Transformer oil model identification method and system

Abstract

The invention discloses a transformer oil model identification method and system. The method comprises the following steps: acquiring first data of a to-be-identified oil product; performing first preprocessing on the first data to obtain second data after the first preprocessing; and inputting the second data into a pre-trained oil product model identification model to obtain a model identification result of the to-be-identified oil product, the pre-trained oil product model identification model being any model of which the input is the second data and the output is the model identification result of the oil product. Compared with the prior art that the type of the oil product can be determined only by sending the oil product to be identified to a professional chemical laboratory for analysis, the identification method provided by the invention can rapidly identify the type of the oil product to be identified.

Description

Technical Field

[0001] The invention relates to the technical field of transformer oil, and in particular to a transformer oil model identification method and system. Background Art

[0002] Transformer oil is the main liquid insulating medium inside the transformer, and plays an important role in electrical insulation, air isolation, and circulating cooling. According to different components and sources, transformer oil can be divided into mineral insulating oil, natural ester insulating oil, mixed insulating oil, etc. When filling new transformers with oil and recycling waste transformer oil, effective identification of the type of insulating oil can achieve the rational use and efficient recycling of transformer oil, which is of great value. The identification of insulating oil models mainly relies on the electrical and physical and chemical properties of insulating oil, such as dielectric constant, conductivity, dielectric loss tangent, kinematic viscosity, etc., but since the electrical and physical and chemical properties of insulating oil will change significantly during the operation of the transformer, the detection of electrical and physical and chemical properties may not be accurate enough. In addition, some types of transformer oil have similar electrical and physical and chemical properties, such as KI25 and KI45. Therefore, it is of great significance to propose a more effective method for identifying transformer oil models.

[0003] At present, chemical analysis methods are most commonly used in the analysis and identification technologies of different types of transformer oils, such as gas chromatography, chemical titration, thermogravimetric analysis, etc. Among them, gas chromatography is the most effective method for detecting oil components. The chemical composition of oil is analyzed by heating the oil to volatilize and separate its different components. Since mineral oil, natural ester oil and synthetic oil have different peaks on the chromatogram, oil signal analysis can be performed through different peaks.

[0004] However, methods such as gas chromatography require the oil to be sent to a professional chemical laboratory for analysis, which results in a long oil model identification cycle. Summary of the invention

[0005] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0006] In view of the above existing problems, the present invention is proposed.

[0007] Therefore, the present invention provides a transformer oil model identification method and system, which can solve the problems mentioned in the background technology.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0009] In a first aspect, the present invention provides a transformer oil model identification method, comprising:

[0010] Acquire first data of the oil product to be identified;

[0011] Performing a first preprocessing on the first data to obtain second data after the first preprocessing;

[0012] The second data is input into a pre-trained oil product model recognition model to obtain a model recognition result of the oil product to be identified. The pre-trained oil product model recognition model is an arbitrary model whose input is the second data and whose output is the oil product model recognition result.

[0013] As a preferred solution of the transformer oil model identification method of the present invention, the pre-trained oil model identification model includes: a plurality of different discrimination models;

[0014] The number of the discrimination models depends on the number of the first data of the oil product to be identified;

[0015] The first data of any oil product to be identified includes several.

[0016] As a preferred solution of the transformer oil model identification method of the present invention, wherein: the inputs of the several different discrimination models are all second data, and the outputs are all oil model identification results;

[0017] The inputs of the different discriminant models are respectively second data after first preprocessing is performed on the first data;

[0018] And the number of the several different discriminant models is the same as the number of the second data after the first preprocessing.

[0019] As a preferred solution of the transformer oil model identification method of the present invention, the first preprocessing of the first data includes:

[0020] Performing denoising processing on the first data to obtain denoised first data;

[0021] Performing baseline correction on the denoised first data to obtain corrected first data;

[0022] Performing data dimension reduction on the corrected first data to obtain second data after the first preprocessing.

[0023] As a preferred solution of the transformer oil model identification method of the present invention, wherein: performing data dimension reduction on the corrected first data to obtain the second data after the first preprocessing includes:

[0024] Acquire a covariance matrix of the corrected first data;

[0025] Calculating the eigenvalues ​​and eigenvectors of the covariance matrix;

[0026] Calculate the variance contribution rate according to the eigenvalues, and use the eigenvectors corresponding to the first preset number of eigenvalues ​​whose cumulative contribution rates exceed the preset ratio as the conversion matrix;

[0027] The product of the conversion matrix and the corrected first data is determined as the first preprocessed second data.

