Transformer drying time sequence prediction method and system

Abstract

The invention belongs to the technical field of transformer manufacturing, and provides a transformer drying time sequence prediction method and system, and the method comprises the steps: obtaining transformer drying whole process information from a plurality of sources to form an original data set, obtaining corresponding classification information, and carrying out the classification and superior product screening of the original data set, obtaining a superior product data set and carrying out data processing and integration to obtain a learning data set comprising an input feature set, a drying time sequence vector set and an energy consumption data set, constructing a drying prediction model based on a hybrid neural network, training the drying prediction model by adopting the learning data set, and judging whether the prediction accuracy of the drying prediction model reaches the standard or not; and if the drying process parameters reach the standard, reasoning prediction is carried out to obtain a predicted drying time sequence vector and predicted energy consumption data, whether the drying process parameters need to be optimized or not is judged, and if yes, collaborative optimization and adjustment based on a prediction result are carried out on the drying process parameters through a self-adaptive control algorithm in combination with the drying prediction model.

Description

Technical Field

[0001] The present invention belongs to the technical field of transformer manufacturing, and in particular relates to a transformer drying time sequence prediction method and system. Background Art

[0002] In the field of power equipment manufacturing, transformer drying is a critical step in ensuring stable performance and long-term reliable operation. During the transformer manufacturing process, the insulation material will contain a certain amount of moisture. If this moisture is not completely removed, it will seriously affect the transformer's insulation performance and service life, and even the safe and stable operation of the power system. Therefore, accurately predicting the transformer drying sequence and rationally controlling the drying process are crucial to improving drying quality, reducing energy consumption, and increasing production efficiency.

[0003] During actual production, technicians discovered that existing transformer drying timing prediction methods have many shortcomings. On the one hand, traditional methods often rely on a single data source or limited drying process information, making it difficult to fully cover the various influencing factors in the entire transformer drying process, such as the structural differences between different types of transformers, the characteristics of insulation materials, the operating status of drying equipment, and environmental conditions. This leads to inaccurate predictions of drying timing, and drying time is often too long or too short. Too long may increase energy consumption and production cycle, while too short may lead to incomplete drying, affecting product quality.

[0004] On the other hand, the existing technology lacks an effective classification and screening mechanism when processing data from the drying process. A large amount of low-quality or irrelevant data enters the model, which not only increases the difficulty and cost of data processing, but also reduces the reliability of the prediction model. At the same time, traditional prediction models mostly use a single neural network or other simple algorithms, which have insufficient ability to fit the nonlinear characteristics of complex drying processes and cannot adapt to the dynamic changes of different drying stages, resulting in prediction accuracy that is difficult to meet actual production needs. In addition, in terms of adjusting drying process parameters, existing methods are usually based on experience or fixed rules and cannot be adaptively optimized according to real-time prediction results, making it difficult to achieve refined control of the drying process.

[0005] In response to the above problems, the present invention proposes a transformer drying timing prediction method and system. Summary of the Invention

[0006] In order to address the deficiencies in the prior art and solve at least one of the technical problems raised in the background art, the present invention provides a transformer drying timing prediction method and system.

