Power equipment fire prediction method and device, equipment and storage medium

Abstract

The embodiment of the invention discloses a power equipment fire prediction method and device, equipment and a storage medium, and the method comprises the steps: training a preset XGBoost model according to historical fire case data, and obtaining an initial fire prediction model; a data set of the power equipment is collected, the data set comprises a plurality of training samples, and each training sample comprises historical environment data corresponding to any time point; adjusting a training weight in the initial fire prediction model according to the historical environment data to obtain an adjusted training weight; applying the adjusted training weight to the initial fire prediction model to obtain an adjusted target fire prediction model; and inputting the environment data and the power data of the power equipment, which are collected in real time, into the target fire prediction model to obtain a fire prediction result output by the target fire prediction model. By optimizing the fire prediction model, the fire prediction accuracy of the power equipment is improved, and the disaster prevention and reduction capability and the operation and maintenance decision level of the power grid are improved.

Description

Technical Field

[0001] The present invention relates to the field of fire prediction technology, and in particular to a method and device, equipment and storage medium for predicting fire in power equipment. Background Art

[0002] In recent years, my country's economy has developed rapidly, people's living standards have continued to improve, and higher requirements have been placed on the safety and reliability of power supply. In particular, the risk of fire in tourist areas and industrial parks is extremely high. During the peak tourist season, the number of tourists has skyrocketed, the flow of people is dense, and the demand for energy has increased. Overload of power equipment and lines has occurred frequently, and there are relatively higher fire hazards than in other areas. The fire risk in industrial parks is more prominent and more dangerous. Therefore, it is necessary to provide early warning of fires.

[0003] Existing fire monitoring methods for power equipment and lines have problems such as late reporting, missed reporting and false reporting of electrical fires, which greatly endangers the safe and stable operation of transmission lines and nearby equipment. Summary of the invention

[0004] Based on this, it is necessary to propose a method and device, equipment and storage medium for predicting fires in power equipment to improve the accuracy of predicting the probability of fire occurrence, so as to timely discover fire hazards and reduce the probability of fire in power equipment.

[0005] To achieve the above-mentioned purpose, the first aspect of the present application provides a method for predicting fire in electric power equipment, the method comprising:

[0006] Obtain historical fire case data of power equipment, and train a preset XGBoost model based on the historical fire case data to obtain an initial fire prediction model;

[0007] Collecting a data set of the electric power equipment, wherein the data set includes a plurality of training samples, and each training sample includes historical environmental data corresponding to any time point;

[0008] Adjusting the training weights in the initial fire prediction model according to the historical environmental data to obtain adjusted training weights;

[0009] Applying the adjusted training weights to the initial fire prediction model to obtain an adjusted target fire prediction model;

[0010] The environmental data and power data of the electric power equipment collected in real time are input into the target fire prediction model to obtain the fire prediction result output by the target fire prediction model.

[0011] Furthermore, the training sample also includes historical power data of the power equipment under the historical environment data, and the method further includes:

[0012] Performing model training based on the historical environmental data and the historical power data to obtain a threshold prediction model, wherein the threshold prediction model includes a relationship between the environmental data and the power data;

[0013] Inputting the environmental data of the electric power equipment collected in real time into the threshold prediction model to obtain the target threshold of the electric power data output by the threshold prediction model;

[0014] If the power data of the power equipment collected in real time is greater than the target threshold, a protection action is taken on the power equipment and an alarm signal is issued.

[0015] Furthermore, the historical environment data includes environment data types and historical environment parameters corresponding to each environment data type, and the historical power data includes power data types and historical power parameters corresponding to each power data type;

[0016] The performing model training according to the historical environmental data and the historical power data to obtain a threshold prediction model specifically includes:

[0017] Performing a correlation analysis based on the historical environmental parameters and the historical power parameters, and calculating the correlation coefficient between each of the power data types and each of the environmental data types;

[0018] Determine a target environment data type having a maximum correlation coefficient with the target power data type, wherein the target power data type is any one of all power data types;

[0019] Model training is performed according to historical environmental parameters corresponding to the target environmental data type and historical power parameters corresponding to the target power data type to obtain a target threshold prediction model.

