Power supply equipment temperature characteristic detection method and system

Abstract

The invention provides a power supply equipment temperature characteristic detection method and system. The method comprises the following steps: acquiring temperature data sets of power supply equipment under different working conditions; performing partial derivative calculation on the mathematical model, quantifying the contribution degree of each working condition to the temperature characteristic, and determining a main influence factor and a secondary influence factor; a time sequence analysis method is introduced, the temperature data set is analyzed, the dynamic change trend of the working conditions is captured, and nonlinear relation characteristics between the working conditions and temperature characteristics are identified; a self-adaptive filtering algorithm is adopted, changes of working conditions are tracked in real time, and parameters and structures in the mathematical model are dynamically adjusted, so that the mathematical model can adapt to the dynamic changes of the working conditions; and training and optimizing the adjusted mathematical model by using a support vector machine algorithm, thereby improving the prediction precision and generalization ability of the mathematical model.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method and system for detecting temperature characteristics of power supply equipment. Background Art

[0002] Traditional power supply equipment temperature characteristic detection methods usually test and analyze a single working condition, ignoring the mutual influence and coupling between different working conditions. Although this method is simple to operate, it cannot fully reflect the temperature characteristics of the equipment under actual complex working conditions. The influence mechanisms of different working conditions such as workload, ambient temperature, humidity, air flow velocity, etc. on the temperature characteristics of the equipment are different, and there may be nonlinear relationships and dynamic changes between the conditions. How to reasonably describe the influence of each working condition on the temperature characteristics in a mathematical model and quantify the contribution of each condition is a difficult problem that needs to be solved. In addition, how to track the changes in working conditions in real time, adaptively adjust the mathematical model to analyze and predict the temperature characteristics of the power supply equipment under the corresponding working conditions, and then guide the development of the detection work to ensure the pertinence and safety of the detection method is also an important technical problem. Therefore, it is necessary to explore a new detection method that can comprehensively consider multiple working conditions, quantify the contribution of different conditions, and adaptively adjust dynamically. Summary of the invention

[0003] The present invention provides a method for detecting temperature characteristics of a power supply device, which mainly includes:

[0004] Obtain a temperature data set of the power supply device under different working conditions, where the temperature data set includes workload, ambient temperature, humidity, and airflow velocity parameters;

[0005] A multivariate regression analysis method is used to establish a mathematical model between the temperature characteristics of the power supply equipment and various operating conditions based on the temperature data set. The mathematical model is used to describe the influence mechanism of various operating conditions on the temperature characteristics; by calculating the partial derivatives of the mathematical model, the contribution of each operating condition to the temperature characteristics is quantified, and the main influencing factors and secondary influencing factors are determined; the time series analysis method is introduced to analyze the temperature data set, capture the dynamic change trend of the operating conditions, and identify the nonlinear relationship characteristics between the operating conditions and the temperature characteristics; an adaptive filtering algorithm is used to track the changes in the operating conditions in real time, and dynamically adjust the parameters and structures in the mathematical model so that the mathematical model can adapt to the dynamic changes in the operating conditions; the support vector machine algorithm is used to train and optimize the adjusted mathematical model to improve the prediction accuracy and generalization ability of the mathematical model; based on the optimized mathematical model, the temperature characteristic curve of the power supply equipment under different operating conditions is generated, and the temperature characteristic curve is used to reflect the temperature changes of the equipment in the actual working environment.

[0006] The present invention provides a power supply equipment temperature characteristic detection system, which mainly includes: a data acquisition module, which is used to acquire a temperature data set of the power supply equipment under different working conditions;

[0007] A multiple regression analysis module is used to establish a mathematical model between the temperature characteristics of the power supply equipment and various working conditions based on the temperature data set by using the multiple regression analysis method, and the mathematical model is used to describe the influence mechanism of various working conditions on the temperature characteristics; a partial derivative calculation module is used to quantify the contribution of various working conditions to the temperature characteristics by calculating partial derivatives of the mathematical model, and determine the main influencing factors and secondary influencing factors; a time series analysis module is used to introduce a time series analysis method to analyze the temperature data set, capture the dynamic change trend of the working conditions, and identify the nonlinear relationship characteristics between the working conditions and the temperature characteristics; an adaptive filtering module is used to use an adaptive filtering algorithm to track the changes in the working conditions in real time, and dynamically adjust the parameters and structures in the mathematical model so that the mathematical model can adapt to the dynamic changes in the working conditions; a support vector machine optimization module is used to train and optimize the adjusted mathematical model using the support vector machine algorithm to improve the prediction accuracy and generalization ability of the mathematical model; a temperature characteristic curve generation module is used to generate a temperature characteristic curve of the power supply equipment under different working conditions according to the optimized mathematical model, and the temperature characteristic curve is used to reflect the temperature change of the equipment in the actual working environment. The technical solution provided by the embodiment of the present invention may include the following beneficial effects:

[0008] The present invention discloses a method for detecting the temperature characteristics of a power supply device. The method establishes a mathematical model of temperature characteristics and working conditions through multivariate regression analysis, and quantifies the contribution of each factor by using partial derivative calculation. Time series analysis is introduced to capture dynamic change trends and nonlinear relationships, and an adaptive filtering algorithm is used to adjust model parameters in real time. The support vector machine is further used to optimize the model to improve prediction accuracy and generalization ability. Finally, a temperature characteristic curve under different working conditions is generated to reflect the temperature changes in the actual working environment of the equipment. The present invention realizes accurate analysis and prediction of the temperature characteristics of the power supply device during the detection process, provides an important basis for the formulation of detection parameters of the power supply device and the determination of the range of key detection parameters, and effectively improves the pertinence and safety of power supply equipment detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 The present invention is a flow chart of a method for detecting temperature characteristics of a power supply device.

