A line loss rate prediction method based on big data

By using big data technology in the power grid monitoring server, obtaining the data of the target area and adjusting the input layer weight of the prediction model, the problem of high resource and time costs of existing line loss rate prediction methods is solved, and more accurate line loss rate prediction is achieved. .

CN118381011BActive Publication Date: 2024-11-12GUANGDONG POWER GRID CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410482395.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-04-22
Publication Date
2024-11-12
Estimated Expiration
2044-04-22

AI Technical Summary

Technical Problem

When the existing line loss rate prediction method is used in different regions, the resource and time cost is too high, and the prediction results are not accurate enough to adjust the input layer data according to actual conditions.

Method used

The line loss rate prediction method based on big data is adopted to obtain the geographical information and living environment data of the target area, match the corresponding line loss rate prediction model, and based on coefficients such as the duration of the power load, the humidity deviation around the equipment and the equipment damage frequency, Adjust the input layer data weight of the initial prediction model to establish a more accurate prediction model.

Benefits of technology

Reduce resource and time costs, improve the accuracy of line loss rate prediction, make the prediction results more in line with the actual situation, and avoid resource waste and prediction errors.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118381011B_ABST
    Figure CN118381011B_ABST
Patent Text Reader

Abstract

The present invention discloses a line loss rate prediction method based on big data, which relates to the technical field of electric power monitoring; the method comprises matching geographic information data and living environment data of a target area with a line loss rate prediction model to obtain an initial prediction model; compared with establishing a separate prediction model for each area, the method avoids the problem of excessive resource and time costs, and establishes a modifiable coefficient by collecting a power load duration coefficient, a humidity deviation coefficient around equipment, and an equipment damage frequency coefficient, and comparing the modifiable coefficient with a preset threshold value, so as to more accurately adjust the input layer data weight ratio of the initial prediction model and realize accurate prediction of the line loss rate of the area.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of electric power monitoring, and in particular relates to a line loss rate prediction method based on big data. Background Art

[0002] Line loss is the power loss of the power grid, which reflects the planning, design and economic operation level of the power grid. It is also one of the key evaluation indicators of power grid companies. Line loss is one of the important cost factors in the operation of power systems. Therefore, the prediction of line loss rate is crucial to the operation, management and development of power systems. Existing technologies can accurately predict line loss rates, so that power companies can better control costs, optimize resource utilization, and thus improve operational efficiency.

[0003] The traditional method of predicting line loss rate is usually to make predictions by establishing models. In the actual line loss rate calculation process, when predicting line loss rate for different regions, almost all of them establish target prediction models for the region for prediction, which will lead to excessive resource and time costs. When using the existing line loss rate prediction model for prediction, if the focus of the input layer data cannot be changed according to the actual situation, the predicted line loss rate will eventually be not accurate enough. Summary of the invention

[0004] The purpose of the present invention is to solve the problem and to propose a line loss rate prediction method based on big data.

[0005] In a first aspect of the implementation of the present invention, a line loss rate prediction method based on big data is first proposed. The method is applied to a power grid monitoring server, the power grid monitoring server monitors a target power grid area, the target power grid area includes multiple power grid areas to be tested, and the power grid monitoring server stores multiple line loss rate prediction models. The method includes:

[0006] Obtaining geographic information data and living environment data of a target area, matching the geographic information data and living environment data of the area in a plurality of line loss rate prediction models, and obtaining an initial prediction model; the target area is any one of a plurality of power grid areas to be tested;

[0007] The target area is divided into multiple sub-areas, and the power load duration coefficient, the humidity deviation coefficient around the equipment and the equipment damage frequency coefficient of the target sub-area are extracted as correlation coefficients; the target sub-area is any one of the multiple sub-areas;

[0008] Obtaining a coefficient that may be modified according to the correlation coefficient, and comparing the coefficient that may be modified with a preset threshold value, determining a weight ratio of data in the input layer of the initial prediction model to obtain a target prediction model for the target sub-region;

[0009] The line loss rate of the target sub-area is calculated according to the target prediction model.

[0010] Optionally, obtaining geographic information data and living environment data of a target area, matching multiple line loss rate prediction models according to the regional geographic information data and living environment data, and obtaining an initial prediction model includes:

[0011] Acquire the altitude slope change, precipitation difference, and soil thermal conductivity of the target area as the geographic information data;

[0012] Acquire the population density, building usage type, and building area density of the target area as the living environment data;

[0013] Performing weighted summation on the geographic information data and the living environment data to obtain a regional coefficient of the target area;

[0014] The absolute difference between the regional coefficient and the regional coefficient of the preset line loss rate prediction model is calculated. When the absolute difference is less than a preset threshold, the line loss rate prediction model corresponding to the regional coefficient of the preset line loss rate prediction model is used as the initial prediction model.