[0020] Furthermore, the model training is performed according to the historical environmental parameters corresponding to the target environmental data type and the historical power parameters corresponding to the target power data type to obtain the target threshold prediction model, specifically including:

[0021] Performing model training according to the historical environmental parameters corresponding to the target environmental data type and the historical power parameters corresponding to the target power data type to obtain an initial threshold prediction model;

[0022] The initial threshold prediction model is modified according to the correlation coefficient between the target environment data type and the target power data type to obtain a modified target threshold prediction model.

[0023] Furthermore, the target threshold prediction model is:

[0024]

[0025] Y′=Y×(1+Δ)*r i

[0026] Wherein, Δ is the rate of change of the environmental parameter corresponding to the target environmental data type relative to the standard value, X is the environmental parameter corresponding to the target environmental data type collected in real time, is the average value of the historical environmental parameters corresponding to the target environmental data type, Y is the real-time value output by the initial threshold prediction model, Y′ is the target threshold of the target power data type output by the target threshold prediction model, r i is a correlation coefficient between the target environment data type and the target power data type.

[0027] Furthermore, the adjusting the training weights in the initial fire prediction model according to the historical environmental data to obtain the adjusted training weights specifically includes:

[0028] Comparing the historical environmental parameters corresponding to the target environmental data type with the preset environmental parameter thresholds to determine the environmental gain factor;

[0029] The training weights in the initial fire prediction model are adjusted according to the correlation coefficient between the target power data type and the target environment data type, and the environment gain factor to obtain adjusted training weights.

[0030] Furthermore, the adjusted training weight is obtained by the following formula:

[0031] Δ w =W original *f(W)*r

[0032]

[0033] In the formula, Δ w is the adjusted training weight, W original is the initial training weight, r is the correlation coefficient between the target power data type and the target environment data type, and f(W) is the environment gain factor.

[0034] To achieve the above-mentioned purpose, the second aspect of the present application provides a fire prediction device for electric power equipment, the device comprising a data acquisition unit, a weight adjustment unit and a disaster prediction unit;

[0035] The data acquisition unit is used to obtain historical fire case data of power equipment, and train a preset XGBoost model based on the historical fire case data to obtain an initial fire prediction model;

[0036] Collecting a data set of the electric power equipment, wherein the data set includes a plurality of training samples, and each training sample includes historical environmental data corresponding to any time point;

[0037] The weight adjustment unit is used to adjust the training weights in the initial fire prediction model according to the historical environmental data to obtain adjusted training weights;

[0038] Applying the adjusted training weights to the initial fire prediction model to obtain an adjusted target fire prediction model;

[0039] The disaster prediction unit is used to input the environmental data and power data of the power equipment collected in real time into the target fire prediction model to obtain the fire prediction result output by the target fire prediction model.

[0040] To achieve the above-mentioned purpose, the third aspect of the present application provides a computer-readable storage medium storing a computer program, characterized in that when the computer program is executed by a processor, the processor executes the steps of the method described in the first aspect.

[0041] To achieve the above-mentioned purpose, the fourth aspect of the present application provides a computer device, including a memory and a processor, characterized in that the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method described in the first aspect.

[0042] The embodiments of the present invention have the following beneficial effects:

[0043] The embodiment of the present invention proposes a method for predicting fires in power equipment, the method comprising: obtaining historical fire case data of power equipment, and training a preset XGBoost model based on the historical fire case data to obtain an initial fire prediction model; collecting a data set of power equipment, wherein the data set contains a number of training samples, and each training sample contains historical environmental data corresponding to any time point; adjusting the training weights in the initial fire prediction model based on the historical environmental data to obtain adjusted training weights; applying the adjusted training weights to the initial fire prediction model to obtain an adjusted target fire prediction model; inputting the environmental data and power data of the power equipment collected in real time into the target fire prediction model to obtain a fire prediction result output by the target fire prediction model. By optimizing the fire prediction model, the accuracy of fire prediction of power equipment is improved, which is conducive to improving the disaster prevention and mitigation capabilities and operation and maintenance decision-making level of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0045] in:

[0046] Figure 1 A schematic diagram of a flow chart of a method for predicting fire in electric power equipment according to an embodiment of the present invention;

