An intelligent power grid load forecasting system and method

Abstract

The present invention relates to the technical field of electric load forecasting, and specifically provides an intelligent power grid load forecasting system and method. The system includes: a power grid load coupling relationship calculation module, a power grid load dynamic adjustment module, a meteorological impact on power grid load correction module, a load forecasting model input optimization module, and a power grid load forecasting error assessment module. In the present invention, the load change rate is extracted, the load increment and power flow ratio are calculated to depict the load dynamic coupling relationship. The voltage phase angle deviation is calculated in combination with the power grid topology to enhance the regional linkage of load forecasting. The parameters are adjusted based on the past load propagation delay to improve the model's response ability to sudden fluctuations. The load slope is analyzed in combination with meteorological data, and the temperature and humidity sensitive intervals are set to enhance the adaptability to extreme weather. The input parameters are optimized using the load coupling correction value to reduce data distortion and improve the long-term forecasting stability, and to improve the accuracy of load dynamic adjustment, meteorological impact modeling, and input optimization, providing precise support for power grid dispatching.

Description

Technical Field

[0001] The present invention relates to the technical field of electric load forecasting, and particularly to an intelligent power grid load forecasting system and method. Background Art

[0002] The technical field of electric load forecasting includes methods and technologies for analyzing and forecasting the load demand of a power system. The core content of this technical field involves using various data sources such as historical load data, meteorological information, user behavior patterns, and power grid operating status, combined with mathematical modeling, statistical analysis, and machine learning technologies, to forecast future electric loads. The research directions in this field mainly include short-term, medium-term, and long-term load forecasting, and the forecasting methods for different time scales have different focuses. For example, short-term load forecasting is mainly used for real-time scheduling and optimization of the power grid, relying on time series analysis methods and neural network models, while medium- and long-term forecasting relies more on macro factors such as economic growth trends, industrial structure adjustments, and population changes to provide a basis for power planning and investment decisions. In addition, this technical field also involves the uncertainty analysis of load forecasting, abnormal load detection, and the construction of adaptive forecasting models to improve the accuracy and reliability of forecasting.

[0003] Among them, an intelligent power grid load forecasting system refers to a system that combines electric load forecasting technology and uses means such as the Internet of Things, big data processing, and artificial intelligence to intelligently forecast the load demand of the power grid. This system includes multiple links such as data collection, data preprocessing, feature extraction, load forecasting modeling, and forecasting result analysis. The data collection link is responsible for collecting user electricity consumption data, environmental data, and power grid operating status information. The data preprocessing link processes the original data through methods such as denoising, missing value filling, and data normalization. The feature extraction link selects key factors that have a greater impact on load changes. The load forecasting modeling link uses deep learning models, regression analysis, or combined forecasting methods to construct a load forecasting model. The forecasting result analysis link is used to evaluate the forecasting accuracy and adjust and optimize the results. The main goal of this system is to improve the accuracy of load forecasting and provide a scientific basis for the optimal scheduling of the power system, demand response management, and new energy consumption.

[0004] The prior art fails to accurately quantify the load coupling relationship, making it difficult to reflect the load linkage effect between regions, resulting in a large deviation of the prediction model under load mutation conditions. There is a lack of a load propagation time delay correction mechanism, which cannot effectively adjust the prediction parameters to cope with the influence of past load fluctuations and reduces the adaptability to short-term drastic load changes. The processing method of meteorological factors is single, and the temperature and humidity sensitive intervals and the characteristics of power grid voltage fluctuations are not fully considered, resulting in a significant increase in prediction errors under extreme weather conditions. The optimization of input data is insufficient, ignoring the synergistic effect between load dynamic adjustment and environmental factors, and the model parameter configuration lacks real-time adjustment, affecting the stability of long-term prediction. The above problems restrict the reliability of power grid load prediction in complex environments and are not conducive to refined power dispatching. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies existing in the prior art and propose an intelligent power grid load prediction system and method.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions: An intelligent power grid load prediction system includes:

[0007] The power grid load coupling relationship calculation module extracts the load change rate from the regional power grid load data, calculates the load increment and the power flow ratio of adjacent substations, calls the power grid topology parameters to calculate the voltage phase angle offset, and calculates the weight ratio to obtain the power grid load coupling weight matrix;

[0008] The power grid load dynamic adjustment module calls the power grid load coupling weight matrix, obtains the real-time load change rate, calculates the load mutation factor, extracts the supply-demand offset and the power flow adjustment amount, calculates the voltage offset change, and after adjusting the weight matrix, calls the past load fluctuation propagation time delay to obtain the load coupling dynamic correction value;

[0009] The meteorological influence on power grid load correction module obtains the past meteorological data of the region, combines the load change rate, calculates the load slope and obtains the temperature sensitive interval and the humidity critical value, and calls the power grid voltage fluctuation range to obtain the meteorological load correction factor;

[0010] The load prediction model input optimization module calls the load coupling dynamic correction value, adjusts the load weights of adjacent regions in the load prediction model, calls the meteorological load correction factor, corrects the proportion of meteorological parameters, and calculates the input adjustment amplitude to obtain the optimized prediction input parameters;

[0011] The power grid load prediction error evaluation module calls the optimized prediction input parameters, runs the load prediction model, and calculates the error rate according to the power grid load data and the prediction error distribution trend to obtain the power grid load prediction error evaluation result.

[0012] As a further solution of the present invention, the power grid load coupling weight matrix includes load increment, adjacent substation power flow ratio, voltage phase angle offset, and weight ratio. The load coupling dynamic correction value includes load mutation factor, supply-demand offset, power flow adjustment amount, voltage offset change, and propagation delay of past load fluctuations. The meteorological load correction factor includes load slope, temperature sensitive interval, humidity critical value, and power grid voltage fluctuation range. The optimized predicted input parameters include load coupling dynamic correction value, adjacent area load weight, meteorological load correction factor, meteorological parameter proportion, and input adjustment amplitude. The load prediction error evaluation result includes power grid load data, prediction error distribution trend, and error rate.

[0013] As a further solution of the present invention, the power grid load coupling relationship calculation module includes:

[0014] The load change rate extraction sub-module obtains the regional power grid load data, extracts the load change amount in multiple time periods, calculates the load change rate per unit time, and filters out abnormal data according to the load change trend to obtain the load change rate sequence;

[0015] The load increment calculation sub-module calculates the load increment in adjacent time periods based on the load change rate sequence, collects the power flow change data of the substation, and matches the load increment distribution according to the power flow change trend to obtain the load increment matching sequence;

[0016] The load coupling weight matrix calculation sub-module calculates the voltage phase angle offset by invoking the power grid topology parameters based on the load increment matching sequence, constructs the load distribution relationship between substations according to the offset, and uses the formula:

[0017] ;

[0018] Calculate the load coupling weight ratio between multiple substations, collect the calculation results, and establish the power grid load coupling weight matrix;

[0019] Wherein, represents the load coupling weight between substation and , represents the load power of substation , represents the voltage phase angle of substation , represents the sum of the load differences between substation and all substations, represents the sum of the squares of the voltage phase angle differences between substation and all substations.