[0010] Figure 2 A schematic diagram of a method and system for detecting temperature characteristics of a power supply device according to the present invention.

[0011] Figure 3 It is another schematic diagram of a method and system for detecting temperature characteristics of a power supply device according to the present invention.

[0012] Figure 4 The present invention is a schematic diagram of the structure of a method and system for detecting temperature characteristics of power supply equipment. DETAILED DESCRIPTION

[0013] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the accompanying drawings and embodiments. The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are only used to explain the relevant inventions, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0014] like Figures 1 to 4 The present embodiment provides a method and system for detecting temperature characteristics of a power supply device, which may specifically include:

[0015] S101, such as Figure 2 As shown, a multivariate regression analysis method is used to establish a mathematical model between the temperature characteristics of the power supply equipment and various operating conditions based on the temperature data set. The mathematical model is used to describe the influence mechanism of various operating conditions on the temperature characteristics.

[0016] Obtain a temperature data set of power equipment containing multiple dimensions, preprocess the temperature data set, remove outliers and missing values, and standardize the data to make the data conform to the normal distribution; select the multiple regression analysis method according to the characteristics of the temperature data set. When the independent variables are linearly correlated, select multiple linear regression; when there is a nonlinear relationship between the independent variables and the dependent variable, select polynomial regression; when the number of independent variables exceeds the preset threshold, select stepwise regression; take the temperature characteristics as the dependent variable and each working condition as the independent variable, and establish the regression equation between the temperature characteristics and the working condition; use the least squares method to estimate the regression coefficient and obtain the parameter estimation value of the regression model; perform a significance test on the regression coefficient through the t test and eliminate insignificant independent variables. The F test is used to test the significance of the model as a whole, and the determination coefficient is calculated to evaluate the goodness of fit of the model. If the regression model passes the significance test and the goodness of fit is higher than the preset threshold, it is used as a mathematical model to describe the influence mechanism between temperature characteristics and operating conditions. The k-fold cross-validation method is used to randomly divide the temperature data set into k subsets, and one of the subsets is selected as the test set each time, and the remaining subsets are used as training sets. The validation process is repeated k times, and the mean square error and determination coefficient of each validation are calculated to evaluate the prediction performance and generalization ability of the mathematical model. According to the size and sign of the regression coefficient of each variable in the mathematical model, the degree and mode of influence of each operating condition on the temperature characteristics are analyzed, providing a quantitative basis for temperature control and optimization.

[0017] For example, the temperature data set of power supply equipment usually contains multiple dimensions, such as ambient temperature, load power, cooling wind speed, etc. In the preprocessing stage, outliers need to be removed first, for example, data that exceeds the normal operating temperature range of the equipment needs to be deleted. For missing values, mean filling or interpolation can be used for processing. Standardization can use the Z-score method to convert the data into a standard normal distribution with a mean of 0 and a standard deviation of 1. The choice of multiple regression analysis method depends on the data characteristics. When there is a linear correlation between the independent variables, multiple linear regression is a suitable choice. For example, analyze the impact of ambient temperature and load power on device temperature. If it is found that the device temperature and load power show a nonlinear relationship, consider using polynomial regression and introduce quadratic or cubic terms to describe this nonlinear relationship. In the case of many independent variables, such as considering factors such as ambient temperature, load power, cooling wind speed, working time, etc. at the same time, stepwise regression can effectively screen out the most influential variables. When establishing the regression equation, the device temperature is used as the dependent variable and each working condition is used as the independent variable. The least squares method is used to estimate the regression coefficients, and an equation such as T=β0+β1X1+β2X2+...+βnXn is obtained, where T is the equipment temperature and X1, X2, etc. are the operating conditions. The significance test of each regression coefficient is performed by t-test, and the independent variables with p values ​​greater than the significance level (such as 0.05) are eliminated. The F test is used to evaluate the significance of the entire model, and the determination coefficient R² is calculated to measure the goodness of fit of the model. Cross-validation is an effective method to evaluate the generalization ability of the model. For example, using 5-fold cross-validation, the data set is randomly divided into 5 parts, 4 parts are used as training sets each time, and 1 part is used as a test set, and repeated 5 times. The mean square error (MSE) and determination coefficient (R²) of each validation are calculated, and the average value is taken as the estimate of model performance. This method can effectively avoid overfitting problems and improve the reliability of the model. Analyze the impact of various operating conditions based on the regression coefficient. Assume that the equation T=30+0.8X1-0.5X2+0.3X3 is obtained, where X1 is the ambient temperature, X2 is the cooling wind speed, and X3 is the load power. It can be concluded that for every 1°C increase in ambient temperature, the device temperature rises by 0.8°C; for every 1m / s increase in cooling wind speed, the device temperature drops by 0.5°C; for every 1kW increase in load power, the device temperature rises by 0.3°C. These quantitative relationships provide direct guidance for device temperature control. For example, in a high temperature environment, the cooling wind speed can be appropriately increased to suppress the temperature rise.

[0018] S102, such as Figure 3 As shown, by calculating the partial derivative of the mathematical model, the contribution of each operating condition to the temperature characteristics is quantified, and the main and secondary influencing factors are determined.