[0028] As a preferred solution of the transformer oil model identification method of the present invention, the step of inputting the second data into a pre-trained oil model identification model to obtain the model identification result of the oil to be identified includes:

[0029] Inputting the second data into a plurality of different discriminant models to obtain a plurality of different model outputs;

[0030] Using a variety of evaluation indicators to perform performance evaluation on the outputs of each of the several different models, and determining the evaluation result of each discriminant model;

[0031] The discriminant model with the best evaluation result is determined as the oil product type recognition model.

[0032] As a preferred solution of the transformer oil model identification method described in the present invention, the oil product to be identified is a sample of one or several different oil samples.

[0033] In a second aspect, the present invention provides a transformer oil model identification system, comprising:

[0034] An acquisition module, used for acquiring first data of the oil product to be identified;

[0035] A processing module, used for performing a first preprocessing on the first data to obtain second data after the first preprocessing;

[0036] The identification module is used to input the second data into a pre-trained oil model identification model to obtain the model identification result of the oil to be identified. The pre-trained oil model identification model is any model whose input is the second data and whose output is the oil model identification result.

[0037] In a third aspect, the present invention 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 above-mentioned method when executing the computer program.

[0038] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the method described above when executed by a processor.

[0039] Compared with the prior art, the present invention has the following beneficial effects: the present invention proposes a transformer oil model identification method and system, which obtains first data of the oil to be identified; performs first preprocessing on the first data to obtain second data after the first preprocessing; inputs the second data into a pretrained oil model identification model to obtain a model identification result of the oil to be identified, and the pretrained oil model identification model is an arbitrary model that inputs the second data and outputs the oil model identification result. Compared with the prior art, which requires the oil to be identified to be sent to a professional chemical laboratory for analysis before the oil model can be determined, the identification method of the present invention can quickly identify the model of the oil to be identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0041] Figure 1 A method flow chart of a transformer oil model identification method and system provided by one embodiment of the present invention;

[0042] Figure 2 A schematic flow chart of a method for identifying transformer oil type and a spectrum acquisition method of insulating oil in a system provided by an embodiment of the present invention;

[0043] Figure 3 A schematic diagram of original near-infrared spectra of three types of transformer oils in a transformer oil model identification method and system provided by one embodiment of the present invention;

[0044] Figure 4 A schematic diagram of near-infrared spectra of three types of insulating oils after correction according to a transformer oil model identification method and system provided by an embodiment of the present invention;

[0045] Figure 5 A schematic diagram of the system structure of a transformer oil model identification method and system provided by one embodiment of the present invention;

[0046] Figure 6 An internal structural diagram of a computer device of a transformer oil model identification method and system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0048] Example 1

[0049] Reference Figure 1-Figure 6 , which is the first embodiment of the present invention, provides a transformer oil model identification method and system, including:

[0050] There are some shortcomings in the existing related technologies, for example, professional operation is required, the analysis process is time-consuming, and the cost is high. This application provides a method that can effectively solve the above-mentioned problems. Next, how to implement the transformer oil model identification method will be described in detail in combination with multiple embodiments;

[0051] Figure 1 A method flow chart of a transformer oil model identification method and system is shown, including:

[0052] S101, obtaining first data of the oil product to be identified;

[0053] In the embodiment of the present application, the oil product to be identified is one or several different oil samples, for example, one or several different types of transformer oil.

[0054] In an optional embodiment, the first data may be near infrared spectral data, or may be Fourier transform infrared spectral data, Raman spectral data, etc. In the present application, near infrared spectral data is used as a detailed description, but other types of data of the oil product to be identified can be selected as the first data of the oil product to be identified and can be successfully implemented through the technical solution of the present application.

[0055] In an optional embodiment, the method for obtaining the first data of the oil product to be identified (i.e., near-infrared spectral data) can be that the server controls the near-infrared spectrometer to collect the near-infrared spectral data of the oil product to be identified, and the near-infrared spectrometer directly sends the collected near-infrared spectral data to the server.

[0056] In an optional embodiment, the method for obtaining the near-infrared spectral data of the oil product to be identified may also be to obtain the original near-infrared spectral data of the oil product to be identified collected multiple times by a near-infrared spectrometer; and determine the average value of the original near-infrared spectral data collected multiple times as the near-infrared spectral data of the oil product to be identified.

[0057] It should be noted that in order to ensure the accuracy of the near-infrared spectral data of the oil to be identified, the near-infrared spectrometer can be controlled to collect the near-infrared spectral data of the oil to be identified multiple times, and the near-infrared spectral data can be used as the original near-infrared spectral data, and the average value of multiple original near-infrared spectral data can be calculated, and the average value can be determined as the near-infrared spectral data of the oil to be identified.