[0020] As a further solution of the present invention, the power grid load dynamic adjustment module includes:

[0021] The load change calculation sub-module calls the power grid load coupling weight matrix, calculates the real-time load change rate, extracts the load mutation factors during the operation of the power grid, screens the key load fluctuation nodes based on the load change rate, calculates the load propagation ratio between multiple nodes, and generates the key load change ratio;

[0022] The load offset adjustment sub-module calls the key load change ratio, extracts the supply-demand difference data, calculates the supply-demand offset, and combines the load propagation ratio to obtain the power flow adjustment amount of multiple load nodes, using the formula:

[0023] ;

[0024] Operate to obtain the power adjustment value of multiple load nodes, integrate the adjusted voltage offset data, and obtain the voltage offset change;

[0025] Among them, represents the power flow adjustment amount, represents the weight of load node for the overall load adjustment, represents load node 's current load demand value, represents load node 's current supply power value, is a small parameter to avoid a zero denominator, representing the total number of power grid load nodes;

[0026] The load coupling dynamic correction sub-module calls the voltage offset change, adjusts the power grid load coupling weight matrix, and combines the past load fluctuation propagation delay to calculate the load coupling correction coefficient to obtain the load coupling dynamic correction value.

[0027] As a further solution of the present invention, the meteorological impact on power grid load correction module includes:

[0028] The meteorological data extraction sub-module obtains the past meteorological data of the region, extracts the parameters of temperature and humidity, screens the meteorological information that meets the data integrity requirements, excludes the data missing items, calculates the temperature and humidity change ranges in different time periods, statistically analyzes the temperature change trend and humidity change rate, and adjusts the abnormal data fluctuation range to generate the regional meteorological parameter set;

[0029] The meteorological load change calculation sub-module, based on the regional meteorological parameter set, calls the power grid load data, calculates the load change trend under different meteorological conditions, obtains the load change rate according to the time series data, and combines the temperature and humidity change rate, using the formula:

[0030] ;

[0031] Calculate to obtain the load slope;

[0032] Among them, represents the load slope, represents the load change within the time period and represents the weight coefficient of the time period and represents the temperature change, represents the humidity value of the time period and represents the humidity critical value, represents the number of total time periods considered when calculating the load slope;

[0033] The meteorological load correction factor calculation sub-module calls the load slope, combines the regional meteorological parameter set, screens the temperature-sensitive interval range, calculates the proportion of the effect of temperature on load change, obtains the humidity critical value, and calls the power grid voltage fluctuation range, calculates the weight of the meteorological influence factor, and combines multiple influence parameters to calculate the meteorological load correction factor.

[0034] As a further solution of the present invention, the load prediction model input optimization module includes:

[0035] The load weight adjustment sub-module adjusts the load weights of adjacent regions in the load prediction model based on the load coupling dynamic correction value, compares the load data of different regions, screens the change trend of the load weights of adjacent regions, adjusts the load weight value according to the load coupling dynamic correction value, calculates the proportion of the corrected regional load, and obtains the corrected load weight coefficient;

[0036] The meteorological parameter correction sub-module calls the meteorological load correction factor, calculates the correction value according to the change of the meteorological parameter ratio, compares the load change under different meteorological conditions, adjusts the proportion of the meteorological parameter in the load prediction, calculates the corrected meteorological influence parameter, and obtains the corrected meteorological influence factor;

[0037] The input parameter optimization calculation sub-module calls the corrected load weight coefficient and the corrected meteorological influence factor, combines the current load prediction input data, and uses the formula:

[0038] ;

[0039] Performs operations to obtain the optimized adjusted input parameter value and obtains the optimized prediction input parameter;

[0040] Among them, represents the optimized prediction input parameter, represents the corrected load weight coefficient, represents the load data of adjacent regions, represents the corrected meteorological influence factor, represents the proportion of meteorological parameters, represents the load change rate factor, represents the time period parameter, represents the time decay factor of the load impact, represents the upper limit of the calculation, which is the maximum value of an index or range.

[0041] As a further solution of the present invention, the system further includes a power grid load prediction error evaluation module;

[0042] The power grid load prediction error evaluation module calls the optimized prediction input parameters, runs the load prediction model, calculates the error rate based on the power grid load data and the prediction error distribution trend, and obtains the load prediction error evaluation result;

[0043] The load prediction error evaluation result includes power grid load data, prediction error distribution trend, and error rate.

[0044] As a further solution of the present invention, the power grid load prediction error evaluation module includes:

[0045] The prediction input optimization sub-module calls the optimized prediction input parameters, obtains the power grid load data and the prediction error distribution data, screens the key variables affecting the load prediction accuracy, calculates the deviation degree of multiple variables, compares the deviation amplitude with the error distribution interval, and adjusts the variable weights to obtain the optimized variable weight value;

[0046] The error rate calculation sub-module, based on the optimized variable weight value, obtains the power grid load data and the prediction error distribution data, calculates the error amplitude in multiple time intervals, monitors the relationship between the error amplitude and the load mean value, and uses the formula:

[0047] ;

[0048] Calculates the error trend change amplitude to obtain the error trend amplitude value;

[0049] where, represents the time period error rate, represents the time period actual load value, represents the time period predicted load value, represents the time period load deviation correction value, represents the th optimized weight of the variable in the time period represents the total number of variables, represents the time period error adjustment parameter of

[0050] Based on the error trend amplitude value, the error evaluation result generation sub-module analyzes the change in error distribution, calculates the deviation degree between the error rate and the error threshold, judges the fluctuation of the error rate, and obtains the load prediction error evaluation result.

[0051] An intelligent power grid load forecasting method, which is executed based on the above intelligent power grid load forecasting system, includes the following steps:

[0052] S1: Obtain the regional power grid load data, extract the load change rates of multiple nodes, calculate the load increment and power flow ratio of adjacent substations, call the power grid topology parameters to calculate the voltage phase angle offset, calculate the load coupling ratio between multiple substations through normalization, and adjust the weight ratio to obtain the power grid load coupling weight matrix;

[0053] S2: Based on the power grid load coupling weight matrix, obtain the real-time load change rate, calculate the load mutation factor, extract the supply-demand offset and power flow adjustment amount according to the power flow ratio, call the voltage phase angle offset to calculate the voltage offset change, combine the load fluctuation propagation delay in the past period to adjust the power grid load coupling weight matrix, and calculate the load coupling dynamic correction value to obtain the load coupling dynamic correction value;

[0054] S3: Obtain the historical meteorological data of the region, calculate the load slope in combination with the load coupling dynamic correction value, extract the temperature sensitive interval and humidity critical value, call the power grid voltage fluctuation range to calculate the meteorological load offset, and obtain the meteorological load correction factor;

[0055] S4: Call the load coupling dynamic correction value to adjust the load weights of adjacent regions in the load prediction input, call the meteorological load correction factor to correct the meteorological parameter ratio, calculate the input adjustment amplitude, and obtain the optimized prediction input parameters;

[0056] S5: Call the optimized prediction input parameters, run the load prediction calculation, analyze the prediction error distribution trend based on the power grid load data, calculate the error rate, and obtain the load prediction error evaluation result.