[0019] Obtain the established mathematical model of temperature characteristics, and use the numerical differentiation method to calculate the partial derivative values ​​of each working condition in the mathematical model; sort the partial derivative values ​​from large to small according to the absolute value of the partial derivative values, determine the working conditions corresponding to the partial derivative values ​​greater than the preset threshold as the main influencing factors, and determine the working conditions corresponding to the partial derivative values ​​less than the preset threshold as the secondary influencing factors; for the working conditions corresponding to the main influencing factors, use the support vector machine algorithm to establish a nonlinear regression model of the main influencing factors and temperature characteristics; use the grid search and cross-validation methods to optimize the hyperparameters of the support vector machine algorithm to obtain the optimized nonlinear regression model; for the working conditions corresponding to the secondary influencing factors, use the decision tree classification algorithm to establish a classification model of the secondary influencing factors and temperature characteristics; use the information gain ratio as the standard for feature selection, prune the decision tree classification model, and obtain the optimized classification model; obtain real-time data of each working condition, and separate the main influencing factors into the main influencing factors. The real-time data of the main influencing factors are input into the optimized nonlinear regression model to obtain the regression prediction value of the temperature characteristics; the real-time data of the secondary influencing factors are input into the optimized classification model to obtain the classification judgment result of the temperature characteristics; the weighted average function is used to weightedly fuse the regression prediction value and the classification judgment result. During the fusion, the weight of the main influencing factors is set to 7, and the weight of the secondary influencing factors is set to 3, so as to obtain the temperature characteristic prediction value after combining various working conditions; the preset temperature characteristic qualified threshold range is obtained to determine whether the temperature characteristic prediction value falls within the qualified threshold range; if so, it is determined that the temperature characteristics under the current working condition combination meet the requirements; if not, it is determined that the temperature characteristics under the current working condition combination do not meet the requirements, and the simulated annealing algorithm is used to automatically adjust the corresponding working condition parameters; the deviation between the temperature characteristic prediction value and the qualified threshold range is used as the optimization objective function, and the power supply equipment working condition parameter combination that minimizes the deviation is found through iterative search, and it is determined as the optimal working condition.

[0020] Exemplarily, the numerical differentiation method is an effective means to calculate the partial derivatives of various working conditions. Taking the ambient temperature as an example, by changing the ambient temperature by ±1°C and observing the change in the equipment temperature, the partial derivative of the ambient temperature to the equipment temperature can be obtained. For other factors such as load power and cooling wind speed, similar methods are used to calculate partial derivatives. Sorting the absolute values ​​of the obtained partial derivatives may result in the following results: ambient temperature 0.8, load power 0.5, and cooling wind speed 0.3. Based on this, it can be determined that the ambient temperature and load power are the main influencing factors, and the cooling wind speed is the secondary influencing factor. For the main influencing factors, the support vector machine regression algorithm can effectively establish a nonlinear relationship model. Taking the ambient temperature as an example, the radial basis function (RBF) is selected as the kernel function, and the penalty parameter C and the kernel parameter γ are optimized through grid search. Assuming that the optimal parameter combination is C=10, γ=0.1, a nonlinear regression model of ambient temperature and equipment temperature can be obtained. In the cross-validation process, the 5-fold cross-validation method can be used to divide the data set into 5 parts, and 4 parts are used for training and 1 part for testing in turn, and the parameter combination with the smallest average error is taken. For secondary influencing factors such as cooling wind speed, the decision tree classification algorithm can be used to establish a classification model. Using the information gain ratio as a feature selection criterion can effectively avoid biased features with more values. For example, the cooling wind speed is divided into three levels: low, medium, and high, corresponding to different ranges of equipment temperature. Through pruning optimization, such as setting the minimum sample number threshold to 5, the risk of overfitting can be reduced and the generalization ability of the model can be improved. In MATLAB, the weighted average function can be used to fuse the results of different models. Assume that the optimized nonlinear regression model predicts that the equipment temperature is 45°C, and the optimized classification model judges that the temperature range is medium (40-50°C). Set the weight ratio to 7:3, and the predicted temperature after fusion is 45×0.7+45×0.3=45°C. This weighted fusion method can comprehensively consider the influence of primary and secondary factors and improve the prediction accuracy. The setting of the qualified threshold range of temperature characteristics can be based on expert experience and historical data. For example, the threshold range is set to 40-50°C. If the predicted temperature is 45°C, it is judged that the current operating condition combination meets the requirements; if the predicted temperature is 55°C, it is judged that it does not meet the requirements. The simulated annealing algorithm can be used to automatically adjust the parameters of the operating conditions. Taking the ambient temperature and load power as an example, the initial temperature is set to 100 and the cooling coefficient is 0.95. Each iteration randomly adjusts the ambient temperature by ±2°C and the load power by ±10%. If the new solution brings the temperature prediction value closer to the center of the threshold, the new solution is accepted; otherwise, the inferior solution is accepted with a certain probability (such as 5%) to avoid falling into the local optimum. Through multiple iterations, the optimal operating condition combination that makes the device temperature closest to 45°C can be found. This iterative optimization method can maximize the performance of the device while ensuring the temperature characteristics requirements.

[0021] S103. Introduce a time series analysis method to analyze the temperature data set, capture the dynamic change trend of the working conditions, and identify the nonlinear relationship characteristics between the working conditions and the temperature characteristics.

[0022] Acquire a temperature data set, divide the temperature data set according to different working conditions of the power supply equipment, and obtain temperature sub-data sets corresponding to different working conditions; detect and remove outliers and noise data for each temperature sub-data set, use the median filter method to preprocess the data, and obtain the preprocessed temperature sub-data set; perform time series analysis on each preprocessed temperature sub-data set, and use the autoregressive moving average model ARMA to establish a temperature time series model that reflects the temperature change trend under different working conditions; based on the ARMA model, use the wavelet analysis method to perform multi-scale decomposition on the temperature time series under each working condition, extract the temperature fluctuation characteristics under different frequency scales, and obtain the wavelet coefficients that reflect the temperature fluctuation characteristics under different working conditions; use the wavelet coefficients as sample features, and use the support vector machine algorithm to establish a nonlinear mapping relationship model between the working conditions and the temperature fluctuation characteristics; use the nonlinear mapping model to identify the temperature fluctuation characteristics under the working conditions of the real-time collected temperature data, and judge whether the temperature fluctuation under the current working conditions exceeds the preset normal range. If it exceeds, trigger a temperature abnormality warning.

