctuation baseline data; Step S24: Calculate the weekly fluctuation baseline of the power grid load weekly fluctuation data through the weekly fluctuation baseline rolling window to generate weekly fluctuation baseline data; Step S25: performing seasonal fluctuation baseline calculation on the seasonal fluctuation data of the power grid load through the seasonal fluctuation baseline rolling window to generate seasonal fluctuation baseline data; Step S26: Perform a weighted combination of dynamic baseline values ​​based on the long-term trend baseline data, the intraday fluctuation baseline data, the weekly fluctuation baseline data, and the seasonal fluctuation baseline data to obtain the dynamic baseline value data of the power grid load.

5. The method for early warning of dynamic threshold value of power grid resources based on multi-source data according to claim 1 is characterized in that: Step S3 includes the following steps: Step S31: Obtain real-time power grid monitoring timestamp; Step S32: Screening effective historical load data of standard historical multi-source power grid load data through real-time power grid monitoring timestamps, and calculating coefficient of variation to generate effective power grid historical load coefficient of variation; Step S33: Analyze the warning source sensitive parameters according to the effective power grid historical load variation coefficient to generate the optimal warning sensitivity parameters; Step S34: Calculate the load warning threshold value for the dynamic baseline value data of the power grid load according to the effective power grid historical load variation coefficient and the optimal alarm sensitivity parameter, and perform adaptive threshold constraints to obtain dynamic load warning threshold curve data.

6. The method for early warning of dynamic threshold value of power grid resources based on multi-source data according to claim 5 is characterized in that: Step S32 includes the following steps: Step S321: Select a historical effective load window for the real-time power grid monitoring timestamp to obtain effective historical load window data; Step S322: Screening the standard historical multi-source power grid load data through the effective historical load window data to obtain the effective historical load data of the power grid; Step S323: Calculate the historical load mean value according to the effective historical load data of the power grid to generate effective historical load mean data; Step S324: Calculate the historical load standard deviation based on the effective historical load data of the power grid to generate effective historical standard deviation data; Step S325: Calculate the coefficient of variation based on the effective historical load mean data and the effective historical standard deviation data to generate the effective power grid historical load coefficient of variation.

7. The method for early warning of dynamic threshold value of power grid resources based on multi-source data according to claim 5 is characterized in that: Step S33 includes the following steps: Step S331: performing a distribution characteristic analysis of the coefficient of variation of the effective power grid historical load, and setting an initial sensitivity parameter to obtain initial sensitivity parameter range data; Step S332: Acquire power grid resource load warning source configuration data; Step S333: connect the warning source according to the power grid resource load warning source configuration data, and read the historical warning record to obtain the original power grid warning record data; Step S334: screening the real alarm records according to the original power grid early warning record data to obtain the real power grid alarm record data; Step S335: Screening the false alarm records according to the original power grid early warning record data to obtain the power grid false alarm record data; Step S336: performing sensitivity parameter cyclic traversal on the initial sensitivity parameter range data using the preset sensitivity traversal step data to obtain traversal sensitivity parameters; Step S337: calculating the alarm recovery rate of the real alarm record data of the power grid and the false alarm record data of the power grid by traversing the sensitivity parameters, and obtaining the alarm recovery rate data of the power grid; Step S338: Calculate the alarm precision rate of the real alarm record data of the power grid and the false alarm record data of the power grid by traversing the sensitivity parameters to obtain the alarm precision rate data of the power grid; Step S339: Calculate the warning performance index of the ergodic sensitivity parameter according to the power grid alarm precision rate data and the power grid alarm recall rate data, and select the optimal sensitivity parameter to obtain the optimal alarm sensitivity parameter.

8. The method for early warning of dynamic threshold value of power grid resources based on multi-source data according to claim 1 is characterized in that: Step S4 includes the following steps: Step S41: Acquire real-time power grid resource equipment status data; Step S42: constructing a power grid monitoring resource topology structure according to the real-time power grid resource device status data to generate power grid monitoring resource topology data; Step S43: collecting real-time loads of power grid monitoring nodes according to the power grid monitoring resource topology data to obtain real-time load data of power grid resources; Step S44: performing Bayesian network node processing according to the power grid monitoring resource topology data, and constructing a power grid load prediction model through a preset convolutional neural network model, thereby obtaining a power grid load prediction model to be trained; Step S45: performing model training set processing on the standard historical multi-source power grid load data, and performing model training on the power grid load prediction model to be trained to obtain a power grid load prediction model; Step S46: Use the power grid load forecasting model to perform short-term load forecasting on the real-time load data of power grid resources to obtain short-term load forecasting data.

9. The method for early warning of dynamic threshold value of power grid resources based on multi-source data according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: extracting the forecast time point of the short-term load forecast data to obtain the forecast time point data of the power grid resources; Step S52: using the power grid resource prediction time point data to identify the real-time load warning threshold interval of the dynamic load warning threshold curve data, and performing threshold load difference calculation on the short-term load prediction data to obtain predicted load difference data; Step S53: Calculate the load over-limit range according to the predicted load difference data to obtain the power grid predicted load over-limit range data; Step S54: divide the warning levels according to the power grid predicted load over-limit amplitude data, and perform multi-level warning trigger processing through a preset warning level execution strategy to obtain a power grid multi-level warning trigger strategy.

10. A power grid resource dynamic threshold early warning system based on multi-source data, characterized in that: The method for early warning the dynamic threshold value of power grid resources based on multi-source data according to claim 1 is used to execute the method. The system for early warning the dynamic threshold value of power grid resources based on multi-source data comprises: The historical load time-frequency decomposition module is used to obtain standard historical multi-source power grid load data; classify the standard historical multi-source power grid load data by historical load monitoring date, and decompose the power grid load time-frequency components to generate power grid load time-frequency component data; A load dynamic baseline construction module is used to set a grid load rolling window according to the grid load time-frequency component data, and to perform weighted combination of dynamic baseline values ​​to obtain grid load dynamic baseline value data; The adaptive threshold generation module is used to screen the effective historical load data of the standard historical multi-source power grid load data, calculate the coefficient of variation, and generate the effective power grid historical load coefficient of variation; analyze the warning source sensitive parameters according to the effective power grid historical load coefficient of variation, and generate the optimal warning sensitivity parameters; perform adaptive load warning threshold processing on the dynamic baseline value data of the power grid load according to the effective power grid historical load coefficient of variation and the optimal warning sensitivity parameters, and obtain the dynamic load warning threshold curve data; The short-term load forecasting module is used to obtain real-time power grid resource equipment status data; construct the power grid monitoring resource topology structure according to the real-time power grid resource equipment status data to generate power grid monitoring resource topology data; construct a power grid load forecasting model according to the power grid monitoring resource topology data; use the power grid load forecasting model to perform short-term load forecasting and obtain short-term load forecasting data; The multi-level warning trigger module is used to calculate the threshold load difference of short-term load forecast data through dynamic load warning threshold curve data, and perform multi-level warning trigger processing to obtain a multi-level warning trigger strategy for the power grid.

Images