[0058] In the embodiment of the present application, the method of collecting the original near-infrared spectral data of the oil to be identified multiple times can be to use the transmission spectrum collection method to collect the near-infrared spectrum. When collecting spectral data, the oil to be identified is placed in a 20×10×50mm quartz cuvette, and the wide side of the sample pool is JGS3 near-infrared high-transmittance glass with a thickness of 1.25mm. The near-infrared spectrometer band should cover 900-2200nm, the resolution of the near-infrared spectrometer should be better than 7nm, and the signal-to-noise ratio should be better than 5000:1. The integration time of the spectrum acquisition is set to 100ms, and the average number of times is 12 times.

[0059] It should be noted that the benefit of step S101 is to ensure that the near infrared spectrum data collected has high accuracy and repeatability, thereby providing a reliable data basis for subsequent model identification. In actual operation, by repeatedly collecting and calculating the average value, the single measurement error can be effectively reduced, and the stability and reliability of data can be improved. In addition, by selecting a suitable spectrometer and setting reasonable parameters, such as band coverage, resolution, signal-to-noise ratio and integration time, the quality of data collection can be further optimized, and more accurate input data can be provided for subsequent model identification models. Such data collection and pre-processing steps are the key links for realizing rapid and accurate identification of transformer oil models.

[0060] S102, performing a first preprocessing on the first data to obtain second data after the first preprocessing;

[0061] In the embodiment of the present application, when the first data is near-infrared spectrum data, the second data is processed near-infrared spectrum data.

[0062] In the embodiment of the present application, performing first preprocessing on the first data includes:

[0063] Performing denoising processing on the first data to obtain denoised first data;

[0064] Performing baseline correction on the denoised first data to obtain corrected first data;

[0065] Performing data dimension reduction on the corrected first data to obtain second data after first preprocessing.

[0066] In an optional embodiment, the denoising process for the first data may be performed by denoising the near-infrared spectrum data in a moving average smoothing manner to obtain denoised near-infrared spectrum data.

[0067] In an optional embodiment, k wavelength points of the denoised near-infrared spectral data are It is obtained by the following formula:

[0068]

[0069] Among them, x i is the i-th wavelength point in the near-infrared spectrum data, s is the smoothing window size, and N is the total number of wavelength points in the spectrum of the near-infrared spectrum data. At this time, the points at both ends of the spectrum in the denoised near-infrared spectrum data (the points on the leftmost (s-1 / 2) and the rightmost (s-1 / 2)) cannot be calculated using the above formula, but should be calculated using the following approximate smoothing formula, starting with the s-1 / 2 points on the leftmost side, as shown below:

[0070]

[0071] The second is the s-1 / 2 points on the far right, as shown in the following formula.

[0072]

[0073] Therefore, based on the above formula, the denoised near-infrared spectral data is determined.

[0074] It should be noted that the value of the smoothing window size s can be set according to actual needs, and this embodiment will not limit it here.

[0075] In an optional embodiment, denoising can also use a wavelet transform method. Wavelet transform is an effective signal processing tool that can analyze signals in both the time domain and the frequency domain. By selecting appropriate wavelet basis functions and decomposition levels, noise in near-infrared spectral data can be effectively removed while retaining useful information. In this embodiment, the original spectral data is first subjected to wavelet decomposition, and then the wavelet coefficients obtained by the decomposition are threshold processed, and finally the denoised spectral data is obtained by wavelet reconstruction. This method is particularly suitable for processing non-stationary signals containing noise, such as near-infrared spectral data.

[0076] It should be noted that the above denoising operation can effectively remove random noise in near-infrared spectral data and improve the signal-to-noise ratio of the data, thereby providing a clearer spectral signal for subsequent analysis and processing. Denoising is an important step in data preprocessing. It can reduce random fluctuations in the data, make the spectral curve smoother, and facilitate subsequent baseline correction and data dimensionality reduction processing.

[0077] In an optional embodiment, the denoised near-infrared spectral data can be baseline corrected by spectral differential derivation to obtain corrected near-infrared spectral data. After moving average smoothing, spectral differential derivation can be performed to eliminate the influence of baseline drift. The formula for first-order differential derivation of denoised near-infrared spectral data is shown below.

[0078]

[0079] Among them, D xk is the wavelength point of the spectrum after differentiation, g is the difference interval, for the leftmost and rightmost g wavelengths, the forward / backward difference method is used to calculate the first-order derivative. First, for the leftmost g wavelength points:

[0080]

[0081] For the rightmost g wavelength points:

[0082]

[0083] Finally, the corrected near-infrared spectrum data S = [D x1 ,D x2 ,…,D xN ].