[0047] Figure 2 A schematic diagram of a flow chart of a threshold determination method according to an embodiment of the present invention;

[0048] Figure 3 A structural block diagram of a fire prediction device for electric power equipment according to an embodiment of the present invention;

[0049] Figure 4 1 is a diagram showing the internal structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] 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. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Existing fire monitoring methods for power equipment and lines generally adopt manual inspections or drone inspections, or single temperature, smoke and other monitoring methods. They have limitations such as small monitoring range, high labor costs, and high experience requirements. They cannot meet large-scale monitoring needs. In addition, the monitoring algorithm does not have the function of learning to adapt to the environment, timely updating model parameters and judgment thresholds, which easily leads to late reporting, missed reporting and false reporting of electrical fires, which greatly endangers the safe and stable operation of transmission lines and nearby equipment.

[0052] Based on this, the present invention proposes a method for predicting fires in power equipment, which is used to improve the accuracy of fire prediction, and is conducive to improving the disaster prevention and mitigation capabilities and operation and maintenance decision-making level of the power grid. Figure 1 , Figure 1 The following is a flow chart of a method for predicting fire in electric power equipment according to an embodiment of the present invention. The method includes:

[0053] Step 110: Obtain historical fire case data of power equipment, and train a preset XGBoost model based on the historical fire case data to obtain an initial fire prediction model.

[0054] In the embodiment of the present invention, historical fire cases of power equipment are collected, and relevant data when the fire occurs is extracted from the historical fire cases, which may include data such as the time, location, scale, environmental parameters and power parameters of the fire. The prediction model is trained using the historical fire case data, and the decision logic of the prediction algorithm is adjusted to obtain an initial fire prediction model, which can be used to predict the fire of power equipment and obtain the probability of fire in the power equipment.

[0055] Step 120: Collect a data set of power equipment, wherein the data set includes a number of training samples, and each training sample includes historical environmental data corresponding to any time point.

[0056] An embodiment of the present invention provides a fire prediction device for electric power equipment. The device includes a data acquisition unit. The data acquisition unit can be used to collect data sets of the electric power equipment.

[0057] Specifically, the data acquisition unit at least includes current sensors, voltage sensors, optical fiber sensors, temperature and humidity sensors and other devices deployed on power equipment, as well as video surveillance modules and smoke detection modules and other devices deployed on transmission towers; wherein the video surveillance module and the smoke detection module are both connected to a power supply module, and the power supply module includes: a battery, a solar energy module and / or a wind energy module, the solar energy module and the wind energy module are both electrically connected to the battery, and the battery is also electrically connected to the video surveillance module and the smoke detection module, respectively, and the battery supplies power to the video surveillance module and the smoke detection module.

[0058] Current sensors, voltage sensors, optical fiber sensors and other devices are used to monitor the current, voltage, temperature and other data inside the power equipment; temperature and humidity sensors are used to collect environmental temperature and humidity data around the power equipment; the video monitoring module is used to capture real-time dynamic images of the area where the power equipment is located; the smoke detection module is used to obtain the smoke sensor data values ​​of the area where the power equipment is located.

[0059] In one embodiment of the present invention, the historical environmental data is environmental data of the power equipment collected by each device in the data collection unit within a certain period of time in the past.

[0060] In an embodiment of the present invention, the environmental data at least includes the temperature and humidity of the environment in which the power equipment is located, the dynamic image and smoke sensor data of the area in which the power equipment is located, etc. Therefore, the environmental parameters of the power equipment at different time points can be obtained through data acquisition devices such as temperature and humidity sensors, video monitoring modules and smoke detection modules in the data acquisition unit. The temperature and humidity, dynamic image and smoke sensor data at the same time point are used as a training sample, that is, the environmental data such as temperature and humidity, dynamic breast and smoke sensor data collected at the same time point are placed in one training sample. The collection of training samples obtained at different time points is a data set.

[0061] Step 130: Adjust the training weights in the initial fire prediction model according to the historical environmental data to obtain adjusted training weights.

[0062] In the embodiment of the present invention, the decision logic of the initial fire prediction model is adjusted by the environmental data corresponding to different time points. Since seasonal changes will affect the power data of power equipment, the initial fire prediction model is adjusted by the environmental data, so that the prediction results of the adjusted fire prediction model are more accurate.

