Transformer temperature early warning method and system

Abstract

The invention relates to the technical field of power equipment operation state monitoring and fault early warning, in particular to a transformer temperature early warning method and system, and the method comprises the steps: 1, collecting the historical temperature data of a transformer; 2, calculating the size of a sliding window; 3, training a support vector regression model according to the transformer temperature data collected in the step 1 and the size of the sliding window obtained in the step 2; 4, training a random forest model according to the transformer temperature data collected in the step 1 and the size of the sliding window obtained in the step 2; step 5, training a gradient lifting regression model according to the transformer temperature data collected in the step 1 and the size of the sliding window obtained in the step 2; step 6, a Stacking model is obtained, and calculation is carried out; 7, predicting the future temperature; 8, constructing an early warning rule base according to historical data and professional knowledge, and taking the early warning rule base as a judgment standard of early warning after transformer temperature prediction; step 9, performing judgment and early warning; the prediction accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment operating status monitoring and fault early warning, and in particular to a transformer temperature early warning method and system. Background Art

[0002] The transformer is the main equipment in the substation and a key part of the entire power grid system. It is responsible for raising and lowering the voltage to achieve the effective transmission and distribution of electric energy. By increasing the voltage, the transformer can reduce the energy loss in the transmission process and enable the electric energy to cross longer distances without significant impact. At the same time, the safety of the transformer is directly related to the stability and reliability of the power grid. Any failure or overheating of the transformer may cause power outages, equipment damage, and even cause large-scale power accidents. Therefore, ensuring the normal operation of the transformer and timely monitoring its status are important measures to maintain the safety of the power grid.

[0003] Among them, the transformer temperature early warning system can identify potential high temperature faults early and take maintenance measures in time, thereby avoiding equipment damage and power outages and improving the power supply reliability of the power grid. The significance of transformer temperature early warning is mainly reflected in fault prevention, safety assurance, improving operating efficiency, extending equipment life and reducing operating costs. By monitoring the temperature in real time, the early warning system can identify temperature anomalies early and take timely measures to prevent equipment failure and damage. Excessive temperature may lead to safety hazards such as aging of insulation materials, short circuits or fires. Temperature early warning effectively reduces these risks and ensures the safety of equipment and personnel. At the same time, temperature monitoring and early warning can optimize the operating status of the transformer and ensure that it works efficiently within a safe range, thereby improving the overall performance and reliability of the power grid. In addition, timely temperature early warning can prevent damage caused by overheating, extend the service life of the transformer, and reduce the frequency of replacement and maintenance. Through effective temperature management, unexpected power outages and maintenance costs can be reduced, and the economy of the power system can be improved. Therefore, transformer temperature early warning is of great significance to the stability and economy of the entire power system.

[0004] The transformer temperature warning method based on adaptive sliding window and stacking proposed in this application mainly consists of transformer temperature prediction and warning strategy, among which the transformer temperature prediction method is mainly divided into three types: statistical method, machine learning method and deep learning method. Figure 1As shown, the statistical methods include linear regression, which is to establish a linear relationship between variables through statistical analysis; it also includes autoregressive integrated moving average, which is a statistical method for time series analysis and is suitable for modeling and prediction of stationary and non-stationary data. The machine learning methods include ensemble learning, which is to combine multiple weakly supervised models in order to obtain a better and more comprehensive strong supervised model; it includes support vector regression, which is an extension of support vector machine and is used to solve regression problems; it also includes artificial neural network, which is a computational model inspired by biological neural network, composed of multiple levels of neurons, and is widely used in classification and regression problems. The deep learning methods include convolutional neural network, which is a deep learning model specially used to process data with grid structure; it also includes long short-term memory network, which is a special recursive neural network specially used to process and predict long-term dependencies in time series data.

[0005] Statistical methods can provide a scientific theoretical basis and practical technical support in transformer temperature prediction, helping to achieve efficient fault detection and condition monitoring. Among them, linear regression minimizes the sum of squared errors between the predicted value and the actual value, and finds the best fitting line for transformer temperature prediction. This method is simple to implement and has high computational efficiency, but it is only suitable for processing linear relationship predictions and is overly sensitive to outliers; among them, the autoregressive integrated moving average model combines the characteristics of autoregression and moving average, and uses the difference method to make the time series stable for time series prediction. This method can handle non-stationary time series data and can capture the autocorrelation in the time series, but the model parameter selection of this method is relatively complex and cannot handle data mutations and trend changes.