[0084] In an optional embodiment, the correction can also use a polynomial fitting method. This method fits the denoised near-infrared spectral data to correct the baseline drift by selecting an appropriate polynomial function. The specific operation is to first determine the order of a polynomial, and then use mathematical tools such as the least squares method to find the best fitting curve to achieve the correction of the baseline. The advantage of the polynomial fitting method lies in its flexibility, and the appropriate polynomial order can be selected according to the characteristics of the data to achieve the best correction effect. In addition, this method does not introduce additional noise during the processing, thereby ensuring the purity of the data. Through the above-mentioned denoising and baseline correction steps, a more accurate and stable near-infrared spectral data can be obtained, which provides a solid data foundation for subsequent model identification.

[0085] In the embodiment of the present application, performing data dimension reduction on the corrected first data to obtain the second data after the first preprocessing includes:

[0086] Acquire a covariance matrix for correcting the first data;

[0087] Compute the eigenvalues ​​and eigenvectors of the covariance matrix;

[0088] Calculate the variance contribution rate according to the eigenvalues, and use the eigenvectors corresponding to the first preset number of eigenvalues ​​whose cumulative contribution rates exceed the preset ratio as the conversion matrix;

[0089] The product of the conversion matrix and the corrected first data is determined as the second data after the first preprocessing.

[0090] In an optional embodiment, the corrected near-infrared spectrum data may be subjected to data dimension reduction by means of principal component analysis to obtain processed second data (ie, processed near-infrared spectrum data).

[0091] In an optional embodiment, for the corrected near-infrared spectral data S, the covariance matrix V is calculated as shown in the following formula.

[0092]

[0093] In an optional embodiment, the eigenvalues ​​and eigenvectors of the covariance matrix are calculated, and the variance contribution rate is calculated based on the eigenvalues. Then, the eigenvectors corresponding to the first preset number of eigenvalues ​​whose cumulative contribution rates exceed a preset ratio are used as a transformation matrix. Finally, the product of the transformation matrix and the corrected sample near-infrared spectrum data is determined as the processed sample near-infrared spectrum data.

[0094] For example, the eigenvalues ​​of the covariance matrix {σ1,σ2,…σ N} and the corresponding eigenvectors {a1,a2,…,a N}, then select the first m eigenvalues ​​whose cumulative variance contribution rate exceeds 95% as the principal components, and record the transformation matrix P = {a1, a2, …, a m}, the second data after dimensionality reduction (processed near-infrared spectrum data) is recorded as S*=P×S.

[0095] In an optional embodiment, data dimension reduction can be performed by feature selection based on importance score. Specifically, the importance score of each spectral band in the near-infrared spectral data of the calibration sample to the model recognition result is calculated through a pre-trained neural network model; then, the spectral bands with importance scores less than a preset threshold are removed from the near-infrared spectral data of the calibration sample to obtain the processed near-infrared spectral data of the sample.

[0096] Specifically, the neural network model can be a random forest model, which calculates the importance score of each spectral band in the near-infrared spectral data of the calibrated sample to the recognition result, retains the bands with higher scores as key features, and eliminates the bands with less impact on the recognition result.

[0097] In an optional embodiment, data dimension reduction can be performed by feature selection based on chemical composition information: the chemical composition of the sample oil sample corresponding to the calibration sample near-infrared spectrum data is analyzed, and the spectral bands in the calibration sample near-infrared spectrum data related to the model of the sample oil sample are determined as the processed sample near-infrared spectrum data. For example, in combination with the chemical composition analysis of transformer oil, characteristic bands closely related to the oil model are selected, such as characteristic absorption peaks of functional groups such as CH and OH.

[0098] It should be noted that after data cleaning is completed, the near infrared spectrum data of the multiple processed samples are randomly divided into a near infrared spectrum data training set and a near infrared spectrum data test set according to a preset ratio. For example, 80% of the data of the sample near infrared spectrum data set is randomly selected as the near infrared spectrum data training set, and 20% of the data is used as the near infrared spectrum data test set.