[0057] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0058] In the present invention, by extracting the load change rate, calculating the load increment and the power flow ratio, the dynamic coupling relationship between loads is accurately characterized, and the voltage phase angle offset is calculated in combination with the power grid topology, enabling the load prediction to have the characteristic of regional linkage. The prediction parameters are adjusted based on the past load propagation delay to improve the response ability of the model to sudden load fluctuations. The load slope is analyzed in combination with meteorological data, the temperature and humidity sensitive intervals are set, and the weights of meteorological factors are optimized to enhance the adaptability of the prediction model under extreme weather conditions. The input parameter configuration is optimized using the load coupling correction value to reduce data distortion and improve the long-term prediction stability. This solution improves the prediction accuracy in terms of load dynamic adjustment, meteorological impact modeling and input parameter optimization, providing more accurate support for power grid dispatching. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is the system flow chart of the present invention;

[0060] Figure 2 is the flow chart of the power grid load coupling relationship calculation module of the present invention;

[0061] Figure 3 is the flow chart of the power grid load dynamic adjustment module of the present invention;

[0062] Figure 4 is the flow chart of the meteorological impact on power grid load correction module of the present invention;

[0063] Figure 5 is the flow chart of the load prediction model input optimization module of the present invention;

[0064] Figure 6 is the flow chart of the power grid load prediction error evaluation module of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0065] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0066] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention. In addition, in the description of the present invention, the meaning of "a plurality" is two or more unless otherwise specifically defined.

[0067] Embodiment 1: Please refer to Figure 1 , the present invention provides a technical solution: An intelligent power grid load forecasting system includes:

[0068] The power grid load coupling relationship calculation module extracts the load change rate from the regional power grid load data, calculates the load increment and the adjacent substation power flow ratio, calls the power grid topology parameters to calculate the voltage phase angle offset, and calculates the weight ratio to obtain the power grid load coupling weight matrix;

[0069] The power grid load dynamic adjustment module calls the power grid load coupling weight matrix, obtains the real-time load change rate, calculates the load mutation factor, extracts the supply-demand offset and the power flow adjustment amount, calculates the voltage offset change, and calls the past load fluctuation propagation delay after adjusting the weight matrix to obtain the load coupling dynamic correction value;

[0070] The meteorological impact on power grid load correction module obtains the regional past meteorological data, combines the load change rate, calculates the load slope and obtains the temperature sensitive interval and the humidity critical value, and calls the power grid voltage fluctuation range to obtain the meteorological load correction factor;

[0071] The load prediction model input optimization module calls the load coupling dynamic correction value, adjusts the adjacent area load weight in the load prediction model, calls the meteorological load correction factor, corrects the meteorological parameter ratio, and calculates the input adjustment range to obtain the optimized prediction input parameters;

[0072] The power grid load prediction error evaluation module calls the optimized prediction input parameters, runs the load prediction model, and calculates the error rate according to the power grid load data and the prediction error distribution trend to obtain the power grid load prediction error evaluation result.

[0073] The power grid load coupling weight matrix includes the load increment, the adjacent substation power flow ratio, the voltage phase angle offset, and the weight ratio. The load coupling dynamic correction value includes the load mutation factor, the supply-demand offset, the power flow adjustment amount, the voltage offset change, and the past load fluctuation propagation delay. The meteorological load correction factor includes the load slope, the temperature sensitive interval, the humidity critical value, and the power grid voltage fluctuation range. The optimized prediction input parameters include the load coupling dynamic correction value, the adjacent area load weight, the meteorological load correction factor, the meteorological parameter ratio, and the input adjustment range. The power grid load prediction error evaluation result includes the power grid load data, the prediction error distribution trend, and the error rate.

[0074] Please refer to Figure 2 , the power grid load coupling relationship calculation module includes:

[0075] The load change rate extraction sub-module obtains the regional power grid load data, extracts the load change amount in multiple time periods, calculates the load change rate per unit time, and filters out abnormal data according to the load change trend to obtain the load change rate sequence;

[0076] The historical load power value of each substation is recorded by SCADA (Supervisory Control and Data Acquisition). The load data of each time period is stored in the database, as shown in Table 1, which contains the power data of four substations in a certain area at different time points. The system first extracts the data in chronological order, arranges the time series load power values ​​of each substation according to the time axis, and calculates the load change between adjacent time points. The calculation method is: Assume time The load power at the moment is , then the adjacent time intervals The load change rate within is calculated as:

[0077] ;

[0078] ;

[0079] in, The value is 15 minutes (i.e. 0.25 hours), and the load change rate The unit is MW / h. During the calculation process, all substation data must be traversed. When filtering abnormal data, it is necessary to determine whether the load change rate exceeds the set threshold. For example, if the normal load change rate of a substation is MW / h, when the load change rate within a certain period of time exceeds this range (such as the change rate reaches 10MW / h), the data is determined to be an abnormal value and is removed or adjusted. The load change rate sequence can be obtained through the above steps.

[0080] Table 1 Substation load power data:

[0081] ;

[0082] As shown in Table 1, substation A has arrive The load change is MW, the rate of change in 15 minutes is MW / h. If it exceeds the set threshold range, its load value will be removed or adjusted. This result shows that the load fluctuation in this time period is large, and it is necessary to further check whether there is any abnormality in the data source, and adjust the data or perform compensation calculations to ensure the stability and rationality of the final load change rate sequence.

[0083] The load increment calculation submodule calculates the load increments in adjacent time periods based on the load change rate sequence, collects the power flow change data of the substation, matches the load increment distribution according to the power flow change trend, and obtains the load increment matching sequence;

[0084] Based on the selected load change rate sequence, calculate the load increment for adjacent time periods, and the calculation method is as follows:

[0085] ;

[0086] The substation power flow change data needs to be collected through a power flow monitoring system. This system records the active power flow of the substation's incoming and outgoing lines in real time. The load increment needs to be matched with the power flow data. Assume that for a certain substation at the power flow data is as follows:

[0087] Table 2 Substation power flow data:

[0088] ;

[0089] At time , the incoming line power change of Substation A is MW, and the outgoing line power remains unchanged. Then the load power increases by MW, which matches the load change rate. If the load change trend of a certain substation is consistent with the power flow change trend (such as when the load increases, the power flow also increases), then the load increment of this substation is successfully matched. This result indicates that the load change of this substation is consistent with the power flow data and can be used for subsequent load coupling analysis, and ensures the accuracy of the load increment matching sequence.

[0090] Based on the load increment matching sequence, the load coupling weight matrix calculation sub-module calls the grid topology parameters to calculate the voltage phase angle offset, constructs the load distribution relationship between substations according to the offset, and uses the formula:

[0091] ;

[0092] Calculate the load coupling weight ratio between multiple substations, collect the calculation results, and establish the grid load coupling weight matrix;

[0093] Among them, represents the load coupling weight between Substation and , represents the load power of Substation , represents the voltage phase angle of Substation , represents the sum of the load differences between Substation and all substations, represents the sum of the squares of the voltage phase angle differences between Substation and all substations.