[0006] Machine learning methods use a large amount of historical data for training and can automatically extract features and patterns from them. Traditional statistical methods often require a lot of manual intervention and professional knowledge, while machine learning models can automatically learn and optimize, thereby improving prediction efficiency and ensuring high accuracy. Among them, ensemble learning can improve the accuracy and generalization ability of the model by combining multiple learning algorithms, and has a strong anti-interference ability for outliers and noise data, but its effect depends on the selection of base learners and has high computational consumption; the basic idea of ​​support vector regression is to fit data through a hyperplane in a high-dimensional space, and its goal is to minimize the complexity of the model while ensuring that the error between the predicted value and the actual value is within a set threshold. This method can effectively handle high-dimensional data and nonlinear relationships, but its effect depends on parameter selection and is sensitive to feature scaling; artificial neural networks usually include input layers, hidden layers, and output layers, and their learning process includes forward propagation and back propagation. This method can automatically extract features and is suitable for processing large-scale data, but it has high data requirements and is prone to overfitting.

[0007] Deep learning methods use multi-layer neural networks for feature learning and data modeling. They have significant advantages in feature learning and modeling capabilities and are applicable to various complex data types. Among them, convolutional neural networks are specifically used to process image data. They use convolutional layers to extract spatial features. They usually include convolutional layers, pooling layers, and fully connected layers. They can automatically learn effective features and are applicable to large-scale data, but they require a lot of computing power and have high data requirements. Long short-term memory networks effectively overcome the gradient vanishing problem of traditional recursive neural networks in long sequence learning by introducing memory units and gating mechanisms. This method has memory capabilities and can capture complex time series features, but it consumes a lot of computing resources, is sensitive to parameter selection, and has unstable results. Summary of the invention

[0008] In order to solve the above technical problems, the present invention provides a transformer temperature early warning method and system with high prediction accuracy and flexibility.

[0009] A transformer temperature early warning method of the present invention comprises the following steps: step 1: collecting historical temperature data of the transformer, that is, extracting temperature data within a continuous time from a transformer temperature sensor;

[0010] Step 2: Calculate the sliding window size. The sliding window is one of the important algorithms for time series prediction, and its size is a key parameter that affects the effect of the algorithm.

[0011] Step 3: training a support vector regression model based on the transformer temperature data collected in step 1 and the sliding window size obtained in step 2;

[0012] Step 4: training a random forest model based on the transformer temperature data collected in step 1 and the sliding window size obtained in step 2;

[0013] Step 5: training a gradient boosting regression model based on the transformer temperature data collected in step 1 and the sliding window size obtained in step 2;

[0014] Step 6: Stack the models obtained in step 3, step 4, and step 5 to obtain a Stacking model, and calculate whether the prediction mean square error of the current transformer temperature prediction model based on the adaptive sliding window and Stacking is minimized. If it is minimized, proceed to step 7, otherwise return to step 2;

[0015] Step 7: predicting the future temperature according to the transformer temperature prediction model based on adaptive sliding window and stacking that finally obtains the best performance;

[0016] Step 8: Build a warning rule base based on historical data and professional knowledge as a criterion for warning after transformer temperature prediction;

[0017] Step nine, judging and issuing an early warning based on the predicted transformer temperature obtained in step seven and the early warning rule base obtained in step eight; this method uses the Stacking model of support vector regression, random forest and gradient boosting regression for transformer temperature prediction, and predicts the transformer temperature based on Stacking and adaptive sliding window, one of the integrated learning methods, and establishes a transformer temperature early warning rule base, and performs early warning processing according to the predicted transformer temperature and the transformer temperature early warning rule base. This method improves the prediction accuracy and flexibility to achieve effective transformer temperature early warning.

[0018] Preferably, in step 1, the historical temperature data of the transformer is collected, which is the three-phase temperature of optical fibers A, B, and C on the high-voltage side of the main transformer.

[0019] Preferably, the sliding window size is calculated in step 2 to continuously update the sliding window size until the training prediction mean square error is minimized; the sliding window is one of the important algorithms for time series prediction, and its size is a key parameter affecting the effect of the algorithm. Dynamically adjusting the window size according to temperature changes can better capture short-term fluctuations and long-term trends in temperature and improve the accuracy of prediction.