[0099] It should be noted that the first data is subjected to the first preprocessing to obtain the second data after the first preprocessing, which can significantly reduce the dimension of the data while retaining most of the information. This dimensionality reduction process not only simplifies the subsequent data processing process, but also helps to improve the computational efficiency and reduce the demand for storage space. Through methods such as principal component analysis (PCA), the original high-dimensional data can be converted into principal components in a low-dimensional space, and these principal components are linear combinations of the most important features in the original data. In an embodiment of the present invention, by selecting a suitable number of principal components, it can be ensured that the data after dimensionality reduction can still reflect most of the changing trends and features of the original data, thereby providing strong support for the accurate identification of transformer oil models. In addition, the data after dimensionality reduction is also convenient for processing pattern recognition and classification algorithms, which is crucial for realizing an automated transformer oil model identification system.

[0100] S103, input the second data into a pre-trained oil product model recognition model to obtain a model recognition result of the oil product to be recognized. The pre-trained oil product model recognition model is any model that takes the second data as input and outputs the oil product model recognition result.

[0101] In the embodiment of the present application, the pre-trained oil type recognition model includes: a number of different discrimination models;

[0102] The number of discriminant models depends on the number of first data of the oil products to be identified;

[0103] The first data of any oil product to be identified includes several.

[0104] In the embodiment of the present application, the inputs of the several different discrimination models are all the second data, and the outputs are all the oil product model recognition results;

[0105] The inputs of different discriminant models are respectively the second data after the first data is subjected to the first preprocessing;

[0106] And the number of the several different discriminant models is the same as the number of the second data after the first preprocessing.

[0107] In the embodiment of the present application, the second data is input into the pre-trained oil product model recognition model, and the model recognition result of the oil product to be identified is obtained, including:

[0108] Inputting the second data into a plurality of different discriminant models to obtain a plurality of different model outputs;

[0109] Use multiple evaluation indicators to evaluate the performance of several different model outputs and determine the evaluation results of each discriminant model;

[0110] The discriminant model with the best evaluation result is determined as the oil type recognition model.

[0111] In an optional embodiment, the oil product model recognition model can be trained using a deep learning network. The network can be a convolutional neural network (CNN), a recurrent neural network (RNN), or a combination of the two. Other machine learning algorithms such as support vector machines (SVM) can also be used. Other methods other than machine learning can also be used to establish an oil product model recognition model, for example, a decision tree model is used, which identifies the oil product model through a series of judgment rules. The decision tree model is easy to understand and implement, and can process data with nonlinear relationships. In a decision tree, each internal node represents a judgment on an attribute, each branch represents the output of a judgment result, and each leaf node represents a category. By constructing a decision tree, transformer oil can be quickly classified. In practical applications, multiple models can also be combined to use integrated learning methods to improve the accuracy and robustness of recognition. For example, the prediction results of different models such as SVM, decision tree, neural network, etc. can be fused, and the final model recognition result can be obtained by voting or weighted averaging. The integrated learning method can combine the advantages of different models and reduce the deviation and overfitting problems that may exist in a single model. However, no matter what model design is adopted, as long as the model input is the second data and the model output is the oil model identification result, it should be within the protection scope of this application.

[0112] It should be noted that the oil model identification model can determine the model of the oil to be identified, and the model identification result includes the model of the oil to be identified; the processed near-infrared spectral data is input into the oil model identification model to obtain the model identification result output by the oil model model.

[0113] In an optional embodiment, before the oil product to be identified is modeled, it is necessary to first construct an oil product model identification model, and the construction process may be as follows:

[0114] Step 1: Obtain a near infrared spectral data training set and a near infrared spectral data test set; both the near infrared spectral data training set and the near infrared spectral data test set include sample near infrared spectral data of a plurality of different types of sample oil samples; and each type of sample oil sample has multiple sample near infrared spectral data.

[0115] Step 2: Train multiple different machine learning models using the near-infrared spectral data training set to obtain multiple different discriminant models.

[0116] In this embodiment, in order to achieve the best effect, multiple machine learning models are used for trial, and the optimal discriminant model is selected through model optimization. Multiple different machine learning models may include a support vector machine model, a random forest model, and a convolutional neural network model.

[0117] Specifically, in the process of training the support vector machine model through the near infrared spectrum data training set, first, the near infrared spectrum data of the sample extracted in the near infrared spectrum data training set can be normalized to ensure that the magnitude of all features is the same; secondly, the radial basis function (RBF) is selected as the kernel function to map the sample near infrared spectrum data to a high-dimensional space so as to better separate different oil models; and the penalty parameter C and the kernel function parameter γ in the support vector machine (SVM) are adjusted through cross-validation to ensure the good performance of the model on the near infrared spectrum data test set and prevent overfitting. The support vector machine model is trained based on this training method, and the trained support vector machine model is determined as a discriminant model.

