Transformer state prediction method and device based on oil temperature and computer equipment

Abstract

The application relates to a transformer state prediction method and device based on oil temperature, a computer device, a storage medium and a computer program product. The load power of a transformer to be predicted is input into a target oil temperature prediction model, a target oil temperature prediction interval corresponding to a predicted oil temperature is determined according to the predicted oil temperature output by the target oil temperature prediction model and an original oil temperature prediction interval corresponding to the target oil temperature prediction model, and the running state of the transformer to be predicted is determined according to a comparison result of the real oil temperature and the target oil temperature prediction interval. Compared with a traditional method of predicting the oil temperature by using a grey model to determine the transformer state, the present application predicts the predicted oil temperature of the transformer by using the load power of the transformer, determines a prediction interval based on the predicted oil temperature, determines the running state of the transformer by comparing the real oil temperature with the predicted oil temperature, and improves the efficiency of predicting the transformer state based on the oil temperature.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of transformers, in particular to a transformer state prediction method and device based on oil temperature, computer equipment, storage medium and computer program product. BACKGROUND

[0002] With the progress of science and technology and the development of national comprehensive strength, the territory of China's power grid industry is becoming wider and wider, and the use of electricity has been invested in the development of various industries, which has a profound impact on the normal operation of various industries. The power transformer is a very important facility in the entire power system, and plays a decisive role in the normal operation of the entire power grid. The paralysis of the power transformer may cause large-scale power outages, causing incalculable costs. Therefore, ensuring the normal operation of the power transformer is related to the normal life of the public. In order to ensure the normal operation of the transformer, the oil temperature in the transformer needs to be detected to determine whether the transformer is running normally based on the oil temperature. At present, the oil temperature of the transformer is usually predicted by training a gray model with a large number of samples and then predicting the oil temperature by the gray model. However, the prediction by the gray model has a high training complexity due to the large number of training samples, which may result in a low prediction efficiency.

[0003] Therefore, the current transformer state prediction method based on oil temperature has the defect of low prediction efficiency. SUMMARY

[0004] Therefore, it is necessary to provide a transformer state prediction method based on oil temperature which can improve the prediction efficiency.

[0005] In a first aspect, the present application provides a transformer state prediction method based on oil temperature, which comprises:

[0006] obtaining the load power and the real oil temperature of the transformer to be predicted;

[0007] inputting the load power into a target oil temperature prediction model to obtain a predicted oil temperature output by the target oil temperature prediction model; the target oil temperature prediction model is trained based on support vector regression machine on sample load power and sample oil temperature;

[0008] determining a target oil temperature prediction interval corresponding to the predicted oil temperature according to the original oil temperature prediction interval corresponding to the target oil temperature prediction model and the predicted oil temperature; the original oil temperature prediction interval is obtained according to a preset confidence and a model error of the target oil temperature prediction model;

[0009] determining the running state of the transformer to be predicted according to the comparison result of the real oil temperature and the target oil temperature prediction interval.

[0010] In one of the embodiments, the method further comprises:

[0011] obtaining a sample load power and a sample oil temperature corresponding to the sample load power;

[0012] inputting the sample load power into a to-be-trained oil temperature prediction model, performing prediction based on a support vector regression machine and the sample load power by the to-be-trained oil temperature prediction model, and outputting a test oil temperature;

[0013] obtaining a matching degree of the test oil temperature and the sample oil temperature corresponding to the sample load power; inputting a current model parameter of the to-be-trained oil temperature prediction model and a current existing matching degree into a preset fitness function, and determining a root mean square error value corresponding to a global optimal position of a particle in the preset fitness function based on a particle swarm algorithm;

[0014] if the root mean square error value is greater than a preset value, adjusting a model parameter of the to-be-trained oil temperature prediction model, and returning to the step of inputting the sample load power into the to-be-trained oil temperature prediction model, performing prediction based on the support vector regression machine and the sample load power by the to-be-trained oil temperature prediction model, and outputting the test oil temperature;

[0015] if the root mean square error value is less than or equal to the preset value, obtaining a target oil temperature prediction model according to the current model parameter.

[0016] In one of the embodiments, the obtaining of the sample load power and the sample oil temperature corresponding to the sample load power comprises:

[0017] obtaining an original sample load power and a sample oil temperature corresponding to the original sample load power;

[0018] normalizing the original sample load power and the sample oil temperature respectively to obtain a sample load power and a sample oil temperature.

[0019] In one of the embodiments, the inputting of the sample load power into the to-be-trained oil temperature prediction model, the performing of prediction based on the support vector regression machine and the sample load power by the to-be-trained oil temperature prediction model, and the outputting of the test oil temperature comprise:

[0020] inputting the sample load power into the to-be-trained oil temperature prediction model, mapping the sample load power to a preset dimension space by the to-be-trained oil temperature prediction model, constructing a fitting function in the preset dimension space based on the mapped sample load power and a preset Gaussian kernel function, determining a solution of the fitting function as the test oil temperature based on a Lagrange optimization method, and outputting the test oil temperature.

[0021] In one of the embodiments, the determining of the target oil temperature prediction interval corresponding to the prediction oil temperature according to the prediction interval of the original oil temperature corresponding to the target oil temperature prediction model and the prediction oil temperature comprises:

[0022] The target oil temperature prediction model is input with a plurality of sample load powers to obtain a plurality of target test oil temperatures output by the target oil temperature prediction model;

[0023] The model error of the target oil temperature prediction model is determined according to the error value of the sample oil temperature and the corresponding target test oil temperature, and the model error average value is determined according to a plurality of model errors;

[0024] The prediction interval of the original oil temperature is constructed according to the preset confidence and the model error average value;

[0025] The target interval lower limit is obtained according to the sum of the prediction oil temperature and the interval lower limit of the prediction interval of the original oil temperature, and the target interval upper limit is obtained according to the sum of the prediction oil temperature and the interval upper limit of the prediction interval of the original oil temperature;

[0026] The target oil temperature prediction interval is obtained according to the target interval lower limit and the target interval upper limit.

[0027] In one of the embodiments, the determining of the operation state of the transformer to be predicted according to the comparison result of the real oil temperature and the target oil temperature prediction interval comprises:

[0028] If the real oil temperature is within the range of the target oil temperature prediction interval, it is determined that the operation state of the transformer to be predicted is normal;

[0029] If the real oil temperature is not within the range of the target oil temperature prediction interval, it is determined that the operation state of the transformer to be predicted is abnormal.

[0030] In a second aspect, the present application provides a transformer state prediction device based on oil temperature, which comprises:

[0031] The acquisition module is configured to acquire the load power and the real oil temperature of the transformer to be predicted;

[0032] The prediction module is configured to input the load power into a target oil temperature prediction model to obtain a prediction oil temperature output by the target oil temperature prediction model; the target oil temperature prediction model is obtained based on the training of sample load power and sample oil temperature by a support vector regression machine;

[0033] The interval determination module is configured to determine the target oil temperature prediction interval corresponding to the prediction oil temperature according to the prediction interval of the original oil temperature corresponding to the target oil temperature prediction model and the prediction oil temperature; the prediction interval of the original oil temperature is obtained according to the preset confidence and the model error of the target oil temperature prediction model.

[0034] a state determining module configured to determine the operation state of the transformer to be predicted according to a comparison result of the real oil temperature and the target oil temperature prediction interval.

[0035] In a third aspect, the present application provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method described above when executing the computer program.

[0036] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program implements the steps of the method described above when executed by a processor.

[0037] In a fifth aspect, the present application provides a computer program product, comprising a computer program, and the computer program implements the steps of the method described above when executed by a processor.

[0038] The transformer state prediction method based on oil temperature, the device, the computer device, the storage medium and the computer program product described above, by inputting the load power of the transformer to be predicted into the target oil temperature prediction model, determining the target oil temperature prediction interval corresponding to the predicted oil temperature according to the predicted oil temperature output by the target oil temperature prediction model and the original oil temperature prediction interval corresponding to the target oil temperature prediction model, and determining the operation state of the transformer to be predicted according to the comparison result of the real oil temperature and the target oil temperature prediction interval. Compared with the traditional method of predicting oil temperature by using a gray model to determine the state of the transformer, the present application predicts the predicted oil temperature of the transformer by using the load power of the transformer, determines the prediction interval based on the predicted oil temperature, and determines the operation state of the transformer by comparing the real oil temperature with the predicted oil temperature, thereby improving the efficiency of predicting the state of the transformer based on oil temperature. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 It is a flowchart of the transformer state prediction method based on oil temperature in one embodiment;

[0040] Figure 2 It is a flowchart of the particle swarm algorithm in one embodiment;

[0041] Figure 3 It is a schematic diagram of the support vector machine fitting step in one embodiment;

[0042] Figure 4 It is a flowchart of the transformer state prediction method based on oil temperature in another embodiment;

[0043] Figure 5 It is a schematic diagram of the oil temperature sequence under the first confidence level in one embodiment;

[0044] Figure 6A schematic diagram of an oil temperature sequence under a second confidence level in one embodiment;

[0045] Figure 7 A schematic diagram of an oil temperature sequence under a third confidence level in one embodiment;

[0046] Figure 8 A structural block diagram of an oil temperature-based transformer state prediction apparatus in one embodiment;

[0047] Figure 9 An internal structural diagram of a computer device in one embodiment. DETAILED DESCRIPTION

[0048] In order to make the purposes, technical solutions and advantages of the present application clearer, further detailed description will be made to the present application in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0049] In one embodiment, as shown in Figure 1 An oil temperature-based transformer state prediction method is provided, and the present embodiment takes the method applied to a terminal as an example. It should be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is realized through the interaction of the terminal and the server, and includes the following steps:

[0050] In step S202, the load power and the real oil temperature of the transformer to be predicted are obtained.

[0051] The transformer to be predicted can be a power device for which the oil temperature is predicted. The terminal can predict the oil temperature of the transformer, and then the terminal can obtain the operating state of the winding of the transformer to be predicted according to the oil temperature. The terminal can record the oil temperature of the transformer to be predicted at each moment, and obtain the oil temperature data arranged in time sequence. These data can form an oil temperature sequence, that is, the terminal can sort the historically recorded oil temperature data in time sequence to obtain a real oil temperature sequence, so that the terminal can obtain the real oil temperature of the transformer to be predicted from the real oil temperature sequence. The terminal can also obtain the load power of the above-mentioned transformer to be predicted. Specifically, the terminal can obtain the load power of the three-phase in the above-mentioned transformer to be predicted, including the load real power, the reactive power and a column of corresponding oil temperature data, etc. The real power is the energy consumed by the power device in the actual power consumption process per unit time, and the reactive power is the abbreviation of voltage and current, that is: reactive power = input voltage (V) x maximum consumption current (A). Thus, the terminal can predict the oil temperature according to the load power of the transformer to be predicted.

[0052] Step S204, input the load power into the target oil temperature prediction model to obtain a predicted oil temperature output by the target oil temperature prediction model; the target oil temperature prediction model is trained based on a support vector regression machine (SVR) on sample load powers and sample oil temperatures.

[0053] The target oil temperature prediction model can be trained based on an SVR on sample load powers and sample oil temperatures, and can be used to predict a predicted oil temperature of a transformer to be predicted, specifically, the predicted oil temperature can be predicted according to a load power of the transformer to be predicted. The predicted oil temperature predicted by the target oil temperature prediction model can be a normal oil temperature of the transformer to be predicted, that is, an oil temperature data of the transformer in a normal operation state. The load power can be multiple, and the multiple load powers can be recorded in chronological order. The terminal can predict the load powers at multiple time points based on the target oil temperature prediction model to obtain predicted oil temperatures at the multiple time points. The terminal can normalize the load powers and input the normalized load powers into the target oil temperature prediction model to obtain the predicted oil temperatures output by the target oil temperature prediction model.

[0054] Step S206, determining a target oil temperature prediction interval corresponding to the predicted oil temperature according to an original oil temperature prediction interval corresponding to the target oil temperature prediction model and the predicted oil temperature; the original oil temperature prediction interval is obtained according to a preset confidence and a model error of the target oil temperature prediction model.

[0055] The original oil temperature prediction interval can be obtained by the terminal according to a preset confidence and a model error of the target oil temperature prediction model. The model error of the target oil temperature prediction model can be obtained by comparing a sample oil temperature with a target test oil temperature output by the target oil temperature prediction model after inputting a sample load power into the target oil temperature prediction model. Since the sample oil temperature can represent an oil temperature value of the transformer in a normal operation state, the terminal can determine the model error of the target oil temperature prediction model based on a difference between the sample oil temperature and the target test oil temperature. The terminal can determine a target oil temperature prediction interval corresponding to the predicted oil temperature according to the original oil temperature prediction interval and the predicted oil temperature. The predicted oil temperature can be an oil temperature data predicted by the target oil temperature prediction model based on a load power of the transformer to be predicted, and the target oil temperature prediction model can be used to predict a predicted oil temperature of the transformer in a normal operation state under the load power. Thus, the terminal can determine the target oil temperature prediction interval based on the predicted oil temperature and the original oil temperature prediction interval, and determine whether the transformer is in a normal operation state based on the target oil temperature prediction interval.

[0056] Step S208, determining an operation state of the transformer to be predicted according to a comparison result of the real oil temperature and the target oil temperature prediction interval.

[0057] The real oil temperature can be a real oil temperature data generated in the operation of the transformer to be predicted, there can be multiple real oil temperatures, and each real oil temperature corresponds to a corresponding load power. The terminal predicts the load power based on the above target oil temperature prediction model, so as to obtain a predicted oil temperature. That is, one load power corresponds to one predicted oil temperature and one real oil temperature. The predicted oil temperature can form a target oil temperature prediction interval, and the terminal can determine the operation state of the transformer to be predicted according to the comparison result of the real oil temperature of each load power and the target oil temperature prediction interval. For example, in an embodiment, the operation state of the transformer to be predicted is determined according to the comparison result of the real oil temperature and the target oil temperature prediction interval, including: if the real oil temperature is within the target oil temperature prediction interval, it is determined that the operation state of the transformer to be predicted is normal; and if the real oil temperature is not within the target oil temperature prediction interval, it is determined that the operation state of the transformer to be predicted is abnormal. In this embodiment, the load power can include load powers at multiple time points. For each load power, the terminal can compare the target oil temperature prediction interval formed by the real oil temperature corresponding to the load power and the predicted oil temperature corresponding to the load power, to determine whether the real oil temperature is within the target oil temperature prediction interval. If the terminal detects that the real oil temperature is within the target oil temperature prediction interval, the terminal can determine that the operation state of the transformer to be predicted under the load power is normal; and if the terminal detects that the real oil temperature is not within the target oil temperature prediction interval, the terminal can determine that the operation state of the transformer to be predicted under the load power is abnormal.

[0058] Specifically, the above target oil temperature prediction interval can be a confidence interval, different confidence intervals are obtained under different preset confidence levels, and the terminal can give a predicted change range under a given confidence level, so as to determine whether the oil temperature is in a normal interval. In the training stage, the terminal judges the test oil temperature data by using the established trained oil temperature prediction model, the test oil temperature can be the oil temperature data output by the trained oil temperature prediction model based on the sample load power, and the terminal determines the confidence interval under different confidence levels by using the error of the training data set when training the trained oil temperature prediction model. The terminal can observe whether the real value, that is, the above sample oil temperature, is within the prediction interval, if it is within the prediction interval, it is considered that the prediction is accurate; otherwise, it is considered that the prediction fails, and the accuracy of the prediction data can be obtained. After continuous optimization of the accuracy, the terminal can obtain the trained target oil temperature prediction model. In actual application, when the real value of the oil temperature is not within the target oil temperature prediction interval, it indicates that the operation state of the transformer is abnormal, and the terminal can remind the staff to pay attention to the cooling device and the oil level in the transformer, and check the transformer.

[0059] In the transformer state prediction method based on oil temperature, the load power of the transformer to be predicted is input into the target oil temperature prediction model, the predicted oil temperature is determined according to the predicted oil temperature output by the target oil temperature prediction model and the original oil temperature prediction interval corresponding to the target oil temperature prediction model, and the running state of the transformer to be predicted is determined according to the comparison result of the real oil temperature and the target oil temperature prediction interval. Compared with the traditional method of predicting the oil temperature by using the gray model to determine the state of the transformer, the present application predicts the predicted oil temperature of the transformer by using the load power of the transformer, determines the prediction interval based on the predicted oil temperature, determines the running state of the transformer by comparing the real oil temperature with the predicted oil temperature, and improves the efficiency of predicting the state of the transformer based on the oil temperature.

[0060] In one embodiment, the method further comprises: obtaining a sample load power and a sample oil temperature corresponding to the sample load power; inputting the sample load power into the trained oil temperature prediction model to predict the test oil temperature based on the support vector regression machine and the sample load power; obtaining a matching degree of the test oil temperature and the sample oil temperature corresponding to the sample load power; inputting the current model parameter of the trained oil temperature prediction model and the current matching degree into a preset fitness function, and determining a root mean square error value corresponding to a global optimal position of particles in the preset fitness function based on a particle swarm optimization algorithm; if the root mean square error value is greater than a preset value, adjusting the model parameter of the trained oil temperature prediction model, and returning to the step of inputting the sample load power into the trained oil temperature prediction model to predict the test oil temperature based on the support vector regression machine and the sample load power; and if the root mean square error value is less than or equal to the preset value, obtaining a target oil temperature prediction model according to the current model parameter.

[0061] In this embodiment, the terminal can pre-train the oil temperature prediction model to be trained to obtain a target oil temperature prediction model. The terminal can first acquire a sample load power and a sample oil temperature corresponding to the sample load power. The terminal can acquire the oil temperature prediction model to be trained and input the sample load power into the oil temperature prediction model to be trained. The oil temperature prediction model to be trained can perform prediction based on the support vector regression machine and the sample load power and output a test oil temperature. The terminal can acquire a matching degree of the test oil temperature and the sample oil temperature corresponding to the sample load power, and input a preset fitness function according to the current model parameter of the oil temperature prediction model to be trained and the current matching degree. The training can be performed multiple times, and each training can obtain a matching degree. Therefore, there can be multiple matching degrees, and the current matching degree represents the matching degree that has existed until the current training. The preset fitness function can be a fitness function constructed by the terminal based on the particle swarm algorithm. Therefore, the terminal can determine a root mean square error value corresponding to a global optimal position of a particle in the preset fitness function based on the particle swarm algorithm. The terminal can compare the root mean square error value with a preset value. If the terminal detects that the root mean square error value is greater than the preset value, the terminal can adjust the model parameter of the oil temperature prediction model to be trained and return to the step of inputting the sample load power into the oil temperature prediction model to be trained, performing prediction based on the support vector regression machine and the sample load power, and outputting the test oil temperature, that is, performing the next training. If the root mean square error value is less than or equal to the preset value, the terminal can obtain the target oil temperature prediction model according to the current model parameter. The comparison between the root mean square error value and the preset value can be a process of finding a minimum limit. When the global optimal position of the particle corresponding to the fitness function satisfies the minimum limit, it indicates that the root mean square error value is minimum. At this time, the terminal can determine that the training is completed. In addition, the terminal can also determine that the training is completed when detecting that the number of iterations of the training is greater than a preset iteration threshold, and obtain the target oil temperature prediction model according to the current model parameter.

[0062] Specifically, after the terminal performs oil temperature prediction by using the support vector regression machine, the terminal can also optimize the parameters of the model by using the particle swarm algorithm. The idea of the particle swarm algorithm is similar to the idea of bird foraging. The birds search for food in the area near the birds closest to the food location to quickly find the food location. The birds are equivalent to the particles in the particle swarm algorithm. Each particle has two attributes, that is, speed and position. The current position is a candidate solution. The position is converted by the speed and the nearby position until the set number of iterations or the difference between the generations satisfies the minimum limit, the cycle is terminated, and the optimal individual extreme value in the particle swarm is taken as the global optimal solution. The process of the particle swarm algorithm can be as shown in Figure 2 Figure 2 ​A flowchart of a particle swarm algorithm in an embodiment. The specific flow of the particle swarm algorithm is as follows: first, initialize parameters, set the population size, the speed interval of the particle, the search space, and the maximum number of iterations; after the parameter initialization, define the fitness function of the algorithm, calculate the fitness value of each particle according to the fitness function, select the local optimal solution (pbest) of the particle and the global optimal solution (gbest) of the population, and update the speed, position, and historical optimal position of the particle and the fitness function value corresponding to the global optimal position according to the preset update formula. The terminal can end the parameter optimization when the number of iterations of the above particle swarm algorithm reaches the preset iteration threshold or the global optimal position meets the minimum limit, and obtain the iterated model parameters, wherein the update formula is as follows:

[0063]

[0064] wherein, V i k is the speed of particle i at the kth iteration; w is the inertia weight factor, when w is larger, it is more focused on global optimization, and when it is smaller, it is focused on local optimization, so the terminal needs to select an appropriate w to balance global optimization and local optimization; C1 and C2 are learning factors, usually set to C1=C2=2; P i k-1 , is the individual optimal solution of particle i after k-1 iterations and the historical optimal solution of all particles i after k-1 iterations; rand(0,1) is a random number in the range of [0,1]; is the position of particle i at the kth iteration.

[0065] Through the embodiment, the terminal can train the oil temperature prediction model to be trained based on the support vector regression machine and the particle swarm algorithm, so as to obtain the target oil temperature prediction model that can be used for oil temperature prediction, thereby improving the efficiency of transformer state detection based on oil temperature.

[0066] In one embodiment, obtaining the sample load power and the sample oil temperature corresponding to the sample load power comprises: obtaining the original sample load power and the sample oil temperature corresponding to the original sample load power; and normalizing the original sample load power and the sample oil temperature respectively to obtain the sample load power and the sample oil temperature.

[0067] In this embodiment, the terminal can obtain the sample load power and the sample oil temperature when training the to-be-trained oil temperature prediction model. The terminal can obtain the sample load power and the sample oil temperature by normalizing the original data. For example, the terminal can obtain the original sample load power and the sample oil temperature corresponding to the original sample load power, and normalize the original sample load power and the sample oil temperature respectively to obtain the sample load power and the sample oil temperature. Specifically, the original sample load power can be the load power of three phases in the transformer, and the terminal can convert the original sample load power and the sample oil temperature into dimensionless parameters in the interval [0, 1] respectively, thereby realizing data normalization.

[0068] Through this embodiment, the terminal can obtain the sample load power and the sample oil temperature by normalizing the data, and train the oil temperature prediction model, thereby improving the efficiency of transformer state prediction based on oil temperature.

[0069] In one embodiment, the sample load power is input into the to-be-trained oil temperature prediction model, the to-be-trained oil temperature prediction model performs prediction based on the support vector regression machine and the sample load power, and outputs the test oil temperature, including: inputting the sample load power into the to-be-trained oil temperature prediction model, mapping the sample load power to a preset dimension space by the to-be-trained oil temperature prediction model, constructing a fitting function in the preset dimension space based on the mapped sample load power and a preset Gaussian kernel function, determining the solution of the fitting function based on the Lagrange optimization method, and outputting the solution as the test oil temperature.

[0070] In this embodiment, the terminal can perform oil temperature prediction by using the support vector regression machine. The support vector regression machine can realize oil temperature prediction by means of nonlinear fitting and high-dimensional mapping. For example, the terminal can input the sample load power into the to-be-trained oil temperature prediction model, and map the sample load power to a preset dimension space by the to-be-trained oil temperature prediction model. The number of dimensions of the preset dimension space can be higher than the number of dimensions of the sample load power. The terminal can construct a fitting function in the preset dimension space based on the mapped sample load power and a preset Gaussian kernel function, and determine the solution of the fitting function based on the Lagrange optimization algorithm, and output the solution as the test oil temperature. The fitting diagram of the support vector machine is shown in Figure 3 Figure 3 is a schematic diagram of the support vector machine fitting step in one embodiment. The conventional support vector machine is mostly used for classification problems, and aims to find a partition hyperplane to separate two types of data points. The partition hyperplane is generally represented as: y(x)=w T ​x+b. Wherein, w is the normal vector of the hyperplane, b is the intercept of the hyperplane, w and b determine the distance between the hyperplane and the origin. The final optimization problem of support vector machine is converted to a convex quadratic programming problem, for linearly separable training sample set T={(x1, y1), (x2, y2), …, (x i , y i ), …, (x m , y m )}; wherein, x i is an n-dimensional feature vector, yi∈{-1,1}, i=1,2,3, …, m. The optimal hyperplane is determined by maximizing the distance between the nearest points, and the model is established as:

[0071]

[0072] Wherein, w is the normal vector of the hyperplane, b is the constant term x i is the feature vector of the i-th sample, y i is the class to which x i belongs. Support vector regression machine (SVR) is also to find a hyperplane, so that the sample points are as close to the plane as possible, which is an improvement of support vector machine, and is used to solve data fitting problems. SVR is to find a hyperplane that fits the data points as much as possible, so that the data points exist near the hyperplane. Contrary to the principle of support vector machine, SVR determines the hyperplane by minimizing the distance between the farthest points, that is, minimizing the value of ε. Its traditional objective function can be expressed as:

[0073]

[0074] The constraint condition reduces the fitted data points to the range of ε distance from the hyperplane. At this time, since there may be sample points far from the hyperplane, in order to improve the fault tolerance of the model, the terminal can establish a "interval band", that is, the introduction of slack variable ξ. When the error between the data point and the hyperplane is less than the slack variable ξ, it is considered to be accurate fitting, and the loss is not calculated. This is also the difference between this model and other regression algorithms. After introducing the slack variable, the objective function of the model becomes:

[0075]

[0076] The terminal determines the values of w and b by solving the optimization problem, determines the fitting hyperplane, and its fitting function is: f(x)=w T x+b. The terminal can convert the above problem into its dual problem by using the Lagrange optimization method, and the Lagrange equation is:

[0077]

[0078] Wherein, α, α *, mu and mu * is a hyper parameter that determines the width of the interval boundary. The terminal can take partial derivatives of each parameter in the Lagrange equation and set it to 0, then:

[0079] The terminal can substitute the above results into the Lagrange function, so that the dual problem of the original objective function can be obtained:

[0080]

[0081] The terminal can make the above process satisfy the KKT condition, so it needs to require:

[0082] So the w and b values of the model can be:

[0083] Since most of the data in reality cannot be directly linearly fitted, but can be linearly fitted in a higher dimensional space, in order to solve the problem of nonlinear fitting, the terminal can introduce the concept of kernel function, map the data samples to a higher dimensional space, and perform linear regression fitting in the high dimensional space. The commonly used kernel functions are polynomial kernel function, Gaussian kernel function and linear kernel function. The present application selects the Gaussian kernel function for the subsequent experiment. The Gaussian kernel function formula is:

[0084] Where sigma is the width coefficient. After introducing the Gaussian kernel function, the model objective function can be represented as:

[0085] Then after introducing a new sample x, for example, introducing a new sample load power, the fitting function of the new sample is: f(x) = w T + b. As in the solution of the linear fitting model, the terminal can use the Lagrange optimization method to solve the model, and the solution of the model is: The terminal can take the solution as the output of the model training, that is, the above test oil temperature.

[0086] Through the present embodiment, the terminal can predict the oil temperature through the support vector regression machine and the high-dimensional mapping method, thereby improving the efficiency of oil temperature prediction.

[0087] In one embodiment, the target oil temperature prediction interval corresponding to the predicted oil temperature is determined according to the original oil temperature prediction interval corresponding to the target oil temperature prediction model and the predicted oil temperature, comprising: inputting a plurality of sample load powers into the target oil temperature prediction model to obtain a plurality of target test oil temperatures output by the target oil temperature prediction model; determining a model error of the target oil temperature prediction model according to error values of the sample oil temperatures and the corresponding target test oil temperatures, and determining a model error average value according to a plurality of model errors; constructing an original oil temperature prediction interval according to a preset confidence and the model error average value; obtaining a target interval lower limit according to a sum of the predicted oil temperature and an interval lower limit of the original oil temperature prediction interval, and obtaining a target interval upper limit according to a sum of the predicted oil temperature and an interval upper limit of the original oil temperature prediction interval; and obtaining the target oil temperature prediction interval according to the target interval lower limit and the target interval upper limit.

[0088] In the embodiment, the terminal can determine the target oil temperature prediction interval based on the original oil temperature prediction interval and the predicted oil temperature. The terminal can first obtain the original oil temperature prediction interval. For example, the terminal can input a plurality of the above-mentioned plurality of sample load powers into the target oil temperature prediction model, the target oil temperature prediction model can be a model that has been trained, the terminal can obtain a plurality of target test oil temperatures output by the target oil temperature prediction model; determine a model error of the target oil temperature prediction model according to error values of the sample oil temperatures and the corresponding target test oil temperatures, and the terminal can determine a model error average value according to a plurality of model errors. After obtaining the model error average value, the terminal can construct an original oil temperature prediction interval according to a preset confidence and the model error average value, the terminal can obtain a target interval lower limit according to a sum of the predicted oil temperature and an interval lower limit of the original oil temperature prediction interval, and obtain a target interval upper limit according to a sum of the predicted oil temperature and an interval upper limit of the original oil temperature prediction interval. The terminal can obtain the target oil temperature prediction interval according to the target interval lower limit and the target interval upper limit. Among them, since there are many influencing factors of oil temperature, and only load power is selected as the characteristic value of oil temperature prediction in the present application, the terminal introduces the concept of confidence interval, gives the change range of prediction under the premise of given confidence, establishes the change range of error by using the deviation of training samples, and improves the accuracy of model prediction. The present application adopts the pivot measure method as the interval estimation method of error. That is, the terminal can adopt the pivot measure method as the interval estimation method of error. The terminal can let f(x i ) be the predicted value of the i-th sample, i.e. the above-mentioned test oil temperature, y i be the true value of the i-th sample, i.e. the above-mentioned sample oil temperature, then the error is: X i =y i -f(x i ). Since there are a plurality of samples, the model error can be a plurality of, the terminal can let the model error X obey a normal distribution, then use the statistical quantity:

[0089] wherein, μ is the mathematical expectation of the error sequence, σ is the standard deviation, and t(n-1) represents a normal distribution with a degree of freedom of n-1.

[0090] Then, for a given significance level a, the confidence interval of μ at a confidence level of 1-a is:

[0091] For a new sample x, for example, a new sample load power, the prediction interval thereof is:

[0092]

[0093] After the terminal obtains the prediction interval, if during the training, the prediction interval can be used to calculate the accuracy of the to-be-trained oil temperature prediction model. Each sample x has a prediction interval I(x), and the terminal can detect whether the real value, i.e., the sample oil temperature, falls within the prediction interval. If it falls within the prediction interval, it is considered that the prediction is accurate, and if it falls outside the prediction interval, it is considered that the prediction is incorrect, and finally the accuracy of the model is calculated.

[0094] Let l j be the judgment function of the jth test sample, and then:

[0095] The final evaluation index (accuracy) of the model is: where m is the number of test samples.

[0096] Through this embodiment, the terminal can determine the oil temperature prediction interval based on the model error and the confidence, so that the terminal can determine whether the transformer is running normally based on the comparison between the target oil temperature prediction interval and the predicted oil temperature, thereby improving the efficiency of transformer state detection based on oil temperature.

[0097] In one embodiment, as Figure 4 shown, Figure 4 is a flowchart of a transformer state prediction method based on oil temperature in another embodiment. In this embodiment, the terminal can obtain load power data and oil temperature data of the power transformer, and perform normalization processing on the data. The training set is formed according to the normalized load power, and the oil temperature prediction model is trained based on the support vector regression machine method according to the training set, wherein the terminal can use the particle swarm algorithm to optimize the parameters of the support vector regression machine. After the terminal obtains the trained target oil temperature prediction model, it can predict a to-be-tested oil temperature data, i.e., the predicted oil temperature, based on the load power, and determine the predicted change range, i.e., the target oil temperature prediction interval, under the premise of a given confidence. The terminal can determine whether the to-be-tested oil temperature is within the target oil temperature interval. If yes, it indicates that the transformer is running normally, otherwise, it indicates that the transformer is running abnormally, and the staff needs to be reminded to maintain the transformer.

[0098] In addition, an application example is provided, a data set used in the example includes six features, namely three-phase load real power, reactive power, and an oil temperature data. First, the features of the data set are normalized, and 80% of the data set is used as a training set and 20% is used as a test set. Through the establishment of the above model, the model is trained using the training set, and the trained model is used to fit the test set to obtain the predicted value of the test set, and the root mean square error of the model fitting can be obtained. The established model is used to predict the training data set, and the error of the training data set is used to determine the confidence interval of the error under different confidence levels, wherein the confidence levels can include a first confidence level, a second confidence level and a third confidence level, which are 95%, 90% and 85% respectively. The confidence interval is used to change the prediction result on the test set into a prediction interval, and it is observed whether the true value falls within the prediction interval. If it falls within the prediction interval, it is considered that the prediction is accurate; otherwise, it is considered that the prediction fails. Based on this, the accuracy of the predicted data can be obtained.

[0099] The terminal can use a particle swarm algorithm to optimize parameters, and in the optimization process, the root mean square error of the parameters is calculated using five-fold cross-validation method, and the parameters are optimized by minimizing the root mean square error. The final optimal model parameters are as shown in the following table:

[0100] Parameter C gamma Parameter value 111.69767726409407 0.0093173780934372

[0101] Wherein, each of the above parameters is a parameter in the target oil temperature prediction model, namely the Gaussian kernel parameters c and gamma of the support vector machine. Gamma is a parameter of the insensitive loss function, the greater the gamma, the fewer the support vectors, the smaller the gamma value, the more the support vectors, and the greater the RBF (radial basis function) width. C is a penalty coefficient, and the generalization ability becomes poor when C is too large or too small. The terminal determines the model using the determined parameters, trains the model using the training set, and predicts the results of the test set using the model. The experimental results are as follows, wherein the root mean square error on the test set is: RMSE = 7.877375897325667. The results after introducing confidence levels of 95%, 90% and 85% are shown in the following table:

[0102]

[0103] According to the above results, it is more recommended to use a confidence level of 90% as the confidence level of the model.

[0104] Specifically, the terminal can use the first 100 test data samples for plotting, and the results are shown in Figure 5 , Figure 6 and Figure 7 , Figure 5A schematic diagram of oil temperature sequence under a first confidence level in one embodiment. Figure 6 A schematic diagram of oil temperature sequence under a second confidence level in one embodiment. Figure 7 A schematic diagram of oil temperature sequence under a third confidence level in one embodiment. Wherein lines 601, 701 and 801 represent true values respectively, lines 602, 702 and 802 represent predicted values respectively, and the area enclosed by the dotted line is the prediction interval: Figure 3 Results under a 95% confidence level; Figure 4 Results under a 90% confidence level, Figure 5 Results under a 85% confidence level. Through the results in the figure, it is more recommended to use 90% or 85% as the confidence level of the error confidence interval. When the true value of the oil temperature is not in the prediction interval, the system can be prompted to remind the staff to pay attention to the cooling device and oil level in the transformer, and to check the transformer.

[0105] Through the above embodiments, the terminal can predict the predicted oil temperature of the transformer by using the load power of the transformer, and determine the prediction interval based on the predicted oil temperature. The running state of the transformer is determined by comparing the true oil temperature and the predicted oil temperature, which improves the efficiency of predicting the state of the transformer based on the oil temperature.

[0106] It should be understood that although each step in the flowchart involved in each of the above embodiments is displayed in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless otherwise stated herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least part of the steps in the flowchart involved in each of the above embodiments can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least part of other steps or steps or stages in other steps.

[0107] Based on the same inventive concept, the present application also provides an oil temperature based transformer state prediction device for implementing the above-mentioned oil temperature based transformer state prediction method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more oil temperature based transformer state prediction device embodiments provided below can refer to the limitations of the oil temperature based transformer state prediction method in the above text, which will not be repeated here.

[0108] In one embodiment, as Figure 8As shown, a transformer state prediction device based on oil temperature is provided, comprising: an acquisition module 500, a prediction module 502, an interval determination module 504 and a state determination module 506, wherein:

[0109] The acquisition module 500 is configured to acquire a load power and a real oil temperature of a transformer to be predicted.

[0110] The prediction module 502 is configured to input the load power into a target oil temperature prediction model, and acquire a predicted oil temperature output by the target oil temperature prediction model; the target oil temperature prediction model is obtained by training sample load powers and sample oil temperatures based on a support vector regression machine.

[0111] The interval determination module 504 is configured to determine a target oil temperature prediction interval corresponding to the predicted oil temperature according to an original oil temperature prediction interval corresponding to the target oil temperature prediction model and the predicted oil temperature; the original oil temperature prediction interval is obtained according to a preset confidence and a model error of the target oil temperature prediction model.

[0112] The state determination module 506 is configured to determine an operation state of the transformer to be predicted according to a comparison result of the real oil temperature and the target oil temperature prediction interval.

[0113] In one embodiment, the device further comprises a training module configured to input a sample load power into a to-be-trained oil temperature prediction model, to acquire a test oil temperature by predicting the sample load power based on a support vector regression machine and the to-be-trained oil temperature prediction model, to acquire a matching degree of a sample oil temperature corresponding to the test oil temperature and the sample load power, to input a current model parameter of the to-be-trained oil temperature prediction model and the current matching degree into a preset fitness function, to determine a root mean square error value corresponding to a global optimal position of a particle in the preset fitness function based on a particle swarm optimization algorithm, to adjust the model parameter of the to-be-trained oil temperature prediction model if the root mean square error value is greater than a preset value, and to return to the step of inputting the sample load power into the to-be-trained oil temperature prediction model, predicting the sample load power based on the support vector regression machine and the to-be-trained oil temperature prediction model, and outputting the test oil temperature, or to obtain the target oil temperature prediction model according to the current model parameter if the root mean square error value is less than or equal to the preset value.

[0114] In one embodiment, the training module is specifically configured to acquire an original sample load power and a sample oil temperature corresponding to the original sample load power, and to normalize the original sample load power and the sample oil temperature respectively to obtain the sample load power and the sample oil temperature.

[0115] In an embodiment, the training module is configured to input the sample load power into the oil temperature prediction model to be trained, map the sample load power to a preset dimension space by the oil temperature prediction model to be trained, construct a fitting function in the preset dimension space based on the mapped sample load power and a preset Gaussian kernel function, and determine a solution of the fitting function as the test oil temperature and output the test oil temperature.

[0116] In an embodiment, the interval determination module 504 is configured to input a plurality of sample load powers into the target oil temperature prediction model, obtain a plurality of target test oil temperatures output by the target oil temperature prediction model, determine model errors of the target oil temperature prediction model according to error values of the sample oil temperatures and the corresponding target test oil temperatures, and determine a model error average value according to the plurality of model errors. The interval determination module 504 is further configured to construct an original oil temperature prediction interval according to a preset confidence and the model error average value, obtain a target interval lower limit according to a sum of the predicted oil temperature and an interval lower limit of the original oil temperature prediction interval, obtain a target interval upper limit according to a sum of the predicted oil temperature and an interval upper limit of the original oil temperature prediction interval, and obtain the target oil temperature prediction interval according to the target interval lower limit and the target interval upper limit.

[0117] In an embodiment, the state determination module 506 is configured to determine that the running state of the transformer to be predicted is normal if the real oil temperature is within the target oil temperature prediction interval, and determine that the running state of the transformer to be predicted is abnormal if the real oil temperature is not within the target oil temperature prediction interval.

[0118] The various modules in the transformer state prediction device based on oil temperature can be all or partially implemented by software, hardware, and combinations thereof. The various modules can be embedded in or independent of a processor in a computer device in a hardware form, or can be stored in a memory in the computer device in a software form, so as to be called and executed by a processor to perform operations corresponding to the various modules.

[0119] In an embodiment, a computer device is provided, which can be a terminal. An internal structure diagram of the computer device can be as shown in FIG. 8. Figure 9As shown in the figure. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is used for wired or wireless communication with external terminals. Wireless mode can be achieved through WIFI, mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by the processor to implement an oil temperature-based transformer state prediction method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad provided on the shell of the computer device. It can also be an external keyboard, touchpad or mouse, etc.

[0120] Those skilled in the art can understand that, Figure 9 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0121] In one embodiment, a computer device is provided, including a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the oil temperature-based transformer state prediction method described above.

[0122] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by the processor to implement the oil temperature-based transformer state prediction method described above.

[0123] In one embodiment, a computer program product is provided, including a computer program, and the computer program is executed by the processor to implement the oil temperature-based transformer state prediction method described above.

[0124] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties.

[0125] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0126] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0127] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. An oil temperature based transformer condition prediction method, characterized by, The method comprises: obtaining the load power of the transformer to be predicted and the real oil temperature; inputting the load power into a target oil temperature prediction model to obtain a predicted oil temperature output by the target oil temperature prediction model; the target oil temperature prediction model is trained based on a support vector regression machine on sample load power and sample oil temperature; the training process comprises: inputting the sample load power into a to-be-trained oil temperature prediction model, mapping the sample load power to a preset dimension space by the to-be-trained oil temperature prediction model, constructing a fitting function in the preset dimension space based on the mapped sample load power and a preset Gaussian kernel function, determining the solution of the fitting function based on a Lagrange optimization method as a test oil temperature and outputting; obtaining the matching degree of the test oil temperature and the sample oil temperature corresponding to the sample load power; inputting a preset fitness function according to the current model parameters of the to-be-trained oil temperature prediction model and the current matching degree, determining the root mean square error value corresponding to the particle in the global optimal position in the preset fitness function based on five-fold cross-validation, and performing parameter optimization by minimizing the root mean square error; if the root mean square error value is less than or equal to a preset value, obtaining a target oil temperature prediction model according to the current model parameters; the preset Gaussian kernel function is obtained based on the ratio of the square of the Euclidean distance of the feature vectors of two samples to the square of the width coefficient; determining the target oil temperature prediction interval corresponding to the predicted oil temperature according to the original oil temperature prediction interval corresponding to the target oil temperature prediction model and the predicted oil temperature; the original oil temperature prediction interval is obtained according to a preset confidence and the model error of the target oil temperature prediction model; determining the running state of the transformer to be predicted according to the comparison result of the real oil temperature and the target oil temperature prediction interval.

2. The method of claim 1, wherein, The method further comprises: obtaining sample load power and sample oil temperature corresponding to the sample load power; inputting the sample load power into a to-be-trained oil temperature prediction model, and outputting a test oil temperature by the to-be-trained oil temperature prediction model based on a support vector regression machine and the sample load power; obtaining the matching degree of the test oil temperature and the sample oil temperature corresponding to the sample load power; inputting a preset fitness function according to the current model parameters of the to-be-trained oil temperature prediction model and the current matching degree, and determining the root mean square error value corresponding to the particle in the global optimal position in the preset fitness function based on a particle swarm optimization algorithm; if the root mean square error value is greater than a preset value, adjusting the model parameters of the to-be-trained oil temperature prediction model, and returning to the step of inputting the sample load power into a to-be-trained oil temperature prediction model, and outputting a test oil temperature by the to-be-trained oil temperature prediction model based on a support vector regression machine and the sample load power; if the root mean square error value is less than or equal to a preset value, obtaining a target oil temperature prediction model according to the current model parameters.

3. The method of claim 2, wherein, The method further comprises: obtaining sample load power and sample oil temperature corresponding to the sample load power; obtaining original sample load power and sample oil temperature corresponding to the original sample load power; The original sample load power and the sample oil temperature are normalized respectively to obtain a sample load power and a sample oil temperature.

4. The method of claim 2, wherein, The target oil temperature prediction interval corresponding to the prediction oil temperature is determined according to the target oil temperature prediction model. A plurality of sample load powers are input into the target oil temperature prediction model to obtain a plurality of target test oil temperatures output by the target oil temperature prediction model. A model error of the target oil temperature prediction model is determined according to error values of the sample oil temperatures and corresponding target test oil temperatures, and a model error average value is determined according to a plurality of model errors. An original oil temperature prediction interval is constructed according to a preset confidence and the model error average value. A target interval lower limit is obtained according to a sum of the prediction oil temperature and a lower limit of the original oil temperature prediction interval, and a target interval upper limit is obtained according to a sum of the prediction oil temperature and an upper limit of the original oil temperature prediction interval. A target oil temperature prediction interval is obtained according to the target interval lower limit and the target interval upper limit.

5. The method of claim 1, wherein, The running state of the transformer to be predicted is determined according to a comparison result of the real oil temperature and the target oil temperature prediction interval. If the real oil temperature is within the target oil temperature prediction interval, it is determined that the running state of the transformer to be predicted is normal. If the real oil temperature is not within the target oil temperature prediction interval, it is determined that the running state of the transformer to be predicted is abnormal.

6. An oil temperature-based transformer condition prediction apparatus characterized by comprising: The device comprises: An acquisition module is configured to acquire a load power and a real oil temperature of a transformer to be predicted. A prediction module is configured to input the load power into a target oil temperature prediction model to obtain a prediction oil temperature output by the target oil temperature prediction model. The target oil temperature prediction model is trained based on a support vector regression machine for sample load powers and sample oil temperatures. The device further comprises a training module configured to input the sample load powers into a to-be-trained oil temperature prediction model, map the sample load powers to a preset dimension space by the to-be-trained oil temperature prediction model, construct a fitting function in the preset dimension space based on the mapped sample load powers and a preset Gaussian kernel function, determine a solution of the fitting function based on a Lagrange optimization method, and output the solution as a test oil temperature. The device further comprises a matching degree acquisition module configured to acquire a matching degree of the test oil temperature and a sample oil temperature corresponding to the sample load power. The device further comprises a parameter optimization module configured to input current model parameters of the to-be-trained oil temperature prediction model and a current matching degree into a preset fitness function, determine a root mean square error value corresponding to a global optimal position of particles in the preset fitness function based on five-fold cross-validation, and perform parameter optimization by minimizing the root mean square error. If the root mean square error value is less than or equal to a preset value, a target oil temperature prediction model is obtained according to the current model parameters. The preset Gaussian kernel function is obtained based on a ratio of a square of a Euclidean distance of feature vectors of two samples to a square of a width coefficient. The interval determining module is configured to determine a target oil temperature prediction interval corresponding to the predicted oil temperature according to an original oil temperature prediction interval corresponding to the target oil temperature prediction model and the predicted oil temperature, wherein the original oil temperature prediction interval is obtained according to a preset confidence and a model error of the target oil temperature prediction model; The state determining module is configured to determine an operation state of the transformer to be predicted according to a comparison result of the real oil temperature and the target oil temperature prediction interval.

7. The apparatus of claim 6, wherein, The training module is configured to: obtain a sample load power and a sample oil temperature corresponding to the sample load power; input the sample load power into a to-be-trained oil temperature prediction model, perform prediction based on a support vector regression machine and the sample load power by using the to-be-trained oil temperature prediction model, and output a test oil temperature; obtain a matching degree of the test oil temperature and the sample oil temperature corresponding to the sample load power; input a current model parameter of the to-be-trained oil temperature prediction model and a current existing matching degree into a preset fitness function, determine a root mean square error value corresponding to a global optimal position of particles in the preset fitness function based on a particle swarm algorithm; if the root mean square error value is greater than a preset value, adjust the model parameter of the to-be-trained oil temperature prediction model, and return to the step of inputting the sample load power into the to-be-trained oil temperature prediction model, performing prediction based on the support vector regression machine and the sample load power by using the to-be-trained oil temperature prediction model, and outputting the test oil temperature; if the root mean square error value is less than or equal to the preset value, obtain a target oil temperature prediction model according to the current model parameter.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to realize the steps of the method in any one of claims 1 to 5.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method in any one of claims 1 to 5.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method in any one of claims 1 to 5.

Images