[0020] Preferably, the support vector regression model in step 3 is to find an optimal regression hyperplane so that most of the training data points are located within a specific tolerance band, that is, to fit the data using a function that satisfies the relationship:

[0021] f(x)=ω T φ(x)+b

[0022] Among them, ω is the weight vector, x is the input feature, φ(x i ) is the feature map, and b is the bias. The model mainly includes selection parameters, namely, the width of the tolerance band ε and the penalty parameter C, ε defines the tolerance range of the prediction error, and C controls the degree of penalty for the training error; construct an optimization problem, namely, minimize a loss function, balance the complexity of the model and the fitting error, and satisfy the relationship in this application:

[0023]

[0024] Among them, ξ i and is a slack variable used to handle the situation where data points exceed the tolerance band; constraints are added to ensure that the prediction error of each data point does not exceed ε. In this application, the constraints satisfy the relationship:

[0025]

[0026] Among them, y i is the true target value of the i-th training sample, ωT φ(x i )+b represents the predicted value of the model; to solve the optimization problem, the Lagrange multiplier method is used to combine the objective function and the constraints, and the Lagrangian function is constructed and solved to obtain the optimal weight vector ω and bias b; model prediction, predict new data points; the support vector regression model can effectively process high-dimensional data, map the input data to a high-dimensional feature space through a kernel function, and capture complex nonlinear relationships. By using a tolerance band, the support vector regression model has a certain degree of robustness to outliers and noise, because the goal is to maximize the boundary rather than fit all training data points.

[0027] Preferably, according to the random forest model in step 4, the prediction results of multiple decision tree models are combined; the random forest model improves the overall performance by combining the prediction results of multiple decision tree models, so the random forest can handle large-scale data sets and has a certain robustness to missing values ​​and outliers, and by introducing randomness, the random forest is usually less prone to overfitting than a single decision tree.

[0028] Preferably, the gradient boosting regression model in step 5 includes initialization, calculating the mean F of the target value 0 (x), satisfying

[0029] relation:

[0030]

[0031] where y i is the target value of the i-th training sample; iterative training, for each round m=1,2,...,M, calculate the residual respectively, satisfying the relationship:

[0032]

[0033] Then we train the weak learner h m (x), that is, using the residual As the goal, train a new weak learner and then update the model to satisfy the relationship:

[0034] F m (x) = F m―1 (x)+Υ m h m (x)

[0035] Among them, m is the learning rate, which is used to control the influence of the new model on the overall model; the final prediction is to predict the new sample; the gradient boosting regression model can effectively handle missing values ​​by gradually improving the prediction ability of the model, and can adaptively handle the missing parts in the data through the splitting characteristics of the tree.

[0036] Preferably, the transformer temperature prediction model in step seven predicts the three-phase temperatures of optical fibers A, B, and C on the high-voltage side of the main transformer.

[0037] Preferably, the warning rule base in step eight is divided into three warning levels, wherein the first warning is a minor warning, which requires staff to continuously monitor the temperature change of the transformer online, and no further operation is required; the second warning is a medium warning, which requires staff to check the operating status of the transformer on site; the third warning is a serious warning, which requires the transformer to be shut down immediately for inspection, detailed fault analysis and troubleshooting, and inspection of electrical connections, insulation status, etc. The above three warning levels correspond to three temperature change thresholds, respectively represented by ΔT 1 , ΔT 2 , ΔT 3 Since the temperature change in the transformer has a certain time periodicity, the early warning rule library proposed in this application takes one day as a cycle and uses the average value of the data at that time every day in the historical data as the judgment standard. If the difference between the predicted temperature and the average value is less than ΔT 1 , no warning operation is performed; if the difference between the predicted temperature and the mean is greater than or equal to ΔT 1 and less than ΔT 2 A first-level warning operation is performed; if the difference between the predicted temperature and the mean is greater than or equal to ΔT 2 and less than ΔT 3 Then the second-level warning operation is carried out; if the difference between the predicted temperature and the mean is greater than or equal to ΔT 3 A third-level warning operation is then carried out.

[0038] Preferably, step eight has no dependency on steps one to seven.

[0039] A transformer temperature early warning system of the present invention comprises a data collection unit, an algorithm logic unit, a first model training unit, a second model training unit, a third model training unit, a model building unit, a temperature prediction unit, a rule base unit and a judgment and early warning unit;

[0040] A data collection unit, used to collect and store historical temperature data of variable pressure;

[0041] An algorithm logic unit, which is used to receive historical temperature data and continuously update the sliding window size until the training prediction mean square error is minimized;

[0042] A first model training unit, used for reading historical temperature data and training a support vector regression model in coordination with a sliding window size;

[0043] The second model training unit is used to read the historical temperature data and train the random forest model in coordination with the sliding window size;

[0044] The third model training unit is used to read the historical temperature data and train the gradient boosting regression model in coordination with the sliding window size;

[0045] Model building unit; used to stack the support vector regression model, the random forest model and the gradient boosting regression model to obtain the Stacking model, and calculate whether the prediction mean square error of the current transformer temperature prediction model based on the adaptive sliding window and Stacking is minimized;

[0046] A temperature prediction unit is used to predict the future temperature of the three-phase temperature of optical fibers A, B, and C on the high-voltage side of the main transformer and send it;

[0047] The rule base unit is constructed based on historical data and professional knowledge as the evaluation criteria for transformer temperature prediction and post-warning;

[0048] The judgment and early warning unit is used to receive the predicted future temperature sent by the temperature prediction unit and make judgments and early warnings in combination with the rule base unit.

[0049] Compared with the prior art, the beneficial effects of the present invention are as follows: the method uses a stacking model of support vector regression, random forest and gradient boosting regression for transformer temperature prediction, and performs transformer temperature prediction based on Stacking and adaptive sliding windows, one of the integrated learning methods, and establishes a transformer temperature early warning rule base. Early warning processing is performed according to the predicted transformer temperature and the transformer temperature early warning rule base. The method improves the prediction accuracy and flexibility to achieve effective transformer temperature early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is the transformer temperature prediction technology diagram;

[0051] Figure 2 is a flow chart of the transformer temperature early warning method of the present invention;

[0052] Figure 3 The result diagram of the transformer temperature prediction model with fixed sliding window and Stacking;

[0053] Figure 4 The result diagram of the transformer temperature prediction model with adaptive sliding window and Stacking;

[0054] Figure 5 It is a block diagram of the transformer temperature early warning system of the present invention. DETAILED DESCRIPTION

[0055] In order to facilitate understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. The present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0056] Example 1

[0057] A transformer temperature early warning method and system of the present invention comprises the following steps: step 1: collecting transformer historical temperature data, that is, extracting temperature data in a continuous time from a transformer temperature sensor; the transformer historical temperature data is the three-phase temperature of optical fibers A, B, and C on the high-voltage side of the main transformer;

[0058] Step 2: Calculate the sliding window size. The sliding window is one of the important algorithms for time series prediction. Its size is a key parameter that affects the effect of the algorithm. The sliding window size is continuously updated until the mean square error of the training prediction is minimized.

[0059] Step 3: Train the support vector regression model based on the transformer temperature data collected in step 1 and the sliding window size obtained in step 2. The support vector regression model is to find an optimal regression hyperplane so that most training data points are located within a specific tolerance band, that is, to fit the data using a function that satisfies the relationship:

[0060] f(x)=ω T φ(x)+b

[0061] Among them, ω is the weight vector, x is the input feature, φ(x i ) is the feature map, and b is the bias. The model mainly includes selection parameters, namely, the width of the tolerance band ε and the penalty parameter C, ε defines the tolerance range of the prediction error, and C controls the degree of penalty for the training error; construct an optimization problem, namely, minimize a loss function, balance the complexity of the model and the fitting error, and satisfy the relationship in this application:

[0062]

[0063] Among them, ξ i and is a slack variable used to handle the situation where data points exceed the tolerance band; constraints are added to ensure that the prediction error of each data point does not exceed ε. In this application, the constraints satisfy the relationship:

[0064]

[0065] Among them, y i is the true target value of the i-th training sample, ω T φ(x i)+b represents the predicted value of the model; solve the optimization problem, use the Lagrange multiplier method to combine the objective function and the constraints, construct the Lagrange function and solve it to obtain the optimal weight vector ω and bias b; model prediction, predict the new data point;

[0066] Step 4: Train a random forest model based on the transformer temperature data collected in step 1 and the sliding window size obtained in step 2. The random forest model combines the prediction results of multiple decision tree models.

[0067] Step 5: Train a gradient boosting regression model based on the transformer temperature data collected in step 1 and the sliding window size obtained in step 2. The gradient boosting regression model includes initialization, calculation of the mean value F of the target value 0 (x), satisfying the relationship:

[0068]

[0069] where y i is the target value of the i-th training sample; iterative training, for each round m=1,2,...,M, calculate the residual respectively, satisfying the relationship:

[0070]

[0071] Then we train the weak learner h m (x), that is, using the residual As the goal, train a new weak learner and then update the model to satisfy the relationship:

[0072] F m (x) = F m―1 (x)+Υ m h m (x)

[0073] Among them, m is the learning rate, which is used to control the influence of the new model on the overall model; the final prediction is to predict the new sample;

[0074] Step 6: Stack the models obtained in step 3, step 4, and step 5 to obtain a Stacking model, and calculate whether the prediction mean square error of the current transformer temperature prediction model based on the adaptive sliding window and Stacking is minimized. If it is minimized, proceed to step 7, otherwise return to step 2;

[0075] Step 7: predicting the future temperature according to the transformer temperature prediction model based on adaptive sliding window and stacking that finally obtains the best performance;

[0076] Step 8. Build an early warning rule base based on historical data and professional knowledge as the criterion for early warning after transformer temperature prediction. The early warning rule base is divided into three early warning levels. The first-level early warning is a minor early warning, which requires staff to continuously monitor the temperature change of the transformer online, and no further operation is required; the second-level early warning is a medium early warning, which requires staff to check the operating status of the transformer on site; the third-level early warning is a serious early warning, which requires the transformer to be shut down immediately for inspection, detailed fault analysis and troubleshooting, and inspection of electrical connections, insulation status, etc. The above three early warning levels correspond to three temperature change thresholds, respectively represented by ΔT 1 , ΔT 2 , ΔT 3 Since the temperature change in the transformer has a certain time periodicity, the early warning rule library proposed in this application takes one day as a cycle and uses the average value of the data at that time every day in the historical data as the judgment standard. If the difference between the predicted temperature and the average value is less than ΔT 1 , no warning operation is performed; if the difference between the predicted temperature and the mean is greater than or equal to ΔT 1 and less than ΔT 2 A first-level warning operation is performed; if the difference between the predicted temperature and the mean is greater than or equal to ΔT 2 and less than ΔT 3 Then the second-level warning operation is carried out; if the difference between the predicted temperature and the mean is greater than or equal to ΔT 3 Then carry out the third-level warning operation;

[0077] Step nine: make a judgment and issue an early warning based on the predicted transformer temperature obtained in step seven and the early warning rule base obtained in step eight.

[0078] In this embodiment, the present application is characterized in that a Stacking model stacked with support vector regression, random forest and gradient boosting regression is used for transformer temperature prediction. Many Stacking models may only use similar types of regression models. Since the three models used in this application have different advantages and disadvantages, their integration effect is often better than Stacking using only one model. In addition, the present application also applies an adaptive sliding window to the Stacking model training. The sliding window can extract features at different time scales. Combined with the powerful modeling capabilities of support vector regression, random forest and gradient boosting regression, it can effectively capture short-term and long-term temperature change laws, and the adaptive sliding window can dynamically adjust the training set according to the changes in data, which enables the model to better capture the changing trends and periodic characteristics in time series data, thereby improving the accuracy of prediction. Finally, we also construct a transformer temperature warning rule base based on historical data and professional knowledge, so that the predicted temperature obtained by the transformer temperature prediction model based on adaptive sliding window and Stacking can play its actual role. The transformer temperature warning method and system based on adaptive sliding window and Stacking can dynamically adjust the window size and step size according to the trend and pattern of temperature change, improve the sensitivity to emergencies, and optimize the robustness of warning;

[0079] In the transformer temperature early warning system, the most critical factor affecting the early warning function is the temperature prediction of the transformer. If the prediction error is small, the early warning will play a greater role. Therefore, in order to demonstrate the prediction effect of the transformer temperature prediction method based on adaptive sliding window and stacking proposed in this application in transformer temperature prediction, this application conducted an experiment based on the A, B, and C three-phase temperatures of the optical fiber on the high-voltage side of a substation main transformer for 20 consecutive days with a data interval of one hour, and proved the advantages of this method and the role it can play in practical applications through experimental results.

[0080] This application uses the first 15 days of data in the dataset for model training, and the last 5 days of data for testing the prediction effect. Figure 3 and Figure 4 As shown, the present application predicts the three-phase temperatures of the optical fibers A, B, and C on the high-voltage side of the main transformer respectively. In the figure, feature 1 represents the temperature of phase A, feature 2 represents the temperature of phase B, and feature 3 represents the temperature of phase C. Figure 3 It can be seen that the transformer temperature prediction model based on fixed sliding window and Stacking can accurately predict the future temperature. The predicted temperature trend is basically consistent with the actual temperature trend, but there are still some data points with large errors. Figure 4It can be seen that the error is reduced compared with the fixed sliding window and Stacking methods. Therefore, it can be shown that the transformer temperature prediction method based on adaptive sliding window and Stacking improves the accuracy and robustness of transformer temperature prediction;

[0081] Since the prediction performance of the above two methods is good, it is difficult to evaluate their advantages and disadvantages by comparing them only in the figure. This application supplements Table 1 to more intuitively reflect the advantages of the transformer temperature prediction method based on adaptive sliding window and Stacking. From Table 1, it can be seen that the mean square error of the prediction results of the above two methods is small, but the transformer temperature prediction method based on adaptive sliding window and Stacking still reduces the mean square error of the prediction results to a certain extent on the basis of the transformer temperature prediction method based on fixed sliding window and Stacking. It can be concluded that the transformer temperature prediction method based on adaptive sliding window and Stacking reduces the prediction error in the prediction of the three-phase temperature of A, B, and C. Therefore, the transformer temperature early warning method and system based on adaptive sliding window and Stacking achieves effective transformer temperature early warning by improving the accuracy of transformer temperature prediction.