[0063] Specifically, historical environmental data is used to adjust the training weights in the initial fire prediction model. The adjusted training weights adapt to changes in the environment where the power equipment is located, and take into account the influence of environmental factors, so that the prediction results of the trained fire prediction model are more accurate.

[0064] Step 140: Apply the adjusted training weights to the initial fire prediction model to obtain an adjusted target fire prediction model.

[0065] In an embodiment of the present invention, the adjusted training weights are applied to the initial fire prediction model for training to obtain a trained target fire prediction model, so that the fire prediction model can adapt to environmental changes and improve the accuracy of the prediction results.

[0066] Step 150: Input the environmental data and power data of the power equipment collected in real time into the target fire prediction model to obtain the fire prediction result output by the target fire prediction model.

[0067] After obtaining the target fire prediction model, the environmental data and power data of the power equipment can be monitored in real time, and the collected real-time environmental data and power data can be input into the target fire prediction model. The target fire prediction model outputs the probability of a fire in the power equipment under the input environmental data and power data, so as to judge whether a fire has occurred in the power equipment based on the fire probability.

[0068] The electric power equipment fire prediction method proposed in the present invention takes into account the impact of environmental factors on fire, and adjusts the initial fire prediction model through environmental data, so that the adjusted target fire prediction model can adapt to environmental changes at different times, and the prediction results of the adjusted target fire prediction model are more accurate.

[0069] Since there is a close connection between the power data and environmental data of power equipment, the environmental data and power data can be analyzed to determine the abnormal values ​​of the power data under specific environmental data, so as to judge the power data monitored in real time, determine whether there are abnormalities, and discover abnormal conditions of power equipment in time.

[0070] For details, please refer to Figure 2 , Figure 2 This is a flow chart of the threshold determination method proposed in an embodiment of the present invention. In the embodiment of the present invention, the training sample also includes historical power data of the power equipment. It can be understood that the historical power data and the historical environmental data appear in pairs, that is, each training sample includes environmental data and power data measured at the same time. The historical power data includes at least parameters such as the current value, voltage value, and temperature value of the power equipment. In one embodiment of the present invention, the power data can be obtained through data acquisition devices such as current sensors, voltage sensors, and optical fiber sensors in the data acquisition unit to obtain historical power data of the power equipment at different time points.

[0071] Based on this, the threshold determination method specifically includes:

[0072] Step 160: Perform model training based on historical environmental data and historical power data to obtain a threshold prediction model, wherein the threshold prediction model includes a relationship between environmental data and power data.

[0073] By training the model of power data under different environmental data and analyzing the connection between environmental data and power data, a threshold prediction model is obtained, which can predict the threshold size of power data under different environmental data. For example, after the current environmental data is input into the threshold prediction model, the threshold prediction model outputs the threshold of the power data. It can be understood that the threshold is the criterion for judging whether the power equipment has failed, that is, if the power data collected under the current environmental data exceeds the threshold, it means that the power equipment may fail or there is a risk of fire. If it does not exceed the threshold, it means that the power equipment is operating normally without any failure.

[0074] In order to improve the prediction efficiency and accuracy of the fire prediction model, the embodiment of the present invention also pre-processes the environmental data and uses the processed environmental data to adjust the initial fire prediction model. In one embodiment of the present invention, there are multiple types of environmental data and power data, and each type has a corresponding parameter value, that is, the historical environmental data contains environmental data types and historical environmental parameters corresponding to each environmental data type, and the historical power data contains power data types and historical power parameters corresponding to each power data type.

[0075] In order to simplify the training process and improve the accuracy of training, the embodiment of the present invention proposes to analyze the correlation between various types of environmental parameters in the environmental data and various power parameters in the power data to determine the environmental parameters that have the greatest impact on each power parameter. Based on this, model training is performed according to historical environmental data and historical power data to obtain a threshold prediction model, which specifically includes:

[0076] Step 161. Perform correlation analysis based on historical environmental parameters and historical power parameters, and calculate the correlation coefficient between each power data type and each environmental data type.