[0094] Suppose the load powers of Substations and are respectively And , calculate the difference in load power, and sum up the total difference in load power between all substations. The calculation formula is as follows:

[0095] ;

[0096] Among them, represents the voltage phase angle of substation . The voltage phase angle offset can be measured by a PMU (Phasor Measurement Unit). For example, the voltage phase angle data of substations in a certain area is as follows:

[0097] Table 3 Substation Voltage Phase Angle Data:

[0098] ;

[0099] Calculate the load coupling weight between substations A and B. Assume that MW, MW, calculate the load power difference MW. Assume that the total sum of load differences of all substations is 40 MW, then:

[0100] ;

[0101] The voltage phase angle difference is calculated as , and the sum of squares of voltage phase angle differences of all substations is . Calculate the weight coefficient:

[0102] ;

[0103] Final load coupling weight:

[0104] ;

[0105] This result shows that the load coupling degree between substations A and B is 5.7%, indicating that their load changes have a relatively low impact on each other, with a small weight in the load distribution relationship, and can be used for subsequent power grid load optimization analysis.

[0106] Please refer to Figure 3 , the power grid load dynamic adjustment module includes:

[0107] The load change calculation sub-module calls the power grid load coupling weight matrix, calculates the real-time load change rate, extracts the load mutation factors during power grid operation, and based on the load change rate, screens the key load fluctuation nodes, calculates the load propagation ratio between multiple nodes, and generates the key load change ratio;

[0108] Specifically in the calculation, first select the past Historical load data within a time step are used to establish a time series. By calculating the trend of load increase or decrease, the load change rate is determined. The calculation method is as follows:

[0109] ;

[0110] Among them, represents the load change rate, is the current time step, is the observation time window. During the calculation process, the system compares the current load change rate with the normal operation threshold. If it exceeds the set threshold (for example: the set load change rate threshold is ), then this node is marked as a possible load mutation node. Then, all marked load mutation nodes are screened, and it is compared whether the load fluctuation amount exceeds the standard deviation range of the average load fluctuation of the entire network (the set standard deviation is used as a measurement index. If , it is determined as a key load fluctuation node). For example, if the load demand change rate of a certain node is 0.15 p.u., while the set average load change rate of the system is p.u., and the standard deviation is p.u., then this node is determined as a key load fluctuation node.

[0111] The load propagation ratio between the selected key load fluctuation nodes is obtained through historical data calculation. The calculation method is as follows:

[0112] ;

[0113] Among them, represents the load propagation ratio between node and . This ratio is used to evaluate the degree of mutual influence of load fluctuations. For example, if the load demands of node 1 and node 2 are MW and MW respectively, then the propagation ratio is . This result indicates that the load fluctuation of node 1 has a low impact on node 2, meaning that during load adjustment, the fluctuation of node 1 will not significantly affect the power flow of node 2. Therefore, in subsequent load adjustment calculations, the interaction coefficient between node 1 and node 2 should be small.

[0114] The load offset adjustment sub-module calls the key load change ratio, extracts the supply-demand difference data, calculates the supply-demand offset amount, and combines the load propagation ratio to obtain the power flow adjustment amount of multiple load nodes. The formula is used:

[0115] ;

[0116] Calculate the power adjustment value of the multi-load node, integrate the adjusted voltage offset data, and obtain the voltage offset change;

[0117] Among them, represents the power flow adjustment amount, represents the load node weight for overall load adjustment, represents the load node current load demand value, represents the load node current supply power value, To avoid a small parameter with a zero denominator, represents the total number of grid load nodes;

[0118] First, call the key load change ratio data, and extract the supply-demand difference data from the historical load database, that is, calculate the supply-demand offset of each load node. The calculation method of the supply-demand offset is as follows:

[0119] ;

[0120] Among them, represents the current load demand value of the load node , represents the current supply power value of the load node . If , it means that there is a load gap at this node and additional power supply is needed to meet the load demand; if , it means that there is a supply redundancy at this node and the power output of this node needs to be reduced.

[0121] (1) Calculate the supply-demand offset of each node:

[0122] Taking three nodes in a certain power grid as an example, set their current load demands and supply powers as follows:

[0123] Table 4 Power Parameter Table of Load Nodes:

[0124] ;

[0125] As shown in Table 4, the supply-demand offset of Node 1 is the largest, which is 20 MW, indicating that the load demand of this node is relatively high and needs to be adjusted; the supply-demand offset of Node 2 is relatively small, which is 10 MW, and also needs a certain amount of adjustment; the supply power of Node 3 exceeds the demand power, so its supply-demand offset is negative (-20 MW), indicating that there is excess power output at this node and the output needs to be reduced or the load needs to be increased to balance the power grid supply and demand.

[0126] (2) Calculate the power flow adjustment amount:

[0127] To optimize the load distribution, the system needs to calculate the power flow adjustment amount of multiple load nodes , and the calculation formula is as follows:

[0128] ;

[0129] Based on the grid influence weights of the load nodes, the following weight parameters are set:

[0130] Node 1 (with the largest load gap): ;

[0131] Node 2 (with a smaller load gap): ;

[0132] Node 3 (with power supply surplus): ;

[0133] (3) Substitute the data to calculate the power adjustment value:

[0134] Substitute the load data in Table 4 for calculation:

[0135] ;

[0136] Calculate item by item:

[0137] ;

[0138] ;

[0139] ;

[0140] Finally, the calculation result is:

[0141] ;

[0142] (4) Significance and correlation of the numerical results:

[0143] This result indicates that during the current load adjustment process, the system needs to adjust the power of multiple nodes by 1.6253 MW to achieve overall power balance. This means that:

[0144] It is necessary to cut about 0.57 MW of power transmission from the nodes with power supply surplus (such as Node 3);

[0145] It is necessary to provide corresponding power adjustments to the nodes with supply-demand imbalance (such as Node 1 and Node 2) to reduce the load gap.

[0146] The load coupling dynamic correction sub-module calls the voltage offset change, adjusts the grid load coupling weight matrix, and combines the load fluctuation propagation delay in previous periods to calculate the load coupling correction coefficient, and obtains the load coupling dynamic correction value.

[0147] Calculating the voltage change rate based on the power flow data after load adjustment :

[0148] ;

[0149] Among them, is the adjusted node voltage, is the node voltage before adjustment. For example, if the voltage of a certain node before adjustment is 1.02 p.u. and the voltage after adjustment is 1.01 p.u., then the voltage change rate is:

[0150] ;

[0151] According to the voltage change situation, the system adjusts the grid load coupling weight matrix, and the adjustment of this matrix calculates the load coupling correction coefficient based on the historical load fluctuation propagation delay. The calculation method of the load coupling correction coefficient is as follows:

[0152] ;

[0153] Among them, represents the delay of the load adjustment response. Assuming that the adjustment data of the past 5 time steps in the historical record are respectively MW, and the corresponding propagation delay is seconds, then the calculated load coupling correction coefficient is:

[0154] ;

[0155] This result shows that during the historical load adjustment process, the propagation influence degree of the load adjustment of each node is relatively stable. The current load coupling correction value is 5.4 MW, which means that in the subsequent load adjustment, the grid load coupling weight matrix can be optimized based on this correction value to make the load adjustment more accurate and reduce the voltage deviation caused by load fluctuations.