[0082] Table 1 Prediction result evaluation table

[0083]

[0084] Example 2

[0085] A transformer temperature early warning method and system of the present invention comprises a data collection unit, an algorithm logic unit, a first model training unit, a second model training unit, a third model training unit, a model building unit, a temperature prediction unit, a rule base unit and a judgment and early warning unit;

[0086] A data collection unit, used to collect and store historical temperature data of variable pressure;

[0087] An algorithm logic unit, which is used to receive historical temperature data and continuously update the sliding window size until the training prediction mean square error is minimized;

[0088] A first model training unit, used for reading historical temperature data and training a support vector regression model in coordination with a sliding window size;

[0089] The second model training unit is used to read the historical temperature data and train the random forest model in coordination with the sliding window size;

[0090] The third model training unit is used to read the historical temperature data and train the gradient boosting regression model in coordination with the sliding window size;

[0091] Model building unit; used to stack the support vector regression model, the random forest model and the gradient boosting regression model to obtain the Stacking model, and calculate whether the prediction mean square error of the current transformer temperature prediction model based on the adaptive sliding window and Stacking is minimized;

[0092] A temperature prediction unit is used to predict the future temperature of the three-phase temperature of optical fibers A, B, and C on the high-voltage side of the main transformer and send it;

[0093] The rule base unit is constructed based on historical data and professional knowledge as the evaluation criteria for transformer temperature prediction and post-warning;

[0094] The judgment and early warning unit is used to receive the predicted future temperature sent by the temperature prediction unit and make judgments and early warnings in combination with the rule base unit.

[0095] A transformer temperature warning method and system of the present invention, whose installation method, connection method or setting method are all common mechanical methods, can be implemented as long as the beneficial effect can be achieved; Stacking is an integrated learning method, which improves the performance of the model by combining the prediction results of multiple base learners. Its basic idea is to train multiple base learners to learn different parts of the data respectively, and then use a meta-learner to integrate the outputs of these base learners to form the final prediction result. Among them, the base learner is the key to affecting the effect of the Stacking model, which is support vector regression, random forest and gradient boosting regression in this application; the meta-learner is usually a simple model, which is ridge regression in this application. This type of method can reduce the prediction bias and variance by combining multiple base learners, improve the generalization ability of the model, and through the prediction results of different models, the meta-learner can capture more information, and selecting a base learner with greater diversity can improve the effect of Stacking. Therefore, this application selects support vector regression, random forest and gradient boosting regression with their own characteristics as base learners.

[0096] Support vector regression is a regression method based on support vector machines, which aims to make predictions by minimizing the trade-off between model complexity and training error. The support vector regression model is to find an optimal regression hyperplane so that most training data points are within a specific tolerance band, that is, to fit the data using a function that satisfies the relationship:

[0097] f(x)=ω T φ(x)+b

[0098] Among them, ω is the weight vector, x is the input feature, φ(x i) is the feature map, and b is the bias. The model mainly includes selection parameters, namely, the width of the tolerance band ε and the penalty parameter C, ε defines the tolerance range of the prediction error, and C controls the degree of penalty for the training error; construct an optimization problem, namely, minimize a loss function, balance the complexity of the model and the fitting error, and satisfy the relationship in this application:

[0099]

[0100] Among them, ξ i and is a slack variable used to handle the situation where data points exceed the tolerance band; constraints are added to ensure that the prediction error of each data point does not exceed ε. In this application, the constraints satisfy the relationship:

[0101]

[0102] Among them, y i is the true target value of the i-th training sample, ω T φ(x i )+b represents the predicted value of the model; solve the optimization problem, use the Lagrange multiplier method to combine the objective function and the constraints, construct the Lagrange function and solve it to obtain the optimal weight vector ω and bias b; model prediction, predict the new data point. Support vector regression effectively captures the complex relationship in the data by finding the optimal regression hyperplane in the high-dimensional feature space, combining the tolerance band and penalty mechanism, and is suitable for a variety of regression tasks.

[0103] Random forest is an ensemble learning method, mainly used for classification and regression tasks. It improves the accuracy and robustness of the model by building multiple decision trees and combining their prediction results. Its steps include randomly extracting samples with replacement from the original training data set to form multiple different training subsets, each of which is used to train an independent decision tree; when each node of each decision tree splits, a specific number of features are randomly selected for evaluation instead of using all features. This randomness increases the diversity of the model and reduces the risk of overfitting; based on bagged samples and randomly selected features, multiple decision trees are constructed. The structure of each tree can be different because they are trained based on different data subsets and feature sets; finally, for regression tasks, the average of the prediction results of all trees is taken. Random forest uses the integration of multiple decision trees to improve prediction accuracy and robustness. By introducing randomness, random forest performs well in processing complex data. It is suitable for classification and regression problems and is one of the widely used algorithms in machine learning.