[0077] In an embodiment of the present invention, the environmental data type may include: temperature, humidity, dynamic image and smoke sensor data, etc., then the environmental parameters corresponding to the environmental data type are temperature observation values, humidity observation values, dynamic image observation values ​​and smoke observation values, etc.; the power data type may include: current, voltage and temperature, etc., then the power parameters corresponding to the power data type are current observation values, voltage observation values ​​and temperature observation values, etc.

[0078] The correlation between each environmental data type and each power data type is calculated based on the historical environmental parameters and the historical power parameters. For example, the correlation between the temperature observation value and the current observation value is calculated to obtain the correlation between the temperature and the current.

[0079] In the embodiment of the present invention, the correlation between each environment data type and the power data type can be analyzed based on the Pearson correlation coefficient, and the correlation between the environment data type and the power data type can be determined by calculating the correlation coefficient between the two. It can be understood that the larger the correlation coefficient, the greater the impact of the environment data type on the power data type.

[0080] The correlation coefficient between the power data type and the environmental data type can be calculated by the following formula:

[0081]

[0082] Where r is the correlation coefficient between the environmental data type and the power data type; X i is the historical environmental parameter, Yi is the historical power parameter, is the average value of historical environmental parameters, is the average value of historical power parameters.

[0083] Step 162, determine the target environment data type with the largest correlation coefficient with the target power data type, wherein the target power data type is any one of all power data types.

[0084] In an embodiment of the present invention, the correlation coefficient calculation method in Step 161 can be used to determine the correlation coefficient between each environmental data type and each power data type. Retaining the environmental data type with the largest correlation coefficient with each power data type for subsequent analysis can improve analysis efficiency and reduce the impact of other irrelevant factors on the analysis results, making the final prediction value more accurate.

[0085] Step 163: Perform model training based on the historical environmental parameters corresponding to the target environmental data type and the historical power parameters corresponding to the target power data type to obtain a target threshold prediction model.

[0086] In an embodiment of the present invention, the environmental data type with the greatest correlation with the power data type is used for model training. Specifically, the historical environmental parameters corresponding to the target environmental data type and the historical power parameters corresponding to the target power data type are used for model training to obtain a target threshold prediction model. For example, if the environmental data parameter with the greatest correlation with voltage is temperature, the temperature observation value and the voltage observation value are selected for model training to obtain a voltage threshold prediction model, and the voltage threshold under different temperature reference values ​​can be predicted based on the voltage threshold prediction model.

[0087] In one embodiment of the present invention, model training is performed according to historical environmental parameters corresponding to the target environmental data type and historical power parameters corresponding to the target power data type to obtain a target threshold prediction model, specifically including:

[0088] Step 1631, perform model training based on the historical environmental parameters corresponding to the target environmental data type and the historical power parameters corresponding to the target power data type to obtain an initial threshold prediction model.

[0089] Step 1632, modify the initial threshold prediction model according to the correlation coefficient between the target environment data type and the target power data type to obtain a modified target threshold prediction model.

[0090] In one embodiment of the present invention, the target threshold prediction model is:

[0091]

[0092] Y′=Y×(1+Δ)*r i

[0093] Where Δ is the rate of change of the environmental parameter corresponding to the target environmental data type relative to the standard value, X is the environmental parameter corresponding to the target environmental data type collected in real time, is the average value of the historical environmental parameters corresponding to the target environmental data type, Y is the real-time value output by the initial threshold prediction model, Y′ is the target threshold of the target power data type output by the target threshold prediction model, r i is the correlation coefficient between the target environment data type and the target power data type.

[0094] By adjusting the decision logic of the initial threshold prediction model according to environmental changes and changing its judgment criteria for fire prediction, the prediction results of the revised target threshold prediction model are made more accurate.

[0095] Step 170: Input the environmental data of the electric power equipment collected in real time into the threshold prediction model to obtain the target threshold of the electric power data output by the threshold prediction model.

[0096] In an embodiment of the present invention, the environmental parameters of the electric power equipment collected in real time are input into a threshold prediction model to perform threshold prediction, and obtain target thresholds of the electric power parameters under different environmental parameters.

[0097] Step 180: If the power data of the power equipment collected in real time is greater than the target threshold, a protective action is taken on the power equipment and an alarm signal is issued.

[0098] In an embodiment of the present invention, if the collected power parameters of the power equipment are not greater than the target threshold, it means that the power equipment does not have a temporary failure risk and no protection is required; if the collected power parameters of the power equipment are greater than the target threshold, it means that the power equipment has a failure risk and it is necessary to trigger a protection action to avoid the occurrence and deterioration of the failure, and send an alarm signal to prompt the staff to conduct inspections and repairs in a timely manner.

[0099] In one embodiment of the present invention, the training weights in the initial fire prediction model are adjusted according to the historical environmental data to obtain the adjusted training weights, which specifically include:

[0100] Step 131. Compare the historical environmental parameters corresponding to the target environmental data type with the preset environmental parameter thresholds to determine the environmental gain factor.

[0101] Step 132: According to the correlation coefficient between the target power data type and the target environmental data type, and the environmental gain factor, the training weights in the initial fire prediction model are adjusted to obtain adjusted training weights.

[0102] Since seasonal changes will affect the power data of power equipment, the initial fire prediction model is adjusted through environmental data to make the prediction results of the adjusted fire prediction model more accurate. Specifically, the adjusted training weight is obtained by the following formula:

[0103] Δ w =W original *f(W)*r

[0104]

[0105] In the formula, Δ w is the adjusted training weight, W original is the initial training weight, r is the correlation coefficient between the target power data type and the target environment data type, and f(W) is the environment gain factor.

[0106] The training weights are adjusted through historical environmental data so that the adjusted target fire prediction model can adapt to environmental changes and the prediction results can be more accurate.

[0107] The present invention also proposes a fire prediction device for electric power equipment, see Figure 3 , Figure 3 3 is a structural block diagram of a fire prediction device for electric power equipment in an embodiment of the present invention. The device includes a data acquisition unit 301 , a weight adjustment unit 302 and a disaster prediction unit 303 .

[0108] The data acquisition unit 301 is used to obtain historical fire case data of power equipment, and train a preset XGBoost model based on the historical fire case data to obtain an initial fire prediction model;

[0109] A data set of power equipment is collected, wherein the data set includes a number of training samples, and each training sample includes historical environmental data corresponding to any time point.

[0110] The weight adjustment unit 302 is used to adjust the training weights in the initial fire prediction model according to the historical environmental data to obtain adjusted training weights; and apply the adjusted training weights to the initial fire prediction model to obtain an adjusted target fire prediction model.

[0111] The disaster prediction unit 303 is used to input the environmental data and power data of the power equipment collected in real time into the target fire prediction model to obtain the fire prediction result output by the target fire prediction model.

[0112] The electric power equipment fire prediction device proposed in the embodiment of the present invention takes into account the impact of environmental factors on fire, and adjusts the initial fire prediction model through environmental data, so that the adjusted target fire prediction model can adapt to environmental changes at different times, and the prediction results of the adjusted target fire prediction model are more accurate.

[0113] Figure 4 FIG. 1 shows an internal structure diagram of a computer device in one embodiment of the present invention. The computer device may be a terminal or a system. Figure 4 As shown, the computer device includes a processor, a memory and a network interface connected via a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor may implement each step in the above method embodiment. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor may implement each step in the above method embodiment. Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0114] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes each step in the above method embodiment.

[0115] In one embodiment, a computer-readable storage medium is proposed, which stores a computer program. When the computer program is executed by a processor, the processor executes each step in the above method embodiment. A person of ordinary skill in the art can understand that all or part of the processes in the above embodiment method can be completed by instructing the relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium, and the program can include the processes of the embodiments of the above methods when executed. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0116] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0117] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for predicting fire in electric power equipment, characterized in that: The method comprises: Obtain historical fire case data of power equipment, and train a preset XGBoost model based on the historical fire case data to obtain an initial fire prediction model; Collecting a data set of the electric power equipment, wherein the data set includes a plurality of training samples, and each training sample includes historical environmental data corresponding to any time point; Adjusting the training weights in the initial fire prediction model according to the historical environmental data to obtain adjusted training weights; Applying the adjusted training weights to the initial fire prediction model to obtain an adjusted target fire prediction model; The environmental data and power data of the electric power equipment collected in real time are input into the target fire prediction model to obtain the fire prediction result output by the target fire prediction model.