[0015] Optionally, extracting the power load duration coefficient of the target sub-area includes:

[0016] Get the total duration of power overload in the sub-area within H time and mark it as R s ;

[0017] Get the total duration R of the power overload s The total time period elapsed is marked as R d ;

[0018] Get the total time period R d The total duration of the internal power overload is marked as R. h ;

[0019] Calculate the power load duration coefficient, the calculation expression is Lk = R s / R h ×R s .

[0020] Optionally, extracting the humidity deviation coefficient of the device surrounding the target sub-area includes:

[0021] Get the total number of devices G and the actual humidity around each device, marked as S g , g is the number of the actual humidity around the device, g∈[0,G]; get the preset range of the ambient humidity that can be tolerated for each device and mark it as [S ga , S gb], the actual humidity S around each device g At the corresponding preset tolerable ambient humidity [S ga , S gb ] are marked as W q ,q is S g In [S ga , S gb ], q∈[0,W], W∈[0,G]; W represents S g In [S ga , S gb ], and W is an integer, calculate the humidity deviation coefficient around the equipment, and the calculation expression is: Pe=ln(W / G+1); where Pe is the humidity deviation coefficient around the equipment.

[0022] Optionally, extracting the equipment damage frequency coefficient of the target sub-area includes:

[0023] Collect equipment operating hours and failure counts;

[0024] Calculate the Euclidean distance between the device to be predicted and other devices in the data set;

[0025] Select the appropriate K value based on the cross-validation method;

[0026] For each device to be predicted, select its K nearest neighbors;

[0027] The damage frequency coefficient of the equipment to be predicted is estimated by averaging the number of failures of the selected K nearest neighbors;

[0028] The expression to be calculated is:

[0029] Where Vd represents the equipment damage frequency coefficient, and Ni represents the number of failures of the i-th nearest neighbor.

[0030] Optionally, obtaining the coefficient to be modified according to the correlation coefficient includes:

[0031] The power load duration coefficient, the humidity deviation coefficient around the equipment, and the equipment damage frequency coefficient are used to construct a data evaluation model, and the model calculation expression is:

[0032]

[0033] In the formula, Flg is the coefficient that can be modified, Lk, Pe and Vd represent the power load duration coefficient, the equipment surrounding humidity deviation coefficient and the equipment damage frequency coefficient respectively, and a 1 、a 2 、a 3is the proportional coefficient of the power load duration coefficient, the humidity deviation coefficient around the equipment, and the equipment damage frequency coefficient, and a 1 、a 2 、a 3 Both are greater than 0.

[0034] Optionally, comparing the coefficient to be modified with a preset threshold value, determining the weight ratio of the data of the input layer of the initial prediction model to obtain the target prediction model of the target sub-region includes:

[0035] When the required modification coefficient is less than a reference threshold, using the initial prediction model as a target prediction model for the target sub-region, and predicting the line loss rate of the region by using the target prediction model;

[0036] When the modifiable coefficient is not less than a reference threshold, the corresponding preset prediction model input layer data weight ratio correspondence table is queried according to the specific numerical value of the modifiable coefficient, and the weight ratio of the input layer data of the initial prediction model is adjusted accordingly to obtain the target prediction model, and the target prediction model is used to predict the line loss rate of the area.

[0037] Beneficial effects of the present invention:

[0038] The present invention proposes a line loss rate prediction method based on big data. By acquiring regional geographic information data and living environment data of the target area, matching the regional geographic information data and living environment data with the line loss rate prediction model, an initial prediction model is obtained. Compared with establishing a separate prediction model for each area, this method avoids the problem of excessive resource and time costs. By establishing a modifiable coefficient and comparing it with a preset threshold, the input layer data weight ratio of the initial prediction model is adjusted more accurately, thereby achieving accurate prediction of the line loss rate of the area. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The present invention will be further described below in conjunction with the accompanying drawings.

[0040] Figure 1 A flowchart of a line loss rate prediction method based on big data is provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The term "and / or" herein is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the present invention, the description of "first", "second", etc. is only used for descriptive purposes, and cannot be understood as indicating or implying its relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first" and "second" can explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0042] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.

[0043] Example 1

[0044] The embodiment of the present invention provides a line loss rate prediction method based on big data. Figure 1 , Figure 1 A flowchart of a line loss rate prediction method based on big data provided by an embodiment of the present invention. The method is applied to a power grid monitoring server, the power grid monitoring server monitors a target power grid area, the target power grid area includes multiple power grid areas to be tested, and the power grid monitoring server stores multiple line loss rate prediction models. The method includes:

[0045] Obtaining geographic information data and living environment data of a target area, matching multiple line loss rate prediction models according to the regional geographic information data and living environment data, and obtaining an initial prediction model; the target area is any one of multiple power grid areas to be tested;

[0046] The target area is divided into multiple sub-areas, and the power load duration coefficient, the humidity deviation coefficient around the equipment and the equipment damage frequency coefficient of the target sub-area are extracted as correlation coefficients; the target sub-area is any one of the multiple sub-areas;

[0047] Obtain a coefficient that may be modified according to the correlation coefficient, and compare the coefficient with a preset threshold value to determine the weight ratio of the data in the input layer of the initial prediction model to obtain a target prediction model for the target sub-area;

[0048] The line loss rate of the target sub-area is calculated according to the target prediction model.

[0049] By acquiring the regional geographic information data and living environment data of the target area, matching the regional geographic information data and living environment data with the line loss rate prediction model, an initial prediction model is obtained. Compared with establishing a separate prediction model for each area, this method avoids the problem of excessive resource and time costs. By establishing a modifiable coefficient and comparing it with the preset threshold, the input layer data weight ratio of the initial prediction model is adjusted more accurately to achieve accurate prediction of the line loss rate of the area.

[0050] Example 2

[0051] Obtain geographic information data and living environment data of the target area, match them in multiple line loss rate prediction models according to the regional geographic information data and living environment data, and obtain the initial prediction model including:

[0052] Obtain the target area's altitude slope change, precipitation difference, and soil thermal conductivity as geographic information data;

[0053] Obtain the population density, building use type, and building area density of the target area as living environment data;

[0054] Altitude gradient change refers to the spatial instability of the altitude of the grid area to be tested. In some areas, the terrain may show large altitude changes, that is, the altitude is not stable, rather than tending to a specific value. In this case, the efficiency of power transmission may be affected. Because the change in altitude will increase the complexity required for power equipment to transmit electricity, resulting in an increase in the length and complexity of the transmission line, which in turn increases the energy loss during power transmission. Therefore, when the altitude gradient of the grid area to be tested changes greatly, it needs to be used as an important reference factor for selecting the initial prediction model.

[0055] The precipitation difference refers to the spatial difference in precipitation in the grid area to be tested. When there is a significant difference in precipitation between two areas, it may have different degrees of impact on power equipment. In areas with higher precipitation, there may be more humid environments and water vapor, which may increase the corrosion risk of power equipment and affect the performance and life of the equipment. Especially in areas with high precipitation, power equipment is more susceptible to moisture and corrosion, which in turn affects its stable operation and life. Therefore, when the precipitation difference in the grid area to be tested is different, it needs to be used as an important reference factor for selecting the initial prediction model.

[0056] Soil thermal conductivity refers to the ability of soil to conduct heat, which reflects the heat transfer rate in the soil. When the soil thermal conductivity of the power grid area to be tested is high, it means that the soil has better thermal conductivity. The cable is easier to dissipate heat in such a soil environment. In this case, the temperature rise of the cable may be reduced, thereby reducing the loss of the cable and the line loss rate of the line. High soil thermal conductivity helps to keep the cable at a lower temperature and reduce the loss of heat energy. On the contrary, when the soil thermal conductivity is low, the soil's ability to conduct heat is weak, which may cause the temperature around the cable to rise, increase the loss of the cable and the line loss rate of the line. Therefore, when the soil thermal conductivity of the power grid area to be tested is different, it needs to be used as an important reference factor for selecting the initial prediction model.

[0057] By selecting the altitude slope change, precipitation difference, and soil thermal conductivity of the power grid area to be tested, the geographical environment problems of the power grid area to be tested can be basically determined. The reason for selecting the altitude slope change, precipitation difference, and soil thermal conductivity is that compared with other geographic information data, these geographic information data can be obtained more intuitively, and can generally represent the geographical environment of the power grid area to be tested, and can comprehensively reflect the geographical environment problems of the power grid area to be tested, which is helpful for more accurate reference and selection of the initial prediction model.