[0023] For example, for a transformer, temperature sensors can be installed in different parts of the transformer, such as heat sink fins, windings, cores, etc., and the ambient temperature, load rate and other working conditions can be recorded at the same time. According to these working conditions, the data can be divided into multiple subsets, such as full load, light load, overload, etc. For the processing of outliers and noise, median filtering is an effective method. Suppose that at a certain point in time, the temperature of the heat sink fin suddenly appears to be significantly higher, such as 50°C during normal operation, and suddenly a reading of 80°C appears. Median filtering can effectively remove such sudden outliers and maintain the overall trend of the data. In time series analysis, the ARMA model can capture the dynamic characteristics of temperature changes. Taking the temperature of the heat sink fin as an example, it may be found that it has obvious periodicity and trend. The ARMA model can capture the persistence of temperature through the autoregressive term and capture short-term fluctuations through the moving average term, thereby accurately describing the change law of temperature over time. Wavelet analysis has advantages in processing non-stationary signals. For transformer temperature data, Daubechies4 wavelet can be selected for 3-layer decomposition. This allows the temperature signal to be decomposed into different frequency components, such as long-term trends, daily changes, and short-term fluctuations. This decomposition helps identify temperature variation characteristics on different time scales. The support vector machine algorithm performs well in establishing nonlinear mapping relationships. By using wavelet coefficients as input features, the support vector machine model can be trained to obtain a nonlinear mapping relationship model to predict the temperature fluctuation characteristics under different working conditions. For example, the nonlinear mapping relationship model may find that the high-frequency fluctuations in the heat sink fin temperature are more obvious under high load and high ambient temperature conditions. The real-time temperature anomaly warning system can be based on the prediction results of the nonlinear mapping relationship model. Assuming that the nonlinear mapping relationship model predicts that the heat sink fin temperature should fluctuate within the range of 45-65℃ under the current working conditions, but the actual temperature reaches 70℃, the system will trigger an early warning. This early warning mechanism can detect potential equipment anomalies or overloads in a timely manner.

[0024] S104. Adopting an adaptive filtering algorithm, tracking the changes of working conditions in real time, and dynamically adjusting the parameters and structures in the mathematical model, so that the mathematical model can adapt to the dynamic changes of working conditions.

[0025] Acquire real-time data reflecting the current working conditions, pre-process the real-time data to obtain pre-processed real-time data; use an adaptive Kalman filter algorithm to estimate and update the parameters of the mathematical model online and in real time based on the pre-processed real-time data; introduce a forgetting factor into the adaptive Kalman filter algorithm to give different weights to historical data; on the basis of adaptively adjusting the model parameters, change the topological structure of the mathematical model by adding or deleting components of the mathematical model; use the adaptively adjusted mathematical model parameters and structure as input, use the optimization algorithm to optimize the mathematical model in real time to obtain an optimized mathematical model; use the optimized mathematical model to process the input data under the current working conditions to obtain the predicted values ​​of key parameters; based on the predicted results output by the mathematical model, automatically generate an optimization control strategy to achieve optimal control under the working conditions; continuously monitor the key parameters reflecting the working conditions, and when the change range of the key parameters exceeds the preset threshold, trigger the adaptive Kalman filter algorithm to readjust the mathematical model.

[0026] Exemplarily, the adaptive Kalman filter algorithm is a dynamic estimation method that can continuously adjust model parameters according to real-time data. Taking the temperature monitoring of power transformers as an example, real-time operating data such as transformer load rate and ambient temperature can be obtained as input. In the data preprocessing stage, median filtering is used to remove outliers, such as correcting the sudden 80°C heat sink temperature to the normal range. The filtering algorithm dynamically adjusts the state estimation and error covariance matrix by comparing the difference between the predicted value and the measured value. The introduction of the forgetting factor can highlight the impact of new data, for example, giving the data of the last hour a weight of 0.9, while the weight of the data 12 hours ago is reduced to 0.5. This allows the model to respond more quickly to changes in operating conditions, such as a rapid increase in temperature caused by a sudden increase in load. Adaptive adjustment of the model structure is the key to further improving the adaptability of the model. For example, under light load conditions, only two factors, ambient temperature and load rate, may need to be considered; while under heavy load conditions, factors such as cooling efficiency and oil temperature distribution may need to be increased. By dynamically increasing or decreasing model components, the temperature change law under different working conditions can be more accurately described. Optimization algorithms such as the conjugate gradient descent method can further optimize model parameters on the basis of adaptive adjustment. According to the objective function of the optimization problem, a loss function is constructed to evaluate the degree of optimization of the model parameters. The model parameters are initialized, and appropriate hyperparameters such as learning rate and number of iterations are set to prepare for the conjugate gradient descent algorithm. In each iteration, the gradient vector of the loss function under the current parameters is calculated as the basis for the search direction. Using the conjugate gradient algorithm, a new conjugate search direction is determined based on the current gradient and the search direction of the previous step. Along the conjugate search direction, a linear search or other one-dimensional optimization method is used to find the step size that minimizes the loss function. Using the obtained step size, the model parameters are updated to obtain the starting point of a new round of iterations. The gradient search steps are repeated until the convergence condition is met or the preset number of iterations is reached to obtain the final optimized model parameters. Assuming that the model contains parameters such as thermal resistance and thermal capacitance, the optimization algorithm can quickly find the optimal parameter combination at each time step to minimize the error between the model output and the measured data. This real-time optimization ensures that the model can maintain high accuracy under different working conditions. Based on the optimized model, the transformer temperature in the future can be predicted. For example, it is predicted that at the current 90% load rate and 35°C ambient temperature, the heat sink fin temperature may reach 70°C after 2 hours. This prediction provides an important basis for formulating control strategies. Automatically generated control strategies may include adjusting the cooling system operating mode. If the predicted temperature is about to exceed the warning line, the system may recommend starting additional cooling fans or increasing the cooling fan speed. PID controller parameters can also be automatically adjusted based on the prediction results, such as increasing the proportional coefficient to speed up the response when the temperature rises faster. Continuous monitoring of key parameters is key to maintaining optimal performance of the model. For example, when the ambient temperature drops sharply by 5°C in half an hour, it may trigger a model readjustment.This mechanism ensures that the model can adapt to sudden environmental changes or load fluctuations in a timely manner, and always maintain accurate prediction and effective control of transformer temperature. In transformer temperature detection, this method can detect the temperature characteristics of the transformer under extreme load conditions while ensuring equipment safety, thereby increasing the coverage of temperature characteristic detection.

[0027] S105. Using a support vector machine algorithm, train and optimize the adjusted mathematical model to improve the prediction accuracy and generalization ability of the mathematical model.

[0028] Obtain a training data set, which includes multiple sample data and corresponding labels; use a support vector machine algorithm to train the training data set to obtain a preliminary support vector machine model; use a cross-validation method to evaluate the preliminary support vector machine model to obtain the prediction accuracy and generalization error of the preliminary support vector machine model on the validation set; if the prediction accuracy is less than a preset accuracy threshold or the generalization error is greater than a preset error threshold, perform the following steps: analyze the characteristics and sample quality of the training data set, remove noise data and redundant features therein, and obtain an optimized training data set; use a support vector machine algorithm to train the optimized training data set again to obtain an optimized support vector machine model; use a grid search method to determine the optimal hyperparameter combination of the support vector machine algorithm, the optimal hyperparameter combination is a hyperparameter combination with the smallest generalization error obtained by cross-validation for the optimized training data set;

[0029] Retrain the support vector machine model using the optimal hyperparameter combination to obtain the final support vector machine model; obtain a test data set, use the final support vector machine model to predict the test data set, and obtain the prediction accuracy and number of convergence rounds of the final support vector machine model on the test data set; if the prediction accuracy is greater than the preset accuracy threshold and the number of convergence rounds is less than the preset round threshold, apply the final support vector machine model to the actual business scenario for prediction; continuously monitor the online support vector machine model to obtain its prediction error and data distribution changes; if the prediction error exceeds the preset error threshold or the data distribution change exceeds the preset change threshold, obtain newly collected data, and use incremental training to update the final support vector machine model to obtain a new version of the online support vector machine model.

[0030] Exemplarily, first, a large amount of power supply equipment operation data is collected, including features such as load rate, ambient temperature, cooling efficiency, and the corresponding equipment temperature as labels. These data constitute a training set for preliminary training of the SVM model to obtain a preliminary support vector machine model. Cross-validation is an effective method to evaluate the performance of a preliminary support vector machine model. For example, a 5-fold cross-validation is used to divide the data set into 5 parts, 4 parts are used for training and 1 part is used for verification in turn, and the average result is finally taken. Assuming that the prediction accuracy of the preliminary support vector machine model on the verification set is 85% and the generalization error is 0.15, if the preset accuracy threshold is 90% and the error threshold is 0.1, further optimization is required. Data quality directly affects model performance. By analyzing the importance of features, it may be found that some features (such as equipment number) contribute less to the prediction and can be removed. In addition, methods such as box plots are used to identify outliers, such as removing temperature data that obviously deviates from the normal range, so as to obtain an optimized training set. Grid search is a commonly used hyperparameter optimization method. For SVM, the kernel function type, penalty coefficient C and kernel function parameter γ are mainly optimized. For example, the kernel function options are set to {linear, polynomial, RBF}, the range of C is [0.1, 1, 10, 100], and the range of γ is [0.001, 0.01, 0.1, 1]. By traversing all combinations, the hyperparameter combination with the smallest generalization error on the validation set is selected. After retraining the SVM model with the optimal hyperparameters, its performance is evaluated on an independent test set. Assume that the final SVM model achieves 95% prediction accuracy on the test set and converges within 50 rounds. If the preset accuracy threshold is 93% and the round threshold is 100 rounds, the model can be applied to actual business scenarios. In actual applications, it is crucial to continuously monitor the performance of the model. For example, the prediction error is calculated every day, and if the error exceeds 10% for a week in a row, or the KL divergence of the data distribution exceeds 0.5, the model update is triggered. Incremental training allows the model to adapt to new data patterns, such as changes in temperature characteristics caused by seasonal changes, thereby maintaining the prediction accuracy of the model. Through this iterative optimization and dynamic update method, the SVM model can continuously and accurately predict the temperature characteristics of power supply equipment.

[0031] S106. Generate a temperature characteristic curve of the power supply device under different working conditions according to the optimized mathematical model. The temperature characteristic curve is used to reflect the temperature change of the device in the actual working environment.

[0032] According to the optimized mathematical model, the working condition parameters of the power supply equipment are obtained, including the ambient temperature and load power; for each working condition, the temperature characteristic curves of multiple groups of equipment under different working conditions are calculated through the mathematical model; using MATLAB software, according to the generated temperature characteristic curve, a three-dimensional visualization model of the temperature change of the power supply equipment is constructed to intuitively display the temperature change trend. Using MATLAB software, according to the generated temperature characteristic curve, a three-dimensional visualization model of the temperature change of the power supply equipment is constructed to intuitively display the temperature change trend; using the support vector machine algorithm, with the working condition parameters as input and the corresponding temperature characteristic curve as output, a mapping relationship model between the two is trained and established; using the trained mapping relationship model, the temperature change trend of the power supply equipment under unknown working conditions is predicted; the rated operating temperature range of the power supply equipment is obtained, and the predicted temperature change trend is compared with it to determine whether the equipment has an over-temperature risk; if the temperature change trend exceeds the rated operating temperature range, an early warning message is issued to indicate that the equipment may have an over-temperature problem.

[0033] For example, the temperature characteristics of power supply equipment are crucial to its performance and life. In order to accurately predict and monitor the temperature of the equipment, it is necessary to first obtain the parameters of the operating conditions. These parameters usually include ambient temperature, load power, humidity, etc. For example, for a transformer, the following data may be collected: the ambient temperature range is -20℃ to 40℃, the load power is 0 to 100% rated power, and the relative humidity is 30% to 90%. Using the optimized mathematical model, the temperature characteristic curve can be calculated for different operating conditions. Assuming that the ambient temperature is 25℃ and the load power is 80% rated power, the curve of the transformer temperature over time may show the characteristics of first rising rapidly and then stabilizing, and finally stabilizing at around 65℃. Repeating this process can generate a large number of temperature characteristic curves until a preset threshold, such as 1000 curves, is reached. Next, use MATLAB software to build a three-dimensional visualization model. With ambient temperature and load power as the horizontal and vertical coordinates and device temperature as the Z axis, a three-dimensional surface graph can be obtained. This visualization method can directly show the impact of working conditions on equipment temperature, helping engineers quickly identify potential overheating risk areas so as to conduct targeted detection for the corresponding working conditions. The support vector machine (SVM) algorithm plays an important role in this scenario. During the training process, it may be found that some features have a more significant impact on temperature. For example, load power may have a greater impact on transformer temperature than humidity. By adjusting the kernel function (such as radial basis function RBF) and the penalty parameter C, the performance of the mapping relationship model can be optimized. After training, the mapping relationship model can be used to predict the temperature change trend under unknown working conditions. Assume that it is predicted that under certain extreme working conditions (such as ambient temperature 40°C and load power 95%), the transformer temperature may reach 85°C. Comparing this prediction result with the rated operating temperature range of the transformer (such as -25°C to 75°C), it is found that there is an overheating risk. At this time, the system will automatically issue an early warning message to prompt the inspector to take measures. The early warning information may include: current working conditions, predicted temperature, over-temperature range, etc. This early warning mechanism can help inspectors take preventive measures in advance, such as reducing loads and enhancing cooling, so as to avoid damage to power supply equipment during testing. This method not only monitors the temperature status of power supply equipment in real time, but also predicts possible problems in the future, which plays an important role in improving the safety of the detection process.

[0034] like Figure 4As shown, the embodiment of the present invention also provides a power supply equipment temperature characteristic detection system, which mainly includes: a multiple regression analysis module, which is used to adopt a multiple regression analysis method to establish a mathematical model between the temperature characteristics of the power supply equipment and various working conditions according to the temperature data set, and the mathematical model is used to describe the influence mechanism of various working conditions on the temperature characteristics; a partial derivative calculation module is used to quantify the contribution of various working conditions to the temperature characteristics by calculating partial derivatives of the mathematical model, and determine the main influencing factors and secondary influencing factors; a time series analysis module is used to introduce a time series analysis method to analyze the temperature data set, capture the dynamic change trend of the working conditions, and identify The nonlinear relationship characteristics between the working conditions and the temperature characteristics are as follows; the adaptive filtering module is used to adopt the adaptive filtering algorithm to track the changes of the working conditions in real time, dynamically adjust the parameters and structures in the mathematical model, so that the mathematical model can adapt to the dynamic changes of the working conditions; the support vector machine optimization module is used to train and optimize the adjusted mathematical model using the support vector machine algorithm to improve the prediction accuracy and generalization ability of the mathematical model; the temperature characteristic curve generation module is used to generate the temperature characteristic curve of the power supply equipment under different working conditions according to the optimized mathematical model, and the temperature characteristic curve is used to reflect the temperature change of the equipment in the actual working environment. Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the embodiments of the present application. In this way, if these modifications and variations of the embodiments of the present application belong to the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these changes and variations.

Claims

1. A method for detecting temperature characteristics of a power supply device, characterized in that: The method comprises: Obtain a temperature data set of the power supply device under different working conditions, where the temperature data set includes workload, ambient temperature, humidity, and airflow velocity parameters; Using multiple regression analysis method, a mathematical model between the temperature characteristics of power supply equipment and various working conditions is established according to the temperature data set. The mathematical model is used to describe the influence mechanism of various working conditions on the temperature characteristics. By calculating partial derivatives of the mathematical model, the contribution of each operating condition to the temperature characteristics is quantified, and the main and secondary influencing factors are determined; Introduce time series analysis methods to analyze temperature data sets, capture the dynamic change trend of working conditions, and identify the nonlinear relationship characteristics between working conditions and temperature characteristics; Adopting adaptive filtering algorithm, real-time tracking of changes in working conditions, dynamically adjusting parameters and structures in mathematical models, so that mathematical models can adapt to dynamic changes in working conditions; Use the support vector machine algorithm to train and optimize the adjusted mathematical model to improve the prediction accuracy and generalization ability of the mathematical model; According to the optimized mathematical model, the temperature characteristic curve of the power supply equipment under different working conditions is generated. The temperature characteristic curve is used to reflect the temperature change of the equipment in the actual working environment.