[0104] Gradient boosting regression model, including initialization, calculating the mean F of the target value 0 (x), satisfying the relationship:

[0105]

[0106] where y i is the target value of the i-th training sample; iterative training, for each round m=1,2,...,M, calculate the residual respectively, satisfying the relationship:

[0107]

[0108] Then we train the weak learner h m (x), that is, using the residual As the goal, train a new weak learner and then update the model to satisfy the relationship:

[0109] F m (x) = F m―1 (x)+Υ m h m (x)

[0110] Among them, m is the learning rate, which is used to control the influence of the new model on the overall model; the final prediction is to predict the new sample. Gradient boosting regression has performed well in many machine learning competitions and practical applications, and is a very popular choice in regression tasks. By optimizing the loss function, gradient boosting regression can effectively capture complex patterns in the data.

[0111] The meta-learner of the Stacking model in this application is ridge regression, which constrains the complexity of the model by introducing an L2 regularization term, thereby improving the generalization ability of the model. The process includes: selecting a regularization parameter and determining the regularization strength λ, which is used to control the complexity of the model. The larger the value, the simpler the model is and the more likely it is to be underfitted. The smaller the value, the more complex the model is and the more likely it is to be underfitted.

[0112] Among them, β is the regression coefficient, X i is the eigenvector of the i-th sample; solve the optimization problem and obtain the regression coefficient β by minimizing the loss function. The analytical solution of ridge regression satisfies the relationship:

[0113] β * =(X T X+λI) ―1 X T y

[0114] The loss function formula is used to prevent the model from overfitting. The first half is the mean square error, yi is the temperature value in the data set, that is, the actual value, Xiβ is the predicted value, λ is the regularization strength parameter, and βj is the model parameter, which is different from β.

[0115] The adaptive sliding window adapts to the changes of data by continuously adjusting the window size and position, so as to capture the dynamic characteristics of time series data in the prediction process. Its steps include: initialization, defining the initial size and step size of the sliding window; window adjustment, dynamically adjusting the size of the window according to the changes in the data stream. When the data fluctuates greatly, the window size can be reduced to capture short-term changes. When the data is relatively stable, the window size can be increased to capture long-term trends; feature construction, constructing features from the data in the current window, which can include statistical features and time series features; model construction, using the constructed features to input the prediction model and output the predicted value at the current time point; sliding window movement, according to the set step size, move the sliding window to the next time point, and repeat the above steps. In the transformer temperature prediction, the adaptive sliding window, combined with Stacking, can fully utilize the advantages of different models, enhance the accuracy and robustness of the prediction, and is an effective method for processing complex time series data.

[0116] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A transformer temperature early warning method, characterized in that: Step 1: Collect historical temperature data of the transformer, that is, extract temperature data in a continuous time from the transformer temperature sensor; Step 2: Calculate the sliding window size. The sliding window is one of the important algorithms for time series prediction, and its size is a key parameter that affects the effect of the algorithm. Step 3: training a support vector regression model based on the transformer temperature data collected in step 1 and the sliding window size obtained in step 2; Step 4: training a random forest model based on the transformer temperature data collected in step 1 and the sliding window size obtained in step 2; Step 5: training a gradient boosting regression model based on the transformer temperature data collected in step 1 and the sliding window size obtained in step 2; Step 6: Stack the models obtained in step 3, step 4, and step 5 to obtain a Stacking model, and calculate whether the prediction mean square error of the current transformer temperature prediction model based on the adaptive sliding window and Stacking is minimized. If it is minimized, proceed to step 7, otherwise return to step 2; Step 7: predicting the future temperature according to the transformer temperature prediction model based on adaptive sliding window and stacking that finally obtains the best performance; Step 8: Build a warning rule base based on historical data and professional knowledge as a criterion for warning after transformer temperature prediction; Step nine: make a judgment and issue an early warning based on the predicted transformer temperature obtained in step seven and the early warning rule base obtained in step eight.

2. A transformer temperature early warning method according to claim 1, characterized in that: In the step 1, the historical temperature data of the transformer is collected, which is the three-phase temperature of the optical fibers A, B, and C on the high-voltage side of the main transformer.

3. A transformer temperature early warning method according to claim 1, characterized in that: The sliding window size is calculated in step 2, and the sliding window size is continuously updated until the training prediction mean square error is minimized.