[0118] In this embodiment, during the training of the random forest model using the near infrared spectroscopy data training set, multiple random samplings can be used to obtain different training subsets, and a decision tree model can be trained on each subset. During the construction of each decision tree, some features are randomly selected for division to ensure the diversity of the model. By comparing the classification results of multiple decision trees, the most likely classification result of the oil model is finally output. The random forest model is trained based on this training method, and the trained random forest model is determined as a discriminant model.

[0119] In this embodiment, in the process of training the convolutional neural network model through the near-infrared spectral data training set, a one-dimensional convolutional neural network is constructed, and feature extraction is performed on the near-infrared spectral data of the samples in the near-infrared spectral data training set through multiple convolutional layers, and each convolutional layer extracts high-order features of the local area. The convolution result is downsampled by a pooling layer to reduce the data dimension and enhance the generalization ability of the model. Finally, the mapping relationship between the spectral features and the oil model classification is realized through a fully connected layer, and the classification result is output. The convolutional neural network model is trained based on this training method, and the trained convolutional neural network model is determined as a discriminant model.

[0120] Step 3: Use a variety of evaluation indicators to evaluate the performance of each discriminant model on the near-infrared spectral data test set and determine the evaluation results of each discriminant model.

[0121] The evaluation index may include the recognition accuracy, precision and F1 value of the discriminant model, etc. Each discriminant model is tested by using a near infrared spectroscopy data test set to determine the evaluation index value of each discriminant model.

[0122] In this embodiment, each evaluation index may be assigned a weight, and for any discriminant model, the values ​​of each evaluation index may be weighted averaged according to each weight, and the value obtained by the weighted average may be determined as the evaluation result of the discriminant model.

[0123] Step 4: Determine the discriminant model with the best evaluation result as the oil type recognition model.

[0124] In this embodiment, the oil type identification model includes multiple discriminant models trained by different machine learning models;

[0125] In an exemplary embodiment, after obtaining the discriminant models trained by the support vector machine model, the random forest model and the convolutional neural network model, the processed near-infrared spectral data can be respectively input into each discriminant model to obtain multiple oil model identification results, and the model identification result of the oil to be identified can be determined based on the multiple oil model identification results.

[0126] In this embodiment, the oil model identification result is the oil model of the oil to be identified; after obtaining the oil model identification result output by each discrimination model, the number of each oil model in the oil model identification result is counted, and the oil model with the largest number is determined as the model identification result of the oil to be identified.

[0127] Among them, each oil model has a priority. For example, if there is at least one oil model whose quantity is the same and the largest, the model identification result is determined according to the priority; for example, the priority of oil model A is greater than the priority of oil model B. When the quantity of oil model A and oil model B is the same and the largest, oil model A is determined as the model identification result.

[0128] In summary, the present invention proposes a transformer oil model identification method, which includes obtaining first data of the oil to be identified; performing a first preprocessing on the first data to obtain second data after the first preprocessing; inputting the second data into a pretrained oil model identification model to obtain a model identification result of the oil to be identified, wherein the pretrained oil model identification model is an arbitrary model that inputs the second data and outputs the oil model identification result. Compared with the prior art, which requires the oil to be identified to be sent to a professional chemical laboratory for analysis before the oil model can be determined, the identification method of the present invention can quickly identify the model of the oil to be identified.

[0129] This embodiment also provides a transformer oil model identification system, including:

[0130] An acquisition module, used for acquiring first data of the oil product to be identified;

[0131] A processing module, used for performing a first preprocessing on the first data to obtain second data after the first preprocessing;

[0132] The identification module is used to input the second data into a pre-trained oil product model identification model to obtain a model identification result of the oil product to be identified. The pre-trained oil product model identification model is any model whose input is the second data and whose output is the oil product model identification result.

[0133] The above-mentioned unit modules may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above-mentioned modules.

[0134] This embodiment also provides a computer device, which may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 6As shown. 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 operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a transformer oil model identification method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0135] This embodiment further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0136] Acquire first data of the oil product to be identified;

[0137] Performing a first preprocessing on the first data to obtain second data after the first preprocessing;

[0138] The second data is input into a pre-trained oil product model recognition model to obtain a model recognition result of the oil product to be identified. The pre-trained oil product model recognition model is any model whose input is the second data and whose output is the oil product model recognition result.

[0139] Example 2

[0140] Reference Figure 2-Figure 4 , which is an embodiment of the present invention, provides a transformer oil model identification method and system. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0141] In an exemplary embodiment, different types of transformer oil samples are accumulated, and three oil samples are selected, namely mineral insulating oil KI25, KI45 and synthetic ester insulating oil KI50EX. 20 samples of each oil are collected. In this embodiment, considering the type identification of the oil samples in operation, the oil samples obtained by accelerated thermal aging experiment to simulate the real aging conditions are used.

[0142] Spectral acquisition: The transmission spectrum acquisition method is used for near-infrared spectrum acquisition. During spectrum acquisition, the insulating oil is placed in a 20×10×50mm quartz cuvette, and the wide side of the sample pool is JGS3 near-infrared high-transmittance glass with a thickness of 1.25mm. The near-infrared spectrometer used in this embodiment has a wavelength range of 900-2200nm, a spectrometer resolution of 5nm, a signal-to-noise ratio of 10000:1, and a detector pixel number of 512. The integration time of the spectrum acquisition is set to 100ms, and the average number is 12 times. The accuracy and representativeness of the sample data set are guaranteed by taking the average value through multiple measurements. The spectrum acquisition method of insulating oil is as follows Figure 2 The spectra of the three oil samples collected are shown in Figure 3 As shown, the spectra of the two mineral oils KI25 and KI45 are slightly different, with only slight differences at 1550nm and 1950nm, but the spectrum of the synthetic ester oil KI50EX is significantly different from the two, which is due to its different microscopic composition.

[0143] After the spectrum is collected, the spectrum is denoised and baseline corrected by moving average smoothing and differential derivation, and finally the spectrum data set Q = {S1, S2, ..., S M}, where Si = [D x1i ,D x2i ,…,D xNi ], M is the total number of samples; then the spectral data set is subjected to spectral data dimension reduction. The spectrum after denoising and baseline correction is as follows Figure 4 It should be noted that the method of denoising, baseline correction and dimensionality reduction of the spectrum is the same as that in the above embodiment, and will not be described in detail in this embodiment.

[0144] 80% of the data set was randomly selected as the training set and 20% of the data was used as the test set. In order to achieve the best results, multiple machine learning models were tried and the optimal model was selected through model optimization. The specific steps are as follows:

[0145] Support vector machine model: The extracted spectral feature data is normalized to ensure that the magnitude of all features is the same. The radial basis function is selected as the kernel function to map the original spectral data to a high-dimensional space in order to better separate different oil models; the penalty parameter C and the kernel function parameter γ in the SVM are adjusted through cross-validation to ensure good performance of the model on the test data and prevent overfitting. For this embodiment, C is 21.62 and γ is 0.035.

[0146] Random forest model: Use multiple random sampling to obtain different training subsets, and train a decision tree model on each subset. During the construction process of each decision tree, some features are randomly selected for division to ensure the diversity of the model. For this embodiment, the number of decision trees is 20 and the tree depth is 5; by comparing the classification results of multiple decision trees, the most likely oil model classification result is finally output.

[0147] Convolutional neural network: A one-dimensional convolutional neural network is constructed to extract features from spectral data through multiple convolutional layers. Each convolutional layer extracts high-order features of the local area. For this embodiment, two convolutional layers are used and five fully connected layers are used. The pooling layer is used to downsample the convolution results to reduce the data dimension and enhance the generalization ability of the model. Finally, the mapping relationship between the spectral features and the oil model classification is realized through the fully connected layer, and the classification results are output.

[0148] By training the above-mentioned multiple models, for this embodiment, the performance of each model on the test set is evaluated using the classification accuracy, and the test set accuracy is shown in Table 1.

[0149] Table 1

[0150] Model Accuracy Misclassification Support Vector Machine Model 91.7％ The actual value is KI25, but was mistakenly classified as KI45 Random Forest Model 100％ No error Convolutional Neural Network Model 83.3％ Two samples that were actually KI25 were mistakenly classified as KI45

[0151] Therefore, for this embodiment, random forest is selected as the optimal model F(·), that is, the oil type recognition model.

[0152] For transformer oil samples of unknown type, spectrum acquisition, noise reduction and baseline correction are performed according to the above method to obtain spectrum Stest, and data dimension reduction is performed using transformation matrix P to obtain spectrum data Stest* after dimension reduction. Finally, the optimal model F(Stest*) is used to distinguish the category of unknown type samples. For this embodiment, a KI25, a KI45 and a KI50EX are selected as unknown samples for testing. The spectra of the three are as follows: Figure 4 The discrimination results are shown in Table 2.

[0153] Table 2

[0154] True category Prediction Category KI25 KI25 KI45 KI45 KI50EX KI50EX

[0155] The present invention provides a new method for determining the type of transformer oil, which proposes to use the near-infrared spectral characteristics of insulating oil to determine the type of insulating oil. The near-infrared spectral characteristics of insulating oil contain information about its microscopic molecular composition, and have outstanding advantages in type identification, such as fast, non-destructive, and low operating difficulty.