2. The method according to claim 1, characterized in that The training sample also includes historical power data of the power equipment under the historical environment data, and the method further includes: Performing model training based on the historical environmental data and the historical power data to obtain a threshold prediction model, wherein the threshold prediction model includes a relationship between the environmental data and the power data; Inputting the environmental data of the electric power equipment collected in real time into the threshold prediction model to obtain the target threshold of the electric power data output by the threshold prediction model; If the power data of the power equipment collected in real time is greater than the target threshold, a protection action is taken on the power equipment and an alarm signal is issued.

3. The method according to claim 2, characterized in that The historical environment data includes environment data types and historical environment parameters corresponding to each environment data type, and the historical power data includes power data types and historical power parameters corresponding to each power data type; The performing model training according to the historical environmental data and the historical power data to obtain a threshold prediction model specifically includes: Performing a correlation analysis based on the historical environmental parameters and the historical power parameters, and calculating the correlation coefficient between each of the power data types and each of the environmental data types; Determine a target environment data type having a maximum correlation coefficient with the target power data type, wherein the target power data type is any one of all power data types; Model training is performed according to historical environmental parameters corresponding to the target environmental data type and historical power parameters corresponding to the target power data type to obtain a target threshold prediction model.

4. The method according to claim 3, characterized in that The performing model training according to the historical environmental parameters corresponding to the target environmental data type and the historical power parameters corresponding to the target power data type to obtain the target threshold prediction model specifically includes: Performing model training according to the historical environmental parameters corresponding to the target environmental data type and the historical power parameters corresponding to the target power data type to obtain an initial threshold prediction model; The initial threshold prediction model is modified according to the correlation coefficient between the target environment data type and the target power data type to obtain a modified target threshold prediction model.

5. The method according to claim 4, characterized in that The target threshold prediction model is: Y'=Y×(1+Δ)*r i Wherein, Δ is the rate of change of the environmental parameter corresponding to the target environmental data type relative to the standard value, X is the environmental parameter corresponding to the target environmental data type collected in real time, is the average value of the historical environmental parameters corresponding to the target environmental data type, Y is the real-time value output by the initial threshold prediction model, and Y ′ is the target threshold of the target power data type output by the target threshold prediction model, r i is a correlation coefficient between the target environment data type and the target power data type.

6. The method according to claim 3, characterized in that The adjusting the training weights in the initial fire prediction model according to the historical environmental data to obtain the adjusted training weights specifically includes: Comparing the historical environmental parameters corresponding to the target environmental data type with the preset environmental parameter thresholds to determine the environmental gain factor; The training weights in the initial fire prediction model are adjusted according to the correlation coefficient between the target power data type and the target environment data type, and the environment gain factor to obtain adjusted training weights.

7. The method according to claim 6, characterized in that The adjusted training weights are obtained by the following formula: Δ w =W original *f(W)*r In the formula, Δ w is the adjusted training weight, W original is the initial training weight, r is the correlation coefficient between the target power data type and the target environment data type, and f(W) is the environment gain factor.

8. A fire prediction device for electric power equipment, characterized in that: The device comprises a data acquisition unit, a weight adjustment unit and a disaster prediction unit; The data acquisition unit is used to obtain historical fire case data of power equipment, and train a preset XGBoost model based on the historical fire case data to obtain an initial fire prediction model; Collecting a data set of the electric power equipment, wherein the data set includes a plurality of training samples, and each training sample includes historical environmental data corresponding to any time point; The weight adjustment unit is used to adjust the training weights in the initial fire prediction model according to the historical environmental data to obtain adjusted training weights; Applying the adjusted training weights to the initial fire prediction model to obtain an adjusted target fire prediction model; The disaster prediction unit is used to input the environmental data and power data of the power equipment collected in real time into the target fire prediction model to obtain the fire prediction result output by the target fire prediction model.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 7.

10. A computer device comprising a memory and a processor, characterized in that: The memory stores a computer program, and when the computer program is executed by the processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 7.

Images