[0058] Population density refers to the ratio of the population of the grid area to be tested to its area, and the population is closely related to electricity demand. Generally speaking, the higher the population density, the greater the electricity demand in the area, and thus the greater the pressure on power transmission and line loss. On the contrary, low-density areas may have smaller electricity demand and relatively lower line loss rates. Therefore, when the population density of the grid area to be tested is different, it needs to be used as an important reference factor for selecting the initial prediction model.

[0059] Building usage type refers to the distribution of building types in the grid area to be tested. If the industrial area in the grid area to be tested accounts for a large proportion, this may mean that the line loss rate in the area is large. Industrial areas usually have a large number of electrical equipment, including production equipment, mechanical equipment, etc. The large-scale operation of these equipment may lead to higher power loads and power consumption. In this case, the load on the power transmission line is increased, thereby increasing the power loss of the line. In contrast, if the grid area to be tested is mainly residential or commercial, its electrical equipment may be relatively small and the power load is small, so the line loss rate may be relatively low. Therefore, when the building usage types in the grid area to be tested are different, it needs to be used as an important reference factor for selecting the initial prediction model.

[0060] Building area density refers to the density of buildings in an area, that is, the number of buildings per unit area. Areas with higher density usually mean more buildings and higher population density. In such areas, due to the concentration of buildings, load-intensive areas may be formed, and the power load is large. The power demand in these areas may be high, thereby increasing the load on the power transmission line and further increasing the power loss of the line. Therefore, when the building area density of the power grid area to be tested is different, it needs to be used as an important reference factor for selecting the initial prediction model.

[0061] By selecting the population density, building usage type, and building area density of the power grid area to be tested, the living environment problems of the power grid area to be tested can be basically determined. The reason for selecting population density, building usage type, and building area density is that compared with other living environment data, these living environment data can basically determine the living environment problems of the power grid area to be tested in general, and are highly comprehensive and representative, which helps to more accurately reference and select the initial prediction model.

[0062] The initial prediction model is obtained by matching the regional geographic information data and living environment data with the line loss rate prediction model;

[0063] Calculating the area coefficient of the target area includes:

[0064] The regional coefficient of the target area is obtained by weighted summing the altitude slope change, precipitation difference, soil thermal conductivity, population density, building use type, and building area density.

[0065] Matching methods include:

[0066] The regional coefficient is calculated with the regional coefficient of the preset line loss rate prediction model to obtain the absolute difference between the regional coefficient and the regional coefficient of the preset line loss rate prediction model. When the absolute difference is less than the preset threshold, the line loss rate prediction model corresponding to the regional coefficient of the preset line loss rate prediction model is used as the initial prediction model.

[0067] By analyzing the combined impact of geographic information data and living environment data, the initial prediction model of the target area can be obtained from the power grid monitoring server, reducing the frequent establishment of prediction models and reducing resource and time costs.

[0068] Example 3

[0069] Extracting the power load duration coefficient of the target sub-area includes:

[0070] Get the total duration of power overload in the sub-area within time H and mark it as R s ;

[0071] Get the total duration of power overload Rs The total time period elapsed is marked as R d ;

[0072] Get the total time period R d The total duration of the internal power overload is marked as R. h ;

[0073] Calculate the power load duration coefficient, the calculation expression is Lk=R s / R h ×R s .

[0074] For example, suppose there are two sub-areas, named A and B. Within 24 hours, the total duration of power overload in these two sub-areas is counted. Suppose within 24 hours, the total duration of power overload in both sub-areas is 5 (R s ) hours, but in area A, the total duration of power overload from the beginning to the end is from 9 am to 6 pm, and during this period, the number of power overloads experienced is 3 (R h ) times, indicating that between 9:00 a.m. and 3:00 p.m. d ) hours, the total power overload duration in area A is 5 hours, and these 5 hours are the accumulation of 3 power overloads; in area B, the total power overload duration is from the beginning time to the end time from 10 am to 3 pm, and during this period, the power overload occurred once, which means that in the 5 hours from 10 am to 3 pm, the power in area B has been overloaded, and the duration is 5 hours at a time; in this case, the line loss rate in area B may be higher, because area B has experienced a long power overload in a relatively short period of time (5 hours), which may lead to problems such as power line overload, increased line loss, and equipment overheating. In contrast, although the total duration of power overload in area A is also 5 hours in the same 24 hours, these 5 hours are divided into 3 times, which means that the load is relatively more dispersed and the power grid system is more able to withstand such load fluctuations, so the line loss rate may be relatively low. In addition, for area A, since it experiences a larger power load, it can have time to buffer after each load. Due to the buffer time, area A can better adjust and adapt to the changes in power load after each load, which helps the power grid system to remain stable. Compared with area B, the power grid system in area A may be more flexible and able to manage power load more effectively, thereby reducing the risk of line loss rate.