2. The method for detecting temperature characteristics of a power supply device according to claim 1, characterized in that: The multivariate regression analysis method is used to establish a mathematical model between the temperature characteristics of the power supply equipment and various working conditions based on the temperature data set. The mathematical model is used to describe the influence mechanism of various working conditions on the temperature characteristics, including: Obtain a temperature data set of power supply equipment containing multiple dimensions, preprocess the temperature data set, remove outliers and missing values, and standardize the data to make the data conform to the normal distribution; According to the characteristics of the temperature data set, select the multiple regression analysis method. When there is a linear correlation between the independent variables, select multiple linear regression. When there is a nonlinear relationship between the independent variables and the dependent variable, select polynomial regression. When the number of independent variables exceeds the preset threshold, select stepwise regression. Taking the temperature characteristics as the dependent variable and each operating condition as the independent variable, a regression equation between the temperature characteristics and the operating conditions is established; The least squares method is used to estimate the regression coefficients and obtain the parameter estimates of the regression model; The regression coefficient was tested for significance by t-test, insignificant independent variables were eliminated, the overall significance of the model was tested by F-test, and the determination coefficient was calculated to evaluate the goodness of fit of the model; If the regression model passes the significance test and the goodness of fit is higher than the preset threshold, it will be used as a mathematical model to describe the influence mechanism between temperature characteristics and working conditions; The k-fold cross-validation method is used to randomly divide the temperature data set into k subsets. One of the subsets is selected as the test set each time, and the remaining subsets are used as the training set. The validation process is repeated k times, and the mean square error and determination coefficient of each validation are calculated to evaluate the prediction performance and generalization ability of the mathematical model. According to the size and sign of the regression coefficient of each variable in the mathematical model, the degree and mode of influence of each working condition on the temperature characteristics are analyzed to provide a quantitative basis for temperature control and optimization.

3. The method for detecting temperature characteristics of a power supply device according to claim 1, characterized in that: The partial derivative calculation of the mathematical model is performed to quantify the contribution of each operating condition to the temperature characteristics, and the main influencing factors and secondary influencing factors are determined, including: Obtain the established mathematical model of temperature characteristics, and use the numerical differentiation method to calculate the partial derivative values ​​of each working condition in the mathematical model; According to the absolute values ​​of the partial derivative values, the partial derivative values ​​are sorted from large to small, the working conditions corresponding to the partial derivative values ​​greater than the preset threshold are determined as the main influencing factors, and the working conditions corresponding to the partial derivative values ​​less than the preset threshold are determined as the secondary influencing factors; According to the working conditions corresponding to the main influencing factors, the support vector machine algorithm is used to establish a nonlinear regression model of the main influencing factors and temperature characteristics; The grid search and cross-validation methods were used to optimize the hyperparameters of the support vector machine algorithm and obtain the optimized nonlinear regression model; According to the working conditions corresponding to the secondary influencing factors, the decision tree classification algorithm is used to establish a classification model of the secondary influencing factors and temperature characteristics; The information gain ratio is used as the feature selection standard to prune the decision tree classification model and obtain the optimized classification model; Obtain real-time data of various working conditions, input the real-time data of the main influencing factors into the optimized nonlinear regression model, and obtain the regression prediction value of the temperature characteristics; The real-time data of the secondary influencing factors are input into the optimized classification model to obtain the classification judgment results of the temperature characteristics; The weighted average function is used to perform weighted fusion of the regression prediction value and the classification judgment result. During the fusion, the weight of the main influencing factor is set to 7, and the weight of the secondary influencing factor is set to 3, so as to obtain the temperature characteristic prediction value after combining various working conditions. Obtaining a preset temperature characteristic qualified threshold range, and determining whether the temperature characteristic prediction value falls within the qualified threshold range; If so, it is determined that the temperature characteristics under the current operating condition combination meet the requirements; If not, it is determined that the temperature characteristics under the current operating condition combination do not meet the requirements, and the simulated annealing algorithm is used to automatically adjust the corresponding operating condition parameters; The deviation between the predicted value of the temperature characteristic and the qualified threshold range is used as the optimization objective function. The power supply equipment operating parameter combination that minimizes the deviation is found through iterative search and determined as the optimal operating condition.

4. The method for detecting temperature characteristics of a power supply device according to claim 1, characterized in that: The time series analysis method is introduced to analyze the temperature data set, capture the dynamic change trend of the working conditions, and identify the nonlinear relationship characteristics between the working conditions and the temperature characteristics, including: Acquire a temperature data set, divide the temperature data set according to different working conditions of the power supply equipment, and obtain temperature sub-data sets corresponding to different working conditions; For each temperature sub-dataset, outliers and noise data are detected and removed, and the median filtering method is used for data preprocessing to obtain the preprocessed temperature sub-dataset; Time series analysis is performed on each preprocessed temperature sub-dataset, and an autoregressive moving average model (ARMA) is used to establish a temperature time series model that reflects the temperature change trend under different working conditions. Based on the ARMA model, the wavelet analysis method is used to perform multi-scale decomposition of the temperature time series under each working condition, extract the temperature fluctuation characteristics under different frequency scales, and obtain the wavelet coefficients reflecting the temperature fluctuation characteristics under different working conditions. The wavelet coefficients are used as sample features, and the support vector machine algorithm is used to establish a nonlinear mapping relationship model between working conditions and temperature fluctuation characteristics. A nonlinear mapping model is used to identify the temperature fluctuation characteristics of the real-time collected temperature data under working conditions to determine whether the temperature fluctuation under the current working conditions exceeds the preset normal range. If so, a temperature abnormality warning is triggered.