[0156] Secondly, the moving average smoothing and differential derivation method is proposed to remove spectral noise and baseline effects, making the effective spectral information more prominent, easier to distinguish and model, and improving the accuracy of oil type recognition. In addition, the use of multiple model competition method to select the optimal discrimination model can further ensure the effectiveness of the final model.

[0157] It should be noted that, in addition to distinguishing the type of new oil, it is also possible to distinguish the type of transformer oil that has been aged, dampened, or discharged, all it takes is to establish a comprehensive spectral database.

[0158] The present invention also provides a corresponding transformer oil type identification device based on near infrared spectroscopy, such as Figure 5 shown.

[0159] Figure 5 A schematic diagram of a transformer oil model identification device based on near infrared spectroscopy provided by the present invention, the device 800 comprises:

[0160] An acquisition module 801 is used to acquire near infrared spectrum data of the oil product to be identified;

[0161] The processing module 802 is used to perform data cleaning processing on the near infrared spectrum data to obtain processed near infrared spectrum data;

[0162] The identification module 803 is used to input the processed near-infrared spectral data into the oil type identification model to obtain the type identification result of the oil to be identified; the oil type identification model is pre-trained with the sample near-infrared spectral data of sample oils of various types.

[0163] The specific definition of the transformer oil model identification device based on near infrared spectroscopy can be found in the above definition of the transformer oil model identification method based on near infrared spectroscopy, which will not be repeated here. Each module in the above-mentioned transformer oil model identification device based on near infrared spectroscopy can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0164] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should be included in the scope of the claims of the present invention.

[0165] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0166] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0167] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0168] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0169] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0170] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for identifying transformer oil type, characterized in that: include: Acquire first data of the oil product to be identified; Performing a first preprocessing on the first data to obtain second data after the first preprocessing; The second data is input into a pre-trained oil product model recognition model to obtain a model recognition result of the oil product to be identified. The pre-trained oil product model recognition model is an arbitrary model whose input is the second data and whose output is the oil product model recognition result.

2. The transformer oil model identification method according to claim 1, characterized in that: The pre-trained oil product type recognition model includes: a number of different discrimination models; The number of the discrimination models depends on the number of the first data of the oil product to be identified; The first data of any oil product to be identified includes several.

3. The transformer oil model identification method according to claim 2, characterized in that: The inputs of the several different discrimination models are all the second data, and the outputs are all the oil product model recognition results; The inputs of the different discriminant models are respectively second data after first preprocessing is performed on the first data; And the number of the several different discriminant models is the same as the number of the second data after the first preprocessing.

4. The transformer oil model identification method according to claim 3, characterized in that: The performing first preprocessing on the first data comprises: Performing denoising processing on the first data to obtain denoised first data; Performing baseline correction on the denoised first data to obtain corrected first data; Performing data dimension reduction on the corrected first data to obtain second data after the first preprocessing.

5. The transformer oil model identification method according to claim 4, characterized in that: The performing data dimension reduction on the corrected first data to obtain the second data after the first preprocessing comprises: Acquire a covariance matrix of the corrected first data; Calculating the eigenvalues ​​and eigenvectors of the covariance matrix; Calculate the variance contribution rate according to the eigenvalues, and use the eigenvectors corresponding to the first preset number of eigenvalues ​​whose cumulative contribution rates exceed the preset ratio as the conversion matrix; The product of the conversion matrix and the corrected first data is determined as the first preprocessed second data.

6. The transformer oil model identification method according to claim 5, characterized in that: The step of inputting the second data into a pre-trained oil product model recognition model to obtain the model recognition result of the oil product to be recognized includes: Inputting the second data into a plurality of different discriminant models to obtain a plurality of different model outputs; Using a variety of evaluation indicators to perform performance evaluation on the outputs of each of the several different models, and determining the evaluation result of each discriminant model; The discriminant model with the best evaluation result is determined as the oil product type recognition model.

7. The transformer oil model identification method according to claim 6, characterized in that: The oil product to be identified is a sample of one or several different oil samples.

8. A transformer oil model identification system, characterized in that: include: An acquisition module, used for acquiring first data of the oil product to be identified; A processing module, used for performing a first preprocessing on the first data to obtain second data after the first preprocessing; The identification module is used to input the second data into a pre-trained oil model identification model to obtain the model identification result of the oil to be identified. The pre-trained oil model identification model is any model whose input is the second data and whose output is the oil model identification result.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images