[0007] The technical solution adopted by the present invention to solve the technical problem is: a transformer drying time sequence prediction method, comprising: Obtain the entire transformer drying process information from multiple sources to form an original data set. Then, classify and screen the original data set based on the corresponding classification information to obtain a high-quality data set. The high-quality product data set is processed and integrated to obtain a learning data set including an input feature set, a drying time series vector set, and an energy consumption data set; A drying prediction model is constructed based on a hybrid neural network. The drying prediction model is trained using a learning dataset and its prediction accuracy is judged to be up to standard. If it is up to standard, the drying prediction model is used for inference prediction before the current drying process starts, and the predicted drying time series vector and predicted energy consumption data are obtained as the prediction results. Determine whether the drying process parameters need to be optimized based on the prediction results. If necessary, use the adaptive control algorithm combined with the drying prediction model to collaboratively optimize and adjust the drying process parameters based on the prediction results. The method for obtaining the high-quality product data group is as follows: Leveraging SCADA and IoT technologies, during the data collection phase, the data acquisition module acquires information about the entire transformer drying process and corresponding classification information from multiple sources across the production architecture. Data acquisition interface programs are then configured for each source, capable of parsing data in different formats. These data are then converted into a unified format, resulting in the original data set from the data collection period. This data set is then divided into drying data groups based on the corresponding classification information, and filtered based on quality criteria to obtain the quality data group. The prediction result is obtained as follows: Before the current drying process starts, the basic data of transformer accessories, drying process parameters, and environmental data of the transformer to be dried are collected to obtain a collection feature group. The original dimensions and normalized parameters recorded in the data normalization module are obtained to normalize the collection feature group to obtain an input feature vector. The input feature vector is input into a drying prediction model that meets the accuracy standard to obtain a predicted drying time series vector and predicted energy consumption data as the prediction results. The method for determining whether the accuracy meets the standard is as follows: The learning data set is divided into a training set and a test set. The training set is used to train the constructed drying prediction model according to the set parameters. After the training is completed, the drying prediction model is verified using the test set. The determination coefficient and mean square error of the drying prediction model in the test set are used as indicators to evaluate the prediction accuracy of the drying prediction model. If the calculated determination coefficient is greater than 0.9 and the mean square error is less than the preset error threshold, the prediction accuracy of the drying prediction model is judged to meet the standard. The drying prediction model is constructed as follows: A drying prediction model is constructed using a hybrid neural network architecture, which includes an input layer, a feature fusion layer, a time series processing layer, and a fully connected output layer. The input layer receives the input feature set, the feature fusion layer uses a fully connected neural network, and the time series processing layer introduces a long short-term memory network unit based on the time series characteristics of the drying time series. The fully connected output layer contains two branches, which output the drying time series vector and energy consumption data respectively. The method for determining whether the drying process parameters need to be optimized is as follows: Set expected conditions, obtain predicted energy consumption data, calculate the deviation between the predicted energy consumption data and the preset expected energy consumption data, and if the deviation is greater than the allowable deviation threshold, determine that the predicted energy consumption data does not meet the expected conditions; If the predicted drying time series vector does not meet the expected conditions, or the predicted energy consumption data does not meet the expected conditions, it is judged that the prediction result does not meet the expected conditions, and the drying process parameters need to be optimized; The judgment method of whether the predicted drying time series vector does not meet the expected conditions is as follows: Obtain the predicted drying time series vector and the preset expected drying time series vector, calculate the predicted total drying time and the expected total drying time respectively, if the predicted total drying time is greater than the expected total drying time, calculate the Euclidean distance between the predicted drying time series vector and the expected drying time series vector, and compare it with the preset allowable deviation threshold, if it is greater than the allowable deviation threshold, determine that the predicted drying time series vector does not meet the expected conditions; The collaborative optimization and adjustment method is as follows: Combined with the drying prediction model, the particle swarm optimization algorithm is used as the adaptive control algorithm. By continuously updating the particle position and particle velocity until the maximum number of iterations is reached or the fitness function converges, the drying process parameters of the global optimal solution are obtained. The original dimensions and normalized parameters recorded in the data normalization module are obtained, and the drying process parameters and the predicted results are denormalized to obtain the actual drying process parameters and send them to the controller to complete the optimization and adjustment. The fitness function is obtained as follows: The particle position is used as the input feature vector for changing the drying process parameters. Each particle position represents a set of candidate solutions for the drying process parameters. The changed input feature vector is input into the drying prediction model to obtain the predicted drying time series vector and predicted energy consumption data. Based on the Euclidean distance between the predicted drying time series vector and the expected drying time series vector, as well as the deviation between the predicted energy consumption data and the expected energy consumption data, the fitness function is constructed through fusion calculation.

[0008] A transformer drying time sequence prediction system, comprising: Drying Dataset Subsystem: This includes a data acquisition module, a data cleaning module, a data normalization module, and a data set construction module. This module acquires information about the entire transformer drying process from multiple sources to form a raw data set. It then classifies and selects high-quality data sets based on the corresponding classification information to obtain high-quality data sets. This high-quality data set is then processed and integrated to obtain a learning data set containing an input feature set, a drying time series vector set, and an energy consumption data set. Drying prediction model subsystem: This includes a model construction module, a model training module, a prediction evaluation module, and a parameter optimization module. It builds a drying prediction model based on a hybrid neural network, uses a learning data set to train the drying prediction model, and determines whether the prediction accuracy of the drying prediction model meets the standard. If so, the drying prediction model is used to perform inference prediction before the current drying process starts, and the predicted drying time series vector and predicted energy consumption data are obtained as prediction results. Based on the prediction results, it is determined whether the drying process parameters need to be optimized. If necessary, the drying process parameters are collaboratively optimized and adjusted based on the prediction results through the adaptive control algorithm combined with the drying prediction model. Drying model interface subsystem: includes the interface between the drying prediction model and the equipment, the SCADA interaction interface, and the MES interaction interface to achieve precise interaction.