[0156] Please refer to Figure 4 , the meteorological influence on the grid load correction module includes:

[0157] The meteorological data extraction sub-module obtains the regional historical meteorological data, extracts the parameters of temperature and humidity, screens the meteorological information that meets the data integrity requirements, excludes the data missing items, calculates the temperature and humidity change ranges in different time periods, statistics the temperature change trend and humidity change rate, and adjusts the abnormal data fluctuation range to generate the regional meteorological parameter set;

[0158] Call the meteorological historical data of the past five years in the specified area, including temperature and humidity parameters. Extract basic parameters such as daily average temperature, maximum temperature, minimum temperature, and relative humidity from multiple data sources, and filter out the data sets that meet the statistical conditions according to the data integrity requirements. That is, when the proportion of missing values in a certain time period is less than 5%, the data will be included in the calculation range. Otherwise, the data in that time period needs to be excluded or interpolated with adjacent data. Subsequently, calculate the temperature and humidity change ranges in different time periods. The specific method is to calculate the difference between the highest temperature and the lowest temperature within 24 consecutive hours and conduct statistics on a monthly basis to obtain the long-term trend. For example, in January of a certain year in a certain area, the average maximum temperature is 5°C and the average minimum temperature is -2°C, then the temperature change range in that month is 7°C. Similarly, calculate the humidity change range by statistically calculating the difference between the maximum value and the minimum value of the daily average humidity and summarizing it on a weekly or monthly basis, and then evaluate the humidity change trend. In addition, the abnormal data adjustment step needs to eliminate or correct the mutation data in combination with the historical mean. The specific method is to calculate the standard deviation of the data in each time period, and regard the data outside the range of the mean ± 3 times the standard deviation as abnormal values, and replace them with the mean of the data of the previous and next two days. For example, if the temperature data on a certain day in July 2023 suddenly increases to 50°C, while the historical average maximum temperature in that month is only 38°C, then this data should be corrected to the mean of the adjacent two days to ensure that the finally generated regional meteorological parameter set meets the requirements of statistical analysis.

[0159] Based on the regional meteorological parameter set, the meteorological load change calculation sub-module calls the power grid load data, calculates the load change trend under different meteorological conditions, obtains the load change rate according to the time series data, and combines the temperature and humidity change rates. Using the formula:

[0160] ;

[0161] Calculate the load slope;

[0162] Among them, represents the load slope, represents the time period within the load change amount, represents the time period of the weight coefficient, represents the temperature change amount, represents the time period of the humidity value, represents the humidity critical value, represents the total number of time periods considered when calculating the load slope;

[0163] The specific implementation steps are as follows: Extract the load dataset corresponding to the time series of meteorological data. For example, in a certain area, the daily maximum temperature on January 1, 2023 is 5°C, the minimum temperature is -3°C, and the relative humidity is 65%. Under this meteorological condition, the daily maximum load value is 250MW, and the minimum load value is 180MW. Then the load change amount is calculated as 250MW - 180MW = 70MW. Then, calculate the load change rate based on the time series data. The method is to calculate the load increment per hour. For example, if the load rises from 180MW to 200MW from 08:00 to 09:00 on a certain day, then the load change rate for that hour is (200 - 180)MW / 1h = 20MW / h. Next, calculate the load slope based on the temperature and humidity change rates. Use the formula:

[0164] ;

[0165] where the set weight coefficient is 0.8, and the humidity critical value is set to 60%. If the temperature change amounts measured within a week are [3, 5, 2, 4, 6, 3, 5]°C respectively, and the humidity values are [55, 60, 65, 58, 62, 59, 64]% respectively, and the corresponding load change amounts are [15, 18, 12, 16, 22, 14, 20]MW respectively, then the calculation process is as follows:

[0166] ;

[0167] ;

[0168] ;

[0169] The calculation result shows that under this meteorological condition, the change in temperature and humidity will result in a load change slope of 2.13MW / °C, which is used for the subsequent calculation of the meteorological load correction factor.

[0170] The meteorological load correction factor calculation sub-module calls the load slope, combines with the regional meteorological parameter set, screens the temperature-sensitive interval range, calculates the proportion of the effect of temperature on load change, obtains the humidity critical value, and calls the grid voltage fluctuation range, calculates the weight of the meteorological influence factor, and combines multiple influence parameters to calculate the meteorological load correction factor.

[0171] Statistically analyze the annual temperature change data, and screen the interval with the strongest correlation between temperature and load. For example, when the temperature in a certain area is between -5°C and 10°C, the load change is the most significant. Then set this range as the temperature-sensitive interval. Next, calculate the proportion of the effect of temperature on the load change. The method is to calculate the proportion of the load change within this temperature interval in the total load change. For example, if the total annual load change is 500 MW, and the load change within the temperature-sensitive interval is 300 MW, then the proportion of the effect is 300 / 500 = 60%. Then obtain the humidity critical value and call the power grid voltage fluctuation range. The specific operation is to extract the annual humidity and voltage data. For example, if the humidity fluctuation range within the year is 40% - 80%, and the power grid voltage fluctuation range is ±5V, then calculate the humidity influence factor weight using the normalization calculation formula:

[0172] ;

[0173] If the current humidity is 65%, then the humidity influence weight:

[0174] ;

[0175] Finally, calculate the meteorological load correction factor by combining multiple parameters such as temperature, humidity, and voltage fluctuation. Assume the temperature influence weight is 0.6, and the voltage influence weight is 0.4, then the correction factor:

[0176] ;

[0177] This result shows that after comprehensively considering meteorological factors, the calculated meteorological load correction factor is 3.46, which is used to correct the power grid load forecasting model.

[0178] As shown in Table 5:

[0179] Table 5 Meteorological load calculation data table:

[0180] ;

[0181] Table 5 gives the load change data and the calculated load slopes for different time periods, indicating the degree of influence of temperature and humidity changes on the load.

[0182] Please refer to Figure 5 , the load forecasting model input optimization module includes:

[0183] The load weight adjustment sub-module adjusts the load weights of adjacent regions in the load forecasting model based on the load coupling dynamic correction value, compares the load data of different regions, screens the changing trend of the load weights of adjacent regions, adjusts the load weight values according to the load coupling dynamic correction value, calculates the proportion of the corrected regional load, and obtains the corrected load weight coefficient;

[0184] First, obtain the historical load data of the target area and adjacent areas, decompose it into several time periods according to the time series. The data of each time period includes the load peak value, valley value and daily average load value. Calculate the load change rate of the adjacent areas in these time periods. By comparing the differences in the load change rates, select the adjacent areas with similar load change trends, and record their load change ratios as the preliminary weight coefficients. Subsequently, adjust based on the load coupling dynamic correction value, which is obtained by analyzing the load coupling degree between different areas. For example, if an area is greatly affected by meteorological factors, it is necessary to increase the influence weight of the meteorological coupling factor to correct it. If an area is significantly affected by industrial load fluctuations, it is necessary to reduce the meteorological influence weight and increase the industrial load influence weight. The calculation method of the corrected weight is set as follows: If the initial load weight value of the target area is 0.5, the initial load weight value of adjacent area A is 0.3, and that of area B is 0.2. Assuming that the calculated dynamic correction value is 0.1, then the final corrected weight of the target area is 0.5×(1 + 0.1) = 0.55, the weight of adjacent area A is adjusted to 0.3×(1 - 0.1) = 0.27, and that of area B is adjusted to 0.2×(1 - 0.1) = 0.18. When calculating the proportion of the corrected area load, through normalization processing, the sum of the weights of all areas is equal to 1, that is, the final weight value should be adjusted to , so as to obtain the corrected load weight coefficient. This result shows that the corrected load weight can more accurately reflect the dynamic influence relationship of the loads in each area, providing reliable parameter support for the input optimization of the subsequent load prediction model.

[0185] The meteorological parameter correction sub-module calls the meteorological load correction factor, calculates the correction value according to the change in the proportion of meteorological parameters, compares the load changes under different meteorological conditions, adjusts the proportion of meteorological parameters in load prediction, calculates the corrected meteorological influence parameters, and obtains the corrected meteorological influence factor;

[0186] First, obtain meteorological data, including factors such as temperature, humidity, wind speed and solar radiation. Establish a correlation model between each meteorological factor and the load, and calculate the proportion of the influence of each meteorological factor on the load. For example, in summer, the influence of temperature may account for 60%, humidity for 20%, wind speed for 10%, and solar radiation for 10%. In winter, the influence of temperature drops to 40%, wind speed accounts for 30%, humidity for 20%, and solar radiation for 10%. According to the meteorological factor weights calculated in different periods, determine the preliminary meteorological influence factor F. Subsequently, by comparing the meteorological load changes in different periods, calculate the correction value. Assume that the historical data of a certain area shows that when the temperature rises by 5°C, the load increases by 8%. If the current meteorological data change shows that the temperature rises by 3°C, then its influence can be calculated proportionally. Assuming that the initial value of the correction factor is 1.0, the new correction factor can be calculated according to the formula Calculate to obtain the corrected meteorological impact factor, which is used for subsequent load forecasting input optimization to improve the forecasting accuracy. The results show that through the corrected meteorological impact factor, the impact degree of meteorological factors on load changes can be more accurately described, thereby avoiding prediction errors caused by abnormal fluctuations in meteorological conditions and improving the stability and applicability of the load forecasting model.

[0187] The input parameter optimization calculation sub-module calls the corrected load weight coefficient and the corrected meteorological impact factor, combines the current load forecasting input data, and uses the formula:

[0188] ;

[0189] Perform operations to obtain the optimized input parameter values after adjustment, and obtain the optimized forecasting input parameters;

[0190] Among them, represents the optimized forecasting input parameters, represents the corrected load weight coefficient, represents the load data of adjacent regions, represents the corrected meteorological impact factor, represents the meteorological parameter ratio, represents the load change rate factor, represents the time period parameter, represents the time decay factor of load impact, represents the upper limit of the calculation, which is the maximum value of an index or range.

[0191] First, call the corrected load weight coefficient W and the load data L of adjacent regions, and perform weighted calculations on them to reflect the correction degree of the load weight on the input data. Then, call the corrected meteorological impact factor F, combine it with the meteorological parameter ratio P, and calculate the correction value of the input data. To avoid prediction deviations caused by extreme meteorological data, the meteorological parameter part is processed by square root transformation, that is , then, considering the time decay effect of load impact, set the load change rate factor λ, and calculate the exponential decay factor according to the time period parameter T , finally, calculate the optimized forecasting input parameters Taking a specific example to illustrate, assume there are 3 adjacent regions, and their load data L are 500MW, 450MW, and 480MW respectively. The corresponding corrected weight coefficients W are 0.4, 0.35, and 0.25, the corrected meteorological impact factors F are 1.2, 1.1, and 1.3 respectively, the corresponding meteorological parameter ratios P are 0.5, 0.6, and 0.4, the load change rate factor λ is 0.02, and the time period parameter T is 10h. Then the calculation process is as follows:

[0192] ;

[0193] The optimized predicted input parameters are calculated = 382.6 MW.

[0194] Result description:

[0195] This result indicates that after comprehensively considering the adjustment of load weights and the correction of meteorological parameters, the optimized predicted input parameters are 382.6 MW. This value, as the optimized input of the load forecasting model, can more accurately reflect the adjustment of the load affected by the load changes in adjacent regions and meteorological factors. Compared with the original predicted input, its corrected value can more precisely fit the actual load fluctuation trend, avoid prediction errors caused by single-factor deviations, and make the load forecasting model have higher adaptability and reliability under different regions and different meteorological conditions.

[0196] Table 6 Load weight and meteorological influence factor correction calculation table:

[0197] ;

[0198] As shown in Table 6, the corrected weights and meteorological influence factors of each region are all involved in the calculation, and finally the optimized predicted input parameter 382.6 MW is obtained. This value can be used for subsequent load forecasting calculations. This result shows that after considering the regional load influence and meteorological influence, the corrected input parameter can more reasonably serve as the input of the load forecasting model, ensuring that the predicted value is closer to the actual load demand.

[0199] Please refer to Figure 6 , the power grid load forecasting error evaluation module includes:

[0200] The prediction input optimization sub-module calls the optimized predicted input parameters, obtains the power grid load data and prediction error distribution data, screens the key variables affecting the load forecasting accuracy, calculates the deviation degree of multiple variables, compares the deviation amplitude with the error distribution interval, adjusts the variable weights, and obtains the optimized variable weight value;

[0201] Call the historical load data in the power grid operation database. This data includes the actual load values in a certain area at different time periods. Extract the load data for 24 hours each day according to the time series. At the same time, obtain the predicted load data calculated by the previous prediction model, and calculate the error distribution between the actual load value and the predicted load value. For the error distribution data, extract the mean value, standard deviation, and maximum error value of the error in different time periods to obtain the overall error distribution. During the process of screening the key variables that affect the load prediction accuracy, calculate the contribution degree of each variable to the error, set a contribution degree threshold. For example, under the benchmark of 0.05, only retain the variables with a contribution degree greater than this threshold, such as temperature, humidity, electricity price, historical load data, etc., and discard the variables with a low contribution degree. When calculating the deviation degree of multiple variables, normalize the screened variable data to ensure the comparability of variables with different dimensions. When calculating the deviation degree of each variable, use the deviation calculation formula , where is the variable value at a certain moment, is the mean value of this variable, is the standard deviation. For variables with a large deviation degree, that is variables, they are marked as high-deviation variables. During the process of comparing the deviation amplitude with the error distribution interval, compare the deviation degree of the variable with the fluctuation range of the prediction error. For example, if the 99% interval of the error distribution is , then the variables whose deviation degree falls within this interval are normal variables, otherwise they are abnormal variables. Subsequently, adjust the variable weights, re-allocate the weights of the variables according to the deviation degree, and adopt the weight adjustment rule , where is the initial weight value, is the maximum deviation value of all variables. After adjustment, the optimized variable weight value is obtained. This result shows that during the optimization process of variable weights, by calculating the deviation degree of each variable, the weight distribution of the variables can be effectively adjusted, reducing the sensitivity of the prediction error to key variables, thereby improving the stability and accuracy of the overall prediction.