4. A transformer temperature early warning method according to claim 1, characterized in that: The support vector regression model in step 3 is to find an optimal regression hyperplane so that most training data points are located within a specific tolerance band, that is, to fit the data using a function that satisfies the relationship: f(x)=ω T φ(x)+b Among them, ω is the weight vector, x is the input feature, φ(x i ) is the feature map, and b is the bias. The model mainly includes selection parameters, namely, the width of the tolerance band ε and the penalty parameter C, ε defines the tolerance range of the prediction error, and C controls the degree of penalty for the training error; construct an optimization problem, namely, minimize a loss function, balance the complexity of the model and the fitting error, and satisfy the relationship in this application: Among them, ξ i and is a slack variable used to handle the situation where data points exceed the tolerance band; constraints are added to ensure that the prediction error of each data point does not exceed ε. In this application, the constraints satisfy the relationship: Among them, y i is the true target value of the i-th training sample, ω T φ(x i )+b represents the predicted value of the model; to solve the optimization problem, the objective function and the constraints are combined using the Lagrange multiplier method, the Lagrangian function is constructed and solved to obtain the optimal weight vector ω and bias b; model prediction, prediction of new data points.

5. A transformer temperature early warning method according to claim 1, characterized in that: According to the random forest model in step 4, the prediction results of multiple decision tree models are combined.

6. A transformer temperature early warning method according to claim 1, characterized in that: Gradient boosting in step 5 where y i is the target value of the i-th training sample; iterative training, for each round m=1,2,...,M, calculate the residual respectively, satisfying the relationship: Then we train the weak learner h m (x), that is, using the residual As the goal, train a new weak learner and then update the model to satisfy the relationship: F m (x)=F m―1 (x)+Υ m h m (x) Among them, m is the learning rate, which is used to control the influence of the new model on the overall model; the final prediction is to predict the new sample.

7. A transformer temperature early warning method according to claim 2, characterized in that: The transformer temperature prediction model in step seven predicts the three-phase temperatures of optical fibers A, B, and C on the high-voltage side of the main transformer.

8. A transformer temperature early warning method according to claim 1, characterized in that: The warning rule base in step eight is divided into three warning levels, among which the first-level warning is a minor warning, which requires staff to continuously monitor the temperature change of the transformer online, and no further operation is required; the second-level warning is a medium warning, which requires staff to check the operating status of the transformer on site; the third-level warning is a serious warning, which requires the transformer to be shut down for inspection immediately, and detailed fault analysis and troubleshooting are carried out to check the electrical connection, insulation status, etc. The above three warning levels correspond to three temperature change thresholds, represented by ΔT1, ΔT2, and ΔT3 respectively. Since the temperature change in the transformer has a certain time periodicity, the warning rule base proposed in this application takes one day as a cycle, and uses the average value of the data at this time every day in the historical data as the judgment standard. If the difference between the predicted temperature and the mean is less than ΔT1, no warning operation is performed; if the difference between the predicted temperature and the mean is greater than or equal to ΔT1 and less than ΔT2, a first-level warning operation is performed; if the difference between the predicted temperature and the mean is greater than or equal to ΔT2 and less than ΔT3, a second-level warning operation is performed; if the difference between the predicted temperature and the mean is greater than or equal to ΔT3, a third-level warning operation is performed.

9. A transformer temperature early warning method according to claim 1, characterized in that: The step eight has no dependency relationship with the steps one to seven.

10. A transformer temperature early warning system, characterized in that: It includes a data collection unit, an algorithm logic unit, a first model training unit, a second model training unit, a third model training unit, a model building unit, a temperature prediction unit, a rule base unit and a judgment and early warning unit; A data collection unit, used to collect and store historical temperature data of variable pressure; An algorithm logic unit, which is used to receive historical temperature data and continuously update the sliding window size until the training prediction mean square error is minimized; A first model training unit, used for reading historical temperature data and training a support vector regression model in coordination with a sliding window size; The second model training unit is used to read the historical temperature data and train the random forest model in coordination with the sliding window size; The third model training unit is used to read the historical temperature data and train the gradient boosting regression model in coordination with the sliding window size; Model building unit; used to stack the support vector regression model, the random forest model and the gradient boosting regression model to obtain the Stacking model, and calculate whether the prediction mean square error of the current transformer temperature prediction model based on the adaptive sliding window and Stacking is minimized; A temperature prediction unit is used to predict the future temperature of the three-phase temperature of optical fibers A, B, and C on the high-voltage side of the main transformer and send it; The rule base unit is constructed based on historical data and professional knowledge as the evaluation criteria for transformer temperature prediction and post-warning; The judgment and early warning unit is used to receive the predicted future temperature sent by the temperature prediction unit and make judgments and early warnings in combination with the rule base unit.

Images