[0009] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention effectively improves data quality by acquiring full-process information from multiple sources and screening high-quality data groups, providing a reliable foundation for subsequent model training. The drying prediction model constructed based on the hybrid neural network can fully utilize the advantages of different neural networks to improve prediction accuracy. It can perform inference prediction before the drying process starts, and obtain drying time series vectors and energy consumption data in advance, which helps to plan production arrangements in advance, arrange resources reasonably, and avoid production delays or resource waste due to insufficient estimation of drying time or energy consumption.

[0010] 2. The present invention determines whether to optimize the drying process parameters based on the prediction results, and uses an adaptive control algorithm combined with a prediction model to perform collaborative optimization and adjustment, thereby realizing intelligent control of the drying process. This dynamic optimization method can accurately adjust the process parameters according to the drying requirements and actual conditions of different transformers, improve drying efficiency, and reduce energy consumption. At the same time, it helps to improve the drying quality of transformers and reduce problems such as transformer performance degradation caused by improper drying processes, bringing significant economic benefits and quality improvements to enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a flowchart of the steps of the transformer drying timing prediction method according to an embodiment of the present invention; Figure 2This is a system module architecture diagram of the transformer drying timing prediction system described in an embodiment of the present invention. DETAILED DESCRIPTION

[0012] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0013] Example 1 like Figure 1 As shown, a transformer drying time sequence prediction method according to an embodiment of the present invention includes the following steps: S1: Obtain the transformer drying process information from multiple sources to form an original data set, obtain and classify the original data set according to the corresponding classification information, and select the best products to obtain the best product data set; Leveraging SCADA and IoT technologies, during the data collection phase, the data acquisition module acquires full transformer drying process information and corresponding classified information from multiple sources across the production architecture. Data indicators for the full transformer drying process include basic transformer component data, drying process parameters, environmental data, drying timing, and energy consumption data. Corresponding classified information includes the corresponding drying tank model and finished product experimental data. Among them, the basic data of transformer accessories includes coil quality data and core quality data. The coil quality data includes the coil insulation layer thickness and coil resistance value. The coil insulation layer thickness is measured by installing a high-precision coil insulation layer thickness detector on the coil production line and using non-contact measurement technology to measure the insulation layer thickness of each coil. The coil resistance value is obtained by applying a specific voltage and measuring the resistance of the coil using a resistance meter. The core quality data includes the core magnetic permeability and the core lamination gap. The core magnetic permeability is calculated by using a core magnetic permeability tester to test the magnetic permeability of the core under different magnetic field strengths. The core lamination gap is measured optically using a lamination gap measuring device. The drying process parameters include drying temperature and drying vacuum, which are obtained by recording the drying process parameters set in the drying tank during the transformer drying process. The environmental data include ambient temperature and ambient humidity, which are collected by setting temperature and humidity sensors in the environment where the drying tank is located. The drying sequence includes the duration of each drying stage during the transformer drying process in the drying tank. The drying stages include the preheating stage, the heating drying stage, the constant temperature drying stage, and the cooling drying stage. The drying sequence is obtained by automatically recording the timestamps at the start and end of each drying stage through the timing module. Energy consumption data is collected by installing a high-precision electricity meter on the power supply line of the drying equipment. The finished product test data includes the breakdown voltage and dielectric loss of transformer accessories. This data is obtained by taking a number of transformer accessory samples from the same production batch and performing experimental measurements after the drying process, and then combining the measurements with calculations. The calculated finished product test data is the finished product test data of all corresponding transformer accessories in the same production batch. Since the transformer drying process information production system architecture is collected from different sources of the production architecture, corresponding data acquisition interface programs that can parse data in different formats are set for different sources, and the data in different formats are converted into a unified data format. The converted transformer drying process information is integrated through the data transmission network to obtain the original data set within the data collection period. Based on the drying tank model corresponding to each transformer drying process information in the original data set, the transformer drying process information of the same drying tank model is divided into the same drying data group, thereby obtaining several drying data groups; In each drying data group, high-quality products are screened based on the finished product test data corresponding to the transformer drying process information. Specifically, according to the company's production standards and product quality requirements, high-quality product judgment criteria are set, including high-quality product breakdown voltage and high-quality product dielectric loss. For each transformer drying process information, the corresponding finished product test data is compared with the high-quality product judgment criteria; If the breakdown voltage corresponding to the transformer drying process information is greater than the breakdown voltage of a good product, and the corresponding dielectric loss is less than the dielectric loss of a good product, then it is determined that the good product judgment standard is met, and the transformer drying process information is marked as good product data; In each drying data group, only all high-quality data are retained to obtain a high-quality data group corresponding to each drying data group; It should be noted that the purpose of this step is to collect information on the entire transformer drying process through multiple sources, covering multi-dimensional indicators such as basic data of accessories, process parameters, and environmental data, to ensure the comprehensiveness and integrity of the data, and to combine SCADA and Internet of Things technologies to achieve real-time and accurate data collection. Data classification and quality screening based on classified information effectively eliminates unqualified data, improves data quality, provides a reliable data foundation for subsequent analysis, reduces prediction errors caused by data problems, and ensures that the data used for analysis are all high-quality data, so that the subsequent models built based on this data and the predictions made are more in line with actual production needs, which helps to improve production efficiency, reduce scrap rates, and reduce resource waste; S2: Process and integrate the high-quality product data set to obtain a learning data set including an input feature set, a drying time series vector set, and an energy consumption data set; For the obtained high-quality product data group, the high-quality product data group is cleaned by the data cleaning module; Specifically, the data cleaning module uses a median filter algorithm to remove noise, outliers, and missing values ​​from all transformer drying process information in the premium data set. It also uses a density-based local outlier factor (LOF) algorithm to identify outliers and detect abnormal data. The detected abnormal data and missing values ​​are supplemented with a neighboring value filling method. For the high-quality data group after cleaning, the high-quality data group is normalized through the data normalization module; Specifically, the data normalization module uses the maximum-minimum normalization method to map each data indicator in the full-process information of transformer drying in the premium data group to the [0,1] interval. After normalization, the original dimension information and normalization parameters of each data indicator are recorded to ensure the accuracy of the normalization process. The cleaned and normalized high-quality data set is marked as a normalized high-quality data set, and the normalized high-quality data set is integrated into a learning data set including an input feature set, a drying time series vector set, and an energy consumption data set through a data set construction module; Specifically, the full-process information of each transformer drying process in the normalized high-quality data group is integrated respectively. Based on any full-process information of the transformer drying process, the basic data of the transformer accessories, drying process parameters and environmental data in the full-process information of the transformer drying process are integrated into the input feature vector, the drying time series in the full-process information of the transformer drying process is integrated into the drying time series vector, the duration of each drying stage in the drying time series vector is sorted according to the time series, the drying time series vector and the energy consumption data are used as output features, all the input feature vectors, drying time series vectors and energy consumption data are integrated in the same order according to the one-to-one correspondence to obtain the input feature group, the drying time series vector group and the energy consumption data group, that is, the learning data set containing the input feature group, the drying time series vector group and the energy consumption data group is obtained; It should be noted that the purpose of this step is to comprehensively apply multiple data processing algorithms to form a complete data processing process, adopt corresponding solutions for different types of data problems, record the original dimensional information and normalization parameters during the normalization process, provide a basis for the denormalization of the data, ensure the traceability and accuracy of the data before and after processing, and the processed data can better adapt to subsequent model training, so that the model can more accurately learn the laws in the data and improve the prediction accuracy of the model; S3: A drying prediction model is constructed based on a hybrid neural network. The drying prediction model is trained using a learning dataset and its prediction accuracy is determined to be up to standard. If not, the learning dataset is supplemented. If so, the drying prediction model is used for inference prediction before the current drying process starts to obtain a predicted drying time series vector and predicted energy consumption data. Build a drying prediction model, which uses a hybrid neural network architecture, including an input layer, a feature fusion layer, a time series processing layer, and a fully connected output layer; Specifically, the input layer is responsible for receiving the input feature group. The number of neurons in the input layer is consistent with the input feature dimension of the input feature group. Since the input feature vector in the input feature group contains basic data of transformer accessories, drying process parameters, and environmental data, where the basic data of transformer accessories includes coil insulation layer thickness, coil resistance value, core magnetic permeability, and core lamination gap, the drying process parameters include drying temperature and drying vacuum degree, and the environmental data includes ambient temperature and ambient humidity, the input feature dimension of the input feature group is set to 8, and the number of neurons in the input layer is 8; The feature fusion layer uses a fully connected neural network (FCN) to perform deep feature extraction on the input feature group. FCN learns the nonlinear relationship between different data indicators through the full connection between multiple layers of neurons, and maps the input 8-dimensional input feature vector to a 16-dimensional feature vector. The time series processing layer introduces a long short-term memory network (LSTM) unit to address the time series characteristics of the drying time series. The hidden layer dimension is set to 32 to capture the time series dependency between the durations of each drying stage. The fully connected output layer contains two branches, which output the dry time series vector and energy consumption data ; ; in, 、 、 、 They represent the duration of the preheating stage, the heating drying stage, the constant temperature drying stage, and the cooling drying stage, respectively. i represents the drying time series vector and energy consumption data The number in the training set uses the mean square error (MSE) as the loss function, and the formula is: ; Where N is the number of training samples in the training set, and They are respectively the predicted drying time series vector and the predicted energy consumption data obtained by the pre-inference prediction of the drying prediction model; Based on any learning data set, the learning data set is divided into training set and test set in a ratio of 8:2. The Adam optimizer is used to set parameters such as learning rate, batch size, and training cycle. The network weights are updated in each round of training. ; ; Among them, j represents the current training round, represents the learning rate, represents the gradient of the loss function; The drying prediction model was trained using the training set according to the set parameters, and then validated using the test set after training. The coefficient of determination (R²) and mean square error (RMSE) of the drying prediction model in the test set were used as indicators to evaluate the prediction accuracy of the drying prediction model. It should be noted that the coefficient of determination R² reflects the ability of the drying prediction model to explain data variation. The closer the coefficient of determination R² is to 1, the better the drying prediction model fits the data. RMSE measures the average error between the predicted results and the true values. The smaller the RMSE value, the more accurate the prediction results. Set the prediction accuracy compliance conditions. If the calculated determination coefficient is greater than 0.9 and the mean square error is less than the preset error threshold, the drying prediction model meets the prediction accuracy compliance conditions and is judged to be qualified. Otherwise, the drying prediction model is judged to be unqualified. If the prediction accuracy of the drying prediction model does not meet the standard, the data supplement mechanism is triggered. By expanding the time range of the data collection phase, the high-quality data group is supplemented, and the supplemented high-quality data group is used to construct a learning data set to retrain the drying prediction model; If the prediction accuracy of the drying prediction model meets the requirements, the prediction analysis phase will begin; Before the current drying process starts, the basic data of transformer accessories, drying process parameters, and environmental data of the transformer to be dried are collected to obtain a collection feature group. The original dimensions and normalization parameters recorded in the data normalization module are used to normalize the collection feature group to obtain an input feature vector. The input feature vector is input into the drying prediction model to obtain a predicted drying time series vector and predicted energy consumption data as the prediction results. It should be noted that the purpose of this step is to build a drying prediction model based on a hybrid neural network. Combining the advantages of a fully connected neural network (FCN) and a long short-term memory network (LSTM), it can effectively extract data features, capture the temporal dependencies of drying time series, and achieve accurate prediction of drying time series vectors and energy consumption data. It adopts scientific training and evaluation methods, divides the training set and test set, uses the Adam optimizer for training, and uses the coefficient of determination R² and mean square error RMSE as evaluation indicators to ensure the model's prediction accuracy and generalization ability. If the model does not meet the standards, it will be retrained with supplementary data to continuously optimize the model performance. An accurate drying prediction model can provide enterprises with predicted drying time series vectors and predicted energy consumption data before the drying process starts, helping enterprises to plan production processes in advance, arrange resources reasonably, and reduce production costs. Through continuous optimization of the model, the applicability of the model in the actual production environment is improved, providing a scientific basis for the company's production decisions. S4: Determine whether the drying process parameters need to be optimized based on the prediction results. If necessary, optimize and adjust the drying process parameters based on the prediction results through the adaptive control algorithm combined with the drying prediction model. Set expected conditions, including expected drying time series vector, expected energy consumption data and allowable deviation threshold. The expected conditions are combined with the original dimensions, normalization parameters and production expectation settings recorded in the data normalization module to determine whether the prediction results meet the expected conditions. Specifically, based on the obtained predicted drying time series vector and the expected drying time series vector, the predicted total drying time and the expected total drying time are calculated respectively. If the predicted total drying time is greater than the expected total drying time, the Euclidean distance Dt between the predicted drying time series vector and the expected drying time series vector is calculated and compared with a preset allowable deviation threshold. If the calculated Euclidean distance is greater than the allowable deviation threshold, it is determined that the predicted drying time series vector does not meet the expected conditions. On the contrary, if the Euclidean distance is less than or equal to the allowable deviation threshold, or the predicted total drying time is less than or equal to the expected total drying time, it is judged that the predicted drying time series vector meets the expected conditions; Based on the obtained predicted energy consumption data, the difference between the predicted energy consumption data and the expected energy consumption data is calculated as the deviation De. If the deviation is greater than the allowable deviation threshold, it is judged that the predicted energy consumption data does not meet the expected conditions. Conversely, if the deviation is less than or equal to the allowable deviation threshold, it is judged that the predicted energy consumption data meets the expected conditions. If both the predicted drying time series vector and the predicted energy consumption data meet the expected conditions, the prediction result is judged to meet the expected conditions, and the prediction result is denormalized based on the original dimension and normalization parameters recorded in the data normalization module to obtain the actual predicted drying time series and the actual predicted energy consumption data; If the predicted drying time series vector does not meet the expected conditions, or the predicted energy consumption data does not meet the expected conditions, it is judged that the prediction result does not meet the expected conditions, and the drying process parameters need to be optimized. Adaptive control algorithm is used in combination with the drying prediction model to coordinately optimize and adjust the drying process parameters. Specifically, the particle swarm optimization (PSO) algorithm is used as the adaptive control algorithm. In the particle swarm optimization algorithm, each particle position represents a set of candidate solutions for drying process parameters. ; ; in is the normalized drying temperature, is the normalized drying vacuum degree; The particle position As the input feature vector of drying process parameter change, the changed input feature vector is input into the drying prediction model to obtain the predicted drying time series vector and predicted energy consumption data. The fitness function is constructed based on the Euclidean distance Dt between the predicted drying time series vector and the expected drying time series vector, and the deviation De between the predicted energy consumption data and the expected energy consumption data; The particle velocity update formula is: ; in Indicates the particle velocity at the current step k, represents the particle position at the current step k, is the inertia weight, which is used to balance the global search and local search capabilities. and is the acceleration factor, controlling the particles to move towards the individual optimal solution and the global optimal solution The step length of the movement, and A random number in the interval [0,1] increases the randomness of the search; The particle swarm optimization algorithm continuously updates the particle position and particle velocity until the maximum number of iterations is reached or the fitness function converges, obtaining the drying process parameters of the global optimal solution. The drying process parameters of the input feature vector are changed to the drying process parameters of the global optimal solution. The changed input feature vector is input into the drying prediction model to obtain the predicted drying time series vector and predicted energy consumption data as the prediction results. Based on the original dimensions and normalized parameters recorded in the data normalization module, the drying process parameters and the prediction results are denormalized to obtain the actual drying process parameters, the actual predicted drying time series, and the actual predicted energy consumption data. The actual drying process parameters are sent to the controller to complete the optimization and adjustment. It should be noted that the role of this step is to strictly judge the prediction results by setting expected conditions, and to accurately identify whether the prediction results meet production expectations. When the prediction results do not meet expectations, the particle swarm optimization (PSO) algorithm is combined with the drying prediction model to collaboratively optimize and adjust the drying process parameters. The process parameters can be dynamically adjusted according to actual production needs, making the drying process more efficient and energy-saving, improving product quality, and reducing energy consumption costs. In actual production, this step can timely discover possible problems in the production process and solve them by optimizing and adjusting the process parameters. The optimized process parameters can make the drying process more in line with production requirements, improve production efficiency, reduce energy consumption, reduce production costs, and enhance the economic benefits and market competitiveness of the enterprise. The technical solution of an embodiment of the present invention is: obtaining transformer drying process information from multiple sources to form an original data set, obtaining and classifying the original data set and screening for high-quality products according to corresponding classification information to obtain a high-quality data group, performing data processing and integration on the high-quality data group to obtain a learning data set including an input feature group, a drying time series vector group and an energy consumption data group, constructing a drying prediction model based on a hybrid neural network, using the learning data set to train the drying prediction model, and judging whether the prediction accuracy of the drying prediction model meets the standard; if not, supplementing the learning data set; if so, performing inference prediction through the drying prediction model before starting the current drying process to obtain a predicted drying time series vector and predicted energy consumption data, judging whether the drying process parameters need to be optimized based on the prediction results, and if necessary, using an adaptive control algorithm in combination with the drying prediction model to collaboratively optimize and adjust the drying process parameters based on the prediction results.

[0014] Example 2 like Figure 2 As shown, a transformer drying time series prediction system according to an embodiment of the present invention includes: Drying Dataset Subsystem: This includes a data acquisition module, a data cleaning module, a data normalization module, and a data set construction module. This module acquires information about the entire transformer drying process from multiple sources to form a raw data set. It then classifies and selects high-quality data sets based on the corresponding classification information to obtain high-quality data sets. This high-quality data set is then processed and integrated to obtain a learning data set containing an input feature set, a drying time series vector set, and an energy consumption data set. Drying prediction model subsystem: This includes a model construction module, a model training module, a prediction evaluation module, and a parameter optimization module. It builds a drying prediction model based on a hybrid neural network, uses a learning data set to train the drying prediction model, and determines whether the prediction accuracy of the drying prediction model meets the standard. If so, the drying prediction model is used to perform inference prediction before the current drying process starts, and the predicted drying time series vector and predicted energy consumption data are obtained as prediction results. Based on the prediction results, it is determined whether the drying process parameters need to be optimized. If necessary, the drying process parameters are collaboratively optimized and adjusted based on the prediction results through the adaptive control algorithm combined with the drying prediction model. Drying model interface subsystem: includes the interface between the drying prediction model and the equipment, the SCADA interaction interface, and the MES interaction interface to achieve precise interaction.

[0015] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art will appreciate that the present invention is not limited to the foregoing embodiments and that various modifications and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such modifications and improvements are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A transformer drying time series prediction method, characterized by: include: Obtain the entire transformer drying process information from multiple sources to form an original data set. Then, classify and screen the original data set based on the corresponding classification information to obtain a high-quality data set. The high-quality product data set is processed and integrated to obtain a learning data set including an input feature set, a drying time series vector set, and an energy consumption data set; A drying prediction model is constructed based on a hybrid neural network. The drying prediction model is trained using a learning dataset and its prediction accuracy is judged to be up to standard. If it is up to standard, the drying prediction model is used for inference prediction before the current drying process starts, and the predicted drying time series vector and predicted energy consumption data are obtained as the prediction results. Based on the prediction results, it is determined whether the drying process parameters need to be optimized. If necessary, the drying process parameters are collaboratively optimized and adjusted based on the prediction results through the adaptive control algorithm combined with the drying prediction model.

2. The transformer drying time series prediction method according to claim 1, characterized in that: The method for obtaining the high-quality product data group is: With the help of SCADA and Internet of Things technologies, during the data collection stage, the data collection module is used to obtain the full process information and corresponding classification information of transformer drying from different sources of multiple production architectures. Corresponding data collection interface programs that can parse data in different formats are set for different sources, and the data in different formats are converted into a unified data format to obtain the original data set within the data collection period. The original data set is divided into drying data groups based on the corresponding classification information, and screened according to the quality judgment criteria to obtain the high-quality data group.

3. The transformer drying time series prediction method according to claim 1, characterized in that: The prediction results are obtained as follows: Before the current drying process starts, the basic data of transformer accessories, drying process parameters and environmental data of the transformer to be dried are collected to obtain a collection feature group. The original dimensions and normalized parameters recorded in the data normalization module are obtained to normalize the collection feature group to obtain an input feature vector. The input feature vector is input into a drying prediction model with met accuracy standards to obtain a predicted drying time series vector and predicted energy consumption data as prediction results.

4. The transformer drying time series prediction method according to claim 3, characterized in that: The method for judging whether the accuracy meets the standard is as follows: The learning data set is divided into a training set and a test set. The training set is used to train the constructed drying prediction model according to the set parameters, and the test set is used to verify the drying prediction model after the training is completed. The determination coefficient and mean square error of the drying prediction model in the test set are used as indicators to evaluate the prediction accuracy of the drying prediction model. If the calculated determination coefficient is greater than 0.9 and the mean square error is less than the preset error threshold, the prediction accuracy of the drying prediction model is judged to meet the standard.

5. The transformer drying time series prediction method according to claim 1, characterized in that: The drying prediction model is constructed as follows: A hybrid neural network architecture is used to construct a drying prediction model, which includes an input layer, a feature fusion layer, a time series processing layer, and a fully connected output layer. The input layer receives the input feature group, the feature fusion layer adopts a fully connected neural network, and the time series processing layer introduces a long short-term memory network unit based on the time series characteristics of the drying time series. The fully connected output layer contains two branches, which output the drying time series vector and energy consumption data respectively.

6. The transformer drying time series prediction method according to claim 1, characterized in that: The method for determining whether the drying process parameters need to be optimized is as follows: Set expected conditions, obtain predicted energy consumption data, calculate the deviation between the predicted energy consumption data and the preset expected energy consumption data, and if the deviation is greater than the allowable deviation threshold, determine that the predicted energy consumption data does not meet the expected conditions; If the predicted drying time series vector does not meet the expected conditions, or the predicted energy consumption data does not meet the expected conditions, it is judged that the prediction result does not meet the expected conditions, and the drying process parameters need to be optimized.

7. The transformer drying time series prediction method according to claim 6, characterized in that: The judgment method of whether the predicted drying time series vector does not meet the expected conditions is: Obtain the predicted drying timing vector and the preset expected drying timing vector, calculate the predicted total drying time and the expected total drying time respectively, if the predicted total drying time is greater than the expected total drying time, calculate the Euclidean distance between the predicted drying timing vector and the expected drying timing vector, and compare it with the preset allowable deviation threshold, if it is greater than the allowable deviation threshold, judge that the predicted drying timing vector does not meet the expected conditions.

8. The transformer drying time series prediction method according to claim 1, characterized in that: The collaborative optimization and adjustment method is as follows: Combined with the drying prediction model, the particle swarm optimization algorithm is used as the adaptive control algorithm. By continuously updating the particle position and particle velocity until the maximum number of iterations is reached or the fitness function converges, the drying process parameters of the global optimal solution are obtained. The original dimensions and normalized parameters recorded in the data normalization module are obtained, and the drying process parameters and the prediction results are denormalized to obtain the actual drying process parameters and send them to the controller to complete the optimization and adjustment.

9. The transformer drying time sequence prediction method according to claim 8, characterized in that: The fitness function is obtained as follows: The particle position is used as the input feature vector for changing the drying process parameters. Each particle position represents a set of candidate solutions for the drying process parameters. The changed input feature vector is input into the drying prediction model to obtain the predicted drying time series vector and predicted energy consumption data. Based on the Euclidean distance between the predicted drying time series vector and the expected drying time series vector, as well as the deviation between the predicted energy consumption data and the expected energy consumption data, the fitness function is constructed through fusion calculation.

10. A transformer drying time sequence prediction system, the system being used to implement the prediction method according to any one of claims 1 to 9, characterized in that: include: Drying Dataset Subsystem: This includes a data acquisition module, a data cleaning module, a data normalization module, and a data set construction module. This module acquires information about the entire transformer drying process from multiple sources to form a raw data set. It then classifies and selects high-quality data sets based on the corresponding classification information to obtain high-quality data sets. This high-quality data set is then processed and integrated to obtain a learning data set containing an input feature set, a drying time series vector set, and an energy consumption data set. Drying prediction model subsystem: This includes a model construction module, a model training module, a prediction evaluation module, and a parameter optimization module. It builds a drying prediction model based on a hybrid neural network, uses a learning data set to train the drying prediction model, and determines whether the prediction accuracy of the drying prediction model meets the standard. If so, the drying prediction model is used to perform inference prediction before the current drying process starts, and the predicted drying time series vector and predicted energy consumption data are obtained as prediction results. Based on the prediction results, it is determined whether the drying process parameters need to be optimized. If necessary, the drying process parameters are collaboratively optimized and adjusted based on the prediction results through the adaptive control algorithm combined with the drying prediction model. Drying model interface subsystem: includes the interface between the drying prediction model and the equipment, the SCADA interaction interface, and the MES interaction interface to achieve precise interaction.

Images