[0075] Through this calculation method, the degree of power overload can be judged more accurately, and the degree of power overload in different sub-areas can be obtained in detail. When predicting the line loss rate of different sub-areas, the weight ratio of the input layer data of the target prediction model can be accurately adjusted.

[0076] Example 4

[0077] The deviation coefficient of humidity around the device for extracting the target sub-area includes:

[0078] Get the total number of devices G and the actual humidity around each device, marked as S g , g is the number of the actual humidity around the device, g∈[0,G]; get the preset range of the ambient humidity that can be tolerated for each device and mark it as [S ga , S gb ], the actual humidity S around each device g At the corresponding preset tolerable ambient humidity [S ga , S gb ] are marked as W q ,q is S g In [S ga , S gb ], q∈[0,W], W∈[0,G]; W represents S g In [S ga , S gb ], and W is an integer, calculate the humidity deviation coefficient around the equipment, and the calculation expression is: Pe=ln(W / G+1); where Pe is the humidity deviation coefficient around the equipment.

[0079] Through this calculation method, the difference between the actual humidity around the power equipment and the preset tolerable ambient humidity corresponding to the equipment can be judged more accurately, and the weight ratio of the input layer data of the initial prediction model can be adjusted more accurately, making the final line loss rate prediction result more accurate.

[0080] Example 5

[0081] Extracting the equipment damage frequency coefficient of the target sub-area includes:

[0082] Collect the running time and failure number of the equipment, and organize the running time T and failure number N of the equipment into the form of feature vector x;

[0083] Calculate the Euclidean distance between the device to be predicted and other devices in the data set;

[0084] Select the appropriate K value based on the cross-validation method;

[0085] For each device to be predicted, select its K nearest neighbors;

[0086] The damage frequency coefficient of the equipment to be predicted is estimated by averaging the number of failures of the selected K nearest neighbors;

[0087] The expression to be calculated is:

[0088] Where Vd represents the equipment damage frequency coefficient, and Ni represents the number of failures of the i-th nearest neighbor.

[0089] Through this calculation method, the equipment damage frequency can be judged more accurately. The difference in equipment damage frequency will ultimately affect the accuracy of line loss rate prediction. The weight ratio of the input layer data of the initial prediction model can be more accurately adjusted according to the different equipment damage frequency coefficients, making the final line loss rate prediction result more accurate.

[0090] Example 6

[0091] The coefficients that need to be modified based on the correlation coefficient include:

[0092] The power load duration coefficient, the humidity deviation coefficient around the equipment, and the equipment damage frequency coefficient are used to construct a data evaluation model. The model calculation expression is:

[0093]

[0094] In the formula, Flg is the coefficient that can be modified, a 1 、a 2 、a 3 is the proportional coefficient of the power load duration coefficient, the humidity deviation coefficient around the equipment, and the equipment damage frequency coefficient, and a 1 、a 2 、a 3 Both are greater than 0.

[0095] The power load duration coefficient, the humidity deviation coefficient around the equipment, and the equipment damage frequency coefficient are weighted and summed to obtain a more accurate number to evaluate the line loss in the target area, and then more accurately adjust the weight ratio of the input layer data of the initial prediction model.

[0096] Example 7

[0097] Comparing the coefficient to be modified with the preset threshold, determining the weight ratio of the data in the input layer of the initial prediction model to obtain the target prediction model of the target sub-region includes:

[0098] When the required modification coefficient is less than the reference threshold, the initial prediction model is used as the target prediction model of the target sub-region, and the line loss rate of the region is predicted by the target prediction model;

[0099] When the modifiable coefficient is not less than the reference threshold, the corresponding preset prediction model input layer data weight ratio correspondence table is queried according to the specific numerical value of the modifiable coefficient, and the weight ratio of the input layer data of the initial prediction model is adjusted accordingly to obtain the target prediction model, and the target prediction model is used to predict the line loss rate of the area.