[0202] The error rate calculation sub-module, based on the optimized variable weight value, obtains the power grid load data and the prediction error distribution data, calculates the error amplitude in multiple time intervals, monitors the relationship between the error amplitude and the load mean value, and uses the formula:

[0203] ;

[0204] Calculate the change amplitude of the error trend to obtain the error trend amplitude value;

[0205] Among them, represents the time period error rate, represents the time period actual load value, Representative time period Predicted load value Representative time period Load deviation correction value Representing the th variable's optimization weight in the time period Optimization weight Representing the total number of variables Representative time period Error adjustment parameter

[0206] When calculating the error margin of multiple time intervals, different time periods such as hours, days, weeks, etc. are selected for error calculation. The predicted input parameters optimized in the previous period are called to obtain load data, and the error is calculated according to the time dimension. The error calculation formula is:

[0207] ;

[0208] Among them, is the actual load value in the time period , and the actual grid load data for this time period is queried and extracted from the dispatching system database. For example, the load data at 16:00 in a certain area is 980 MW is the predicted load value, which is calculated by the prediction model. For example, the predicted value is 960 MW is the load deviation correction value, and its calculation method is . Assuming the correction coefficient , then the correction value , is the optimization weight of the th variable. For example, the weight of temperature is 0.3, the weight of humidity is 0.2, and the weight of electricity price is 0.5 is the error adjustment parameter, which is used to smooth the influence of errors. Its value is set to 5 MW 2 , substituting the above data into the calculation:

[0209] ;

[0210] This result indicates that the error rate in the current time period is relatively high, indicating that the calculation result of the prediction model has a large deviation from the actual load value, and may be affected by external factors such as temperature, humidity, and sudden changes in grid load. Therefore, it is necessary to further analyze the change trend of the error in combination with the calculation result of the error trend amplitude to determine whether it is necessary to correct the prediction model.

[0211] Based on the error trend amplitude value, the error assessment result generation sub-module analyzes the change in error distribution, calculates the deviation degree between the error rate and the error threshold, judges the fluctuation of the error rate, and obtains the load forecasting error assessment result.

[0212] For example, if the error threshold is set to 5%, compare the calculated error rate. If the error rate is higher than the threshold, it is determined that the prediction error in the current period is large. Monitor the change of the error rate in different time periods. For example, set the time interval to 1 hour, record the error rate in the last 24 hours, use the time series analysis method to calculate the error change trend, judge the fluctuation of the error rate, and calculate the standard deviation of the error. , where is the error mean value. Set the error rate data set in the last 24 hours as , calculate the mean value , calculate the standard deviation:

[0213] ;

[0214] If the standard deviation is greater than 2%, it is determined that the error fluctuation is large. Finally, the load forecasting error assessment result is obtained. This result indicates that the error fluctuation is large, which may indicate that the prediction model has poor adaptability in a specific time period, or the influence of external factors exceeds the compensation ability of the prediction model. Therefore, it is necessary to adjust the prediction parameters or optimize the input variables to reduce the prediction error in the future period and improve the prediction accuracy.

[0215] An intelligent power grid load forecasting method, which is executed based on the above intelligent power grid load forecasting system, includes the following steps:

[0216] S1: Obtain the regional power grid load data, extract the load change rates of multiple nodes, calculate the load increments and power flow ratios of adjacent substations, call the power grid topology parameters to calculate the voltage phase angle offset, normalize and calculate the load coupling ratio between multiple substations, and adjust the weight ratio to obtain the power grid load coupling weight matrix;

[0217] S2: Based on the power grid load coupling weight matrix, obtain the real-time load change rate, calculate the load mutation factor, extract the supply-demand offset and power flow adjustment amount according to the power flow ratio, call the voltage phase angle offset to calculate the voltage offset change, combine the past load fluctuation propagation delay to adjust the power grid load coupling weight matrix, and calculate the load coupling dynamic correction value to obtain the load coupling dynamic correction value;

[0218] S3: Obtain the past meteorological data of the region, calculate the load slope in combination with the load coupling dynamic correction value, extract the temperature sensitive interval and humidity critical value, call the power grid voltage fluctuation range to calculate the meteorological load offset, and obtain the meteorological load correction factor;

[0219] S4: Call the load coupling dynamic correction value to adjust the load weights of adjacent regions in the load prediction input, call the meteorological load correction factor to correct the proportion of meteorological parameters, calculate the input adjustment range, and obtain the optimized prediction input parameters;

[0220] S5: Call the optimized prediction input parameters, run the load prediction calculation, analyze the prediction error distribution trend based on the power grid load data, calculate the error rate, and obtain the load prediction error assessment result.

[0221] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution content of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A smart grid load forecasting system, characterized in that: The system comprises: The grid load coupling relationship calculation module extracts the load change rate from the regional grid load data, calculates the load increment and the power flow ratio of adjacent substations, calls the grid topology parameters to calculate the voltage phase angle offset, and calculates the weight ratio to obtain the grid load coupling weight matrix; The power grid load dynamic adjustment module calls the power grid load coupling weight matrix, obtains the real-time load change rate, calculates the load mutation factor, extracts the supply and demand offset and the power flow adjustment, calculates the voltage offset change, and calls the previous load fluctuation propagation delay after adjusting the weight matrix to obtain the load coupling dynamic correction value; The power grid load dynamic adjustment module includes: The load change calculation submodule calls the grid load coupling weight matrix, calculates the real-time load change rate, extracts the load mutation factor in the grid operation, and screens the key load fluctuation nodes based on the load change rate, calculates the load propagation ratio between multiple nodes, and generates the key load change ratio; The load offset adjustment submodule calls the key load change ratio, extracts the supply and demand difference data, calculates the supply and demand offset, and combines the load propagation ratio to obtain the power flow adjustment of multiple load nodes using the formula: ； Calculate and obtain the power adjustment values ​​of multiple load nodes, integrate the adjusted voltage offset data, and obtain the voltage offset change; in, represents the power flow adjustment, Represents load node The weight of the overall load adjustment, Represents load node The current load demand value, Represents load node The current power supply value, To avoid small parameters with zero denominator, Represents the total number of grid load nodes; The load coupling dynamic correction submodule calls the voltage offset change, adjusts the grid load coupling weight matrix, and calculates the load coupling correction coefficient in combination with the previous load fluctuation propagation delay to obtain the load coupling dynamic correction value; The weather-affected grid load correction module obtains the past regional weather data, combines the load change rate, calculates the load slope, obtains the temperature sensitive range and humidity critical value, calls the grid voltage fluctuation range, and obtains the weather load correction factor; The load forecasting model input optimization module calls the load coupling dynamic correction value, adjusts the load weights of adjacent areas in the load forecasting model, calls the meteorological load correction factor, corrects the proportion of meteorological parameters, and calculates the input adjustment range to obtain optimized forecast input parameters.