5. The method for detecting temperature characteristics of a power supply device according to claim 1, characterized in that: The adaptive filtering algorithm is used to track the changes of working conditions in real time and dynamically adjust the parameters and structures in the mathematical model so that the mathematical model can adapt to the dynamic changes of working conditions, including: Acquire real-time data reflecting current working conditions, pre-process the real-time data, and obtain pre-processed real-time data; Adopting adaptive Kalman filter algorithm, the parameters of mathematical model are estimated and updated online in real time according to pre-processed real-time data; The forgetting factor is introduced into the adaptive Kalman filter algorithm to give different weights to historical data; On the basis of adaptively adjusting the model parameters, the topological structure of the mathematical model is changed by adding or deleting the components of the mathematical model; Taking the adaptively adjusted mathematical model parameters and structure as input, optimizing the mathematical model in real time using the optimization algorithm to obtain the optimized mathematical model; Using the optimized mathematical model, the input data under the current working conditions are processed to obtain the predicted values ​​of key parameters; According to the prediction results output by the mathematical model, the optimization control strategy is automatically generated to achieve the optimal control under the working conditions; Continuously monitor key parameters that reflect working conditions. When the change in key parameters exceeds the preset threshold, the adaptive Kalman filter algorithm is triggered to readjust the mathematical model.

6. The method for detecting temperature characteristics of a power supply device according to claim 1, characterized in that: The support vector machine algorithm is used to train and optimize the adjusted mathematical model to improve the prediction accuracy and generalization ability of the mathematical model, including: Obtain a training data set, which includes multiple sample data and corresponding labels; The support vector machine algorithm is used to train the training data set to obtain a preliminary support vector machine model; The cross-validation method was used to evaluate the preliminary support vector machine model, and the prediction accuracy and generalization error of the preliminary support vector machine model on the validation set were obtained; If the prediction accuracy is less than the preset accuracy threshold or the generalization error is greater than the preset error threshold, the following steps are performed: Analyze the characteristics and sample quality of the training data set, remove the noise data and redundant features therein, and obtain an optimized training data set; The optimized training data set is trained again using the support vector machine algorithm to obtain an optimized support vector machine model; The grid search method is used to determine the optimal hyperparameter combination of the support vector machine algorithm. The optimal hyperparameter combination is the hyperparameter combination with the smallest generalization error obtained by cross-validation for the optimized training data set. Retrain the support vector machine model using the optimal hyperparameter combination to obtain the final support vector machine model; Obtain a test data set, use the final support vector machine model to predict the test data set, and obtain the prediction accuracy and number of convergence rounds of the final support vector machine model on the test data set; If the prediction accuracy is greater than the preset accuracy threshold and the number of convergence rounds is less than the preset round number threshold, the final support vector machine model is applied to the actual business scenario for prediction; Continuously monitor the online support vector machine model to obtain its prediction error and data distribution changes; If the prediction error exceeds the preset error threshold or the data distribution change exceeds the preset change threshold, the newly collected data is obtained, and the final support vector machine model is updated using incremental training to obtain a new version of the online support vector machine model.

7. The method for detecting temperature characteristics of a power supply device according to claim 1, characterized in that: The temperature characteristic curve of the power supply device under different working conditions is generated according to the optimized mathematical model. The temperature characteristic curve is used to reflect the temperature change of the device in the actual working environment, including: According to the optimized mathematical model, the working condition parameters of the power supply equipment are obtained, including the ambient temperature and the load power; For each working condition, the temperature characteristic curves of the equipment under multiple different working conditions are calculated through mathematical models; Using MATLAB software, based on the generated temperature characteristic curve, a three-dimensional visualization model of the temperature change of the power supply equipment is constructed to intuitively display the temperature change trend; Using MATLAB software, based on the generated temperature characteristic curve, a three-dimensional visualization model of the temperature change of the power supply equipment is constructed to intuitively display the temperature change trend; The support vector machine algorithm is used to train and establish a mapping relationship model between the operating condition parameters as input and the corresponding temperature characteristic curve as output; Use the trained mapping relationship model to predict the temperature change trend of power supply equipment under unknown working conditions; Obtain the rated operating temperature range of the power supply equipment and compare the predicted temperature change trend with it to determine whether the equipment has an over-temperature risk; If the temperature change trend exceeds the rated operating temperature range, an early warning message will be issued to indicate that the device may have an over-temperature problem.

8. A power supply equipment temperature characteristic detection system, characterized in that: The system comprises: a data acquisition module for acquiring a temperature data set of the power supply device under different working conditions; The multiple regression analysis module is used to establish a mathematical model between the temperature characteristics of the power supply equipment and various working conditions based on the temperature data set by using the multiple regression analysis method. The mathematical model is used to describe the influence mechanism of various working conditions on the temperature characteristics; The partial derivative calculation module is used to quantify the contribution of various working conditions to the temperature characteristics and determine the main and secondary influencing factors by calculating the partial derivatives of the mathematical model; The time series analysis module is used to introduce the time series analysis method to analyze the temperature data set, capture the dynamic change trend of the working conditions, and identify the nonlinear relationship characteristics between the working conditions and the temperature characteristics; The adaptive filtering module is used to use an adaptive filtering algorithm to track the changes in working conditions in real time and dynamically adjust the parameters and structures in the mathematical model so that the mathematical model can adapt to the dynamic changes in working conditions; The support vector machine optimization module is used to train and optimize the adjusted mathematical model using the support vector machine algorithm to improve the prediction accuracy and generalization ability of the mathematical model; The temperature characteristic curve generation module is used to generate the temperature characteristic curve of the power supply equipment under different working conditions according to the optimized mathematical model. The temperature characteristic curve is used to reflect the temperature change of the equipment in the actual working environment.

Images