[0100] In this way, the line loss rate of the power grid area to be tested can be more accurately predicted. By dividing the target area into different sub-areas and then adjusting the weight ratio of the input layer in the initial prediction model to predict the line loss rate of each sub-area, the prediction model can be more in line with the actual situation of the sub-area, thereby improving the accuracy and reliability of the line loss rate prediction.

[0101] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A line loss rate prediction method based on big data, It is characterized in that The method is applied to a power grid monitoring server, the power grid monitoring server monitors a target power grid area, the target power grid area includes multiple power grid areas to be measured, the power grid monitoring server stores multiple line loss rate prediction models, and the method includes: Obtaining geographic information data and living environment data of a target area, matching the geographic information data and living environment data of the area in a plurality of line loss rate prediction models, and obtaining an initial prediction model; the target area is any one of a plurality of power grid areas to be tested; The target area is divided into multiple sub-areas, and the power load duration coefficient, the humidity deviation coefficient around the equipment and the equipment damage frequency coefficient of the target sub-area are extracted as correlation coefficients; the target sub-area is any one of the multiple sub-areas; Obtaining a coefficient that may be modified according to the correlation coefficient, and comparing the coefficient that may be modified with a preset threshold value, determining a weight ratio of data in the input layer of the initial prediction model to obtain a target prediction model for the target sub-region; Calculating the line loss rate of the target sub-area according to the target prediction model; Obtaining geographic information data and living environment data of a target area, matching the geographic information data and living environment data of the area in a plurality of line loss rate prediction models, and obtaining an initial prediction model includes: Acquire the altitude slope change, precipitation difference, and soil thermal conductivity of the target area as the geographic information data; Acquire the population density, building usage type, and building area density of the target area as the living environment data; Performing weighted summation on the geographic information data and the living environment data to obtain a regional coefficient of the target area; Calculating an absolute difference between the regional coefficient and a regional coefficient of a preset line loss rate prediction model, and when the absolute difference is less than a preset threshold, using the line loss rate prediction model corresponding to the regional coefficient of the preset line loss rate prediction model as an initial prediction model; Extracting the power load duration coefficient of the target sub-area includes: Get the sub-area The total duration of power overload during the time period, marked as ; Get the total duration of the power overload The total time period elapsed is marked as ; Get the total time period The total duration of power overload is recorded as ; Calculate the power load duration coefficient, the calculation expression is: ; The deviation coefficient of humidity around the device for extracting the target sub-area includes: Get the total number of devices , and obtain the actual humidity around each device and mark it as , The actual humidity around the device. ; Get the preset range of ambient humidity that each device can tolerate and mark it as , the actual humidity around each device The ambient humidity is within the preset acceptable range. Mark the device outside the , for In The number of the external device, ; express In The number of devices outside is an integer, and the humidity deviation coefficient around the equipment is calculated. The calculation expression is: ; In the formula, is the humidity deviation coefficient around the equipment; Extracting the equipment damage frequency coefficient of the target sub-area includes: Collect equipment operating hours and failure counts; Calculate the Euclidean distance between the device to be predicted and other devices in the data set; Select the appropriate value; For each device to be predicted, select the closest Neighbors; By selecting The average number of failures of the nearest neighbors is used to estimate the damage frequency coefficient of the equipment to be predicted; The expression to be calculated is: ; In the formula, represents the equipment damage frequency coefficient, Indicates The number of failures of the nearest neighbors; The coefficients that may be modified according to the correlation coefficient include: The power load duration coefficient, the humidity deviation coefficient around the equipment, and the equipment damage frequency coefficient are used to construct a data evaluation model, and the model calculation expression is: In the formula, To modify the coefficient, and They represent the power load duration coefficient, the humidity deviation coefficient around the equipment, and the equipment damage frequency coefficient. is the proportional coefficient of the power load duration coefficient, the humidity deviation coefficient around the equipment, and the equipment damage frequency coefficient, and All are greater than 0; Comparing the coefficient to be modified with a preset threshold value, determining the weight ratio of the data in the input layer of the initial prediction model to obtain the target prediction model of the target sub-region includes: When the required modification coefficient is less than a reference threshold, using the initial prediction model as a target prediction model for the target sub-region, and predicting the line loss rate of the region by using the target prediction model; When the modifiable coefficient is not less than a reference threshold, the corresponding preset prediction model input layer data weight ratio correspondence table is queried according to the specific numerical value of the modifiable coefficient, the weight ratio of the input layer data of the initial prediction model is adjusted accordingly to obtain the target prediction model, and the target prediction model is used to predict the line loss rate of the area.