2. The smart grid load forecasting system according to claim 1, characterized in that: The grid load coupling weight matrix includes load increment, adjacent substation power flow ratio, voltage phase angle offset, and weight ratio; the load coupling dynamic correction value includes load mutation factor, supply and demand offset, power flow adjustment, voltage offset change, and previous load fluctuation propagation delay; the meteorological load correction factor includes load slope, temperature sensitive range, humidity critical value, and grid voltage fluctuation range; the optimized prediction input parameters include load coupling dynamic correction value, adjacent area load weight, meteorological load correction factor, meteorological parameter proportion, and input adjustment amplitude.

3. The smart grid load forecasting system according to claim 2, characterized in that: The grid load coupling relationship calculation module includes: The load change rate extraction submodule obtains regional power grid load data, extracts load changes in multiple time periods, calculates the load change rate per unit time, filters abnormal data based on load change trends, and obtains a load change rate sequence; The load increment calculation submodule calculates the load increments in adjacent time periods based on the load change rate sequence, collects the power flow change data of the substation, matches the load increment distribution according to the power flow change trend, and obtains the load increment matching sequence; The load coupling weight matrix calculation submodule calculates the voltage phase angle offset based on the load increment matching sequence and calls the grid topology parameters, and constructs the load distribution relationship between substations according to the offset, using the formula: ； Calculate the load coupling weight ratios between multiple substations, aggregate the calculation results, and establish a grid load coupling weight matrix; in, Representative substation and The load coupling weight between Representative substation The load power, Representative substation The voltage phase angle, Representative substation The sum of the load differences with all substations, Representative substation The sum of squares of voltage phase angle differences with all substations.

4. The smart grid load forecasting system according to claim 1, characterized in that: The weather-affected power grid load correction module comprises: The meteorological data extraction submodule obtains the past meteorological data of the area, extracts the parameters of temperature and humidity, screens the meteorological information that meets the data integrity requirements, excludes data missing items, calculates the temperature and humidity change ranges in differentiated time periods, calculates the temperature change trend and humidity change rate, and adjusts the abnormal data fluctuation range to generate a regional meteorological parameter set; The meteorological load change calculation submodule calls the power grid load data based on the regional meteorological parameter set, calculates the load change trend under differentiated meteorological conditions, obtains the load change rate based on the time series data, and combines the temperature and humidity change rate to adopt the formula: ； The load slope is calculated; in, represents the load slope, Representative time period The load variation within Representative time period The weight coefficient of represents the temperature change, Representative time period The humidity value, represents the humidity threshold, Represents the total number of time periods considered in calculating the load slope; The meteorological load correction factor calculation submodule calls the load slope, combines the regional meteorological parameter set, screens the temperature sensitive range, calculates the proportion of temperature on load change, obtains the humidity critical value, and calls the grid voltage fluctuation range, calculates the meteorological influencing factor weight, and combines multiple influencing parameters to calculate the meteorological load correction factor.

5. The smart grid load forecasting system according to claim 4, characterized in that: The load forecasting model input optimization module includes: The load weight adjustment submodule adjusts the load weights of adjacent areas in the load forecasting model based on the load coupling dynamic correction value, compares the differentiated regional load data, screens the load weight change trend of adjacent areas, adjusts the load weight value according to the load coupling dynamic correction value, calculates the corrected regional load proportion, and obtains the corrected load weight coefficient; The meteorological parameter correction submodule calls the meteorological load correction factor, calculates the correction value according to the change in the proportion of the meteorological parameters, compares the load changes under differentiated meteorological conditions, adjusts the proportion of the meteorological parameters in the load forecast, calculates the corrected meteorological impact parameters, and obtains the corrected meteorological impact factors; The input parameter optimization calculation submodule calls the modified load weight coefficient and the modified meteorological influence factor, combines the current load forecast input data, and adopts the formula: ； The optimized and adjusted input parameter values ​​are obtained by operation to obtain the optimized predicted input parameters; in, represents the optimized prediction input parameters, represents the corrected load weight coefficient, Represents the adjacent area load data, represents the corrected meteorological impact factor, Represents the proportion of meteorological parameters, represents the load change rate factor, represents the time period parameter, represents the time decay factor of the load effect, Represents the upper limit of the calculation, which is the maximum value of an indicator or range.

6. The smart grid load forecasting system according to claim 5, characterized in that: The system also includes a power grid load forecast error assessment module; The power grid load forecast error evaluation module calls the optimized forecast input parameters, runs the load forecast model, calculates the error rate according to the power grid load data and the forecast error distribution trend, and obtains the load forecast error evaluation result; The load forecast error evaluation result includes power grid load data, forecast error distribution trend, and error rate.

7. The smart grid load forecasting system according to claim 6, characterized in that: The power grid load forecast error evaluation module includes: The prediction input optimization submodule calls the optimized prediction input parameters, obtains the power grid load data and prediction error distribution data, screens the key variables that affect the load prediction accuracy, calculates the deviation degree of multiple variables, compares the deviation amplitude and the error distribution interval, adjusts the variable weights, and obtains the optimized variable weight values; The error rate calculation submodule obtains the grid load data and the prediction error distribution data based on the optimization variable weight value, calculates the error amplitude in multiple time intervals, and monitors the relationship between the error amplitude and the load mean using the formula: ； Calculate the error trend change amplitude to obtain the error trend amplitude value; in, Representative period Error rate, Representative period The actual load value, Representative period Forecast load values, Representative period Load deviation correction value, Representative Variables in the period The optimization weight of represents the total number of variables, Representative period Error adjustment parameters; The error assessment result generation submodule analyzes the error distribution change based on the error trend amplitude value, calculates the degree of deviation between the error rate and the error threshold, determines the fluctuation of the error rate, and obtains the load forecast error assessment result.

8. A smart grid load forecasting method, characterized in that: The smart grid load forecasting system according to any one of claims 1 to 7 comprises the following steps: S1: Obtain regional power grid load data, extract multi-node load change rates, calculate load increments and power flow ratios of adjacent substations, call power grid topology parameters to calculate voltage phase angle offsets, normalize and calculate load coupling ratios between multiple substations, and adjust weight ratios to obtain power grid load coupling weight matrix; S2: Based on the power grid load coupling weight matrix, obtain the real-time load change rate, calculate the load mutation factor, extract the supply and demand offset and the power flow adjustment according to the power flow ratio, call the voltage phase angle offset to calculate the voltage offset change, adjust the power grid load coupling weight matrix in combination with the previous load fluctuation propagation delay, calculate the load coupling dynamic correction value, and obtain the load coupling dynamic correction value; S3: Obtain the regional past meteorological data, calculate the load slope in combination with the load coupling dynamic correction value, extract the temperature sensitive range and humidity critical value, calculate the meteorological load offset by calling the grid voltage fluctuation range, and obtain the meteorological load correction factor; S4: calling the load coupling dynamic correction value to adjust the adjacent area load weight in the load forecast input, calling the meteorological load correction factor to correct the meteorological parameter proportion, calculating the input adjustment range, and obtaining the optimized forecast input parameters; S5: Call the optimized prediction input parameters, run the load prediction calculation, analyze the prediction error distribution trend based on the power grid load data, calculate the error rate, and obtain the load prediction error evaluation result.

Images