Thermal power generating unit main steam temperature prediction method based on adaptive wavelet neural network and related device
Through the time series decomposition and fusion prediction method of adaptive wavelet neural network, the accuracy problem of main steam temperature prediction of thermal power units is solved, and high-precision prediction under multiple operating conditions is achieved.
Patent Information
- Application Number
- CN202510739019.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-04
- Publication Date
- 2025-09-16
AI Technical Summary
The existing technology is not very accurate in predicting the main steam temperature of thermal power units, especially in load fluctuation scenarios where the prediction error is large, making it difficult to effectively capture multi-band characteristics and adapt to dynamic operating conditions.
Adaptive wavelet neural network method is used to divide the main steam temperature data into trend data and periodic data through time series decomposition. Different prediction methods are used for each type of data, and prediction is performed through channel attention mechanism and adaptive wavelet decomposition. Finally, the two results are fused.
The accuracy and precision of the main steam temperature prediction of thermal power units are improved, which can adapt to various operating conditions, reduce prediction errors, and provide more reliable temperature prediction results.
Smart Images
Figure CN120653958A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of main steam temperature prediction of thermal power units, and in particular relates to a main steam temperature prediction method of a thermal power unit based on an adaptive wavelet neural network and a related device. Background Art
[0002] Currently, the methods mostly used in the existing technology for predicting the main steam temperature of thermal power units are: 1) statistics-based methods such as ARIMA (AutoRegressive Integrated Moving Average) and SVR (Support Vector Regression); 2) deep learning models represented by LSTM (Long Short-Term Memory) and Transformer; 3) frequency domain enhancement methods represented by FEDformer (Frequency Enhanced Decomposed Transformer).
[0003] Although statistical methods such as ARIMA and SVR perform well in stationary series, they have a significant lag in predicting sudden changes in the start-up and shutdown processes. The prediction error of the main steam temperature can reach ±8°C under load fluctuation scenarios.
[0004] Although deep learning models represented by LSTM and Transformer can capture long-range dependencies, they do not show the ability to separate multi-band features. Experimental data show that the standard LSTM's ability to capture high-frequency disturbances is 37% lower than that of the wavelet decomposition method.
[0005] Although frequency domain enhancement methods represented by FEDformer introduce wavelet transform, their fixed basis functions are difficult to adapt to dynamic working conditions. Zhou et al. proposed an innovative long-term time series prediction model FEDformer. In variable load tests, they found that the frequency band mismatch rate of the artificial wavelet basis reached 41%. Summary of the Invention
[0006] The purpose of the present invention is to provide a method and related device for predicting the main steam temperature of a thermal power unit based on an adaptive wavelet neural network, so as to solve the problem of low accuracy in predicting the main steam temperature of a thermal power unit in the prior art.
[0007] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for predicting the main steam temperature of a thermal power unit based on an adaptive wavelet neural network, comprising the following steps: Obtain main steam temperature data of thermal power units; Preprocessing the main steam temperature data of the thermal power unit obtained, removing noise and filling missing values, and obtaining the preprocessed main steam temperature data of the thermal power unit; Perform time series decomposition on the pre-processed main steam temperature data of thermal power units to obtain trend temperature data and periodic temperature data; Cluster the trend temperature data to obtain several clusters, and perform linear transformation alignment on the clusters to obtain the trend temperature data prediction results; Performing adaptive wavelet decomposition on the periodic temperature data to obtain several original scale features, wherein the scale features include detail coefficients and approximation coefficients; The channel attention mechanism is used to capture the attention relationship between several scale features in different channels, and several scale features are processed; Reconstructing the processed scale features into periodic temperature data to obtain the periodic temperature data prediction results; The trend temperature data prediction results and the periodic temperature data prediction results are integrated to obtain the final prediction results of the main steam temperature of the thermal power unit; A further improvement of the present invention is that, in the step of performing time series decomposition on the preprocessed main steam temperature data of the thermal power unit to obtain trend temperature data and periodic temperature data, the STL decomposition method is specifically used to perform time series decomposition on the preprocessed main steam temperature data of the thermal power unit to obtain trend temperature data and periodic temperature data.
[0008] A further improvement of the present invention is that the periodic temperature data is decomposed using the following formula:
[0009] in, is the even phase component, is the odd phase component, The number of points representing periodic temperature data, represents the number of decomposition levels, for Layer periodic temperature data, For any one, is a natural number.
[0010] A further improvement of the present invention is that the calculation formula of the STL decomposition method is:
[0011] in, is the main steam temperature data of the thermal power unit after preprocessing, is the periodic temperature data, This is trend temperature data.
[0012] A further improvement of the present invention is that the calculation formula of the detail coefficient is:
[0013] in, is the detail coefficient, is the odd phase component, is a convolution-based prediction operator, is the approximate coefficient; The calculation formula of the approximate coefficient is:
[0014] in, is the approximate coefficient, is the even phase component, is the convolution-based update operator, is the detail coefficient.
[0015] A further improvement of the present invention is that the calculation formula of the channel attention mechanism is:
[0016] in, is processed by the channel attention mechanism Periodic temperature data of the layer, is the channel attention mechanism, for Periodic temperature data of a layer.
[0017] A further improvement of the present invention is that the reconstructed calculation formula is:
[0018] in, Represents the result of reconstructing several scale features after processing, In order to reconstruct several scale features after processing, is the even phase component, is the odd phase component.
[0019] In a second aspect, the present invention provides a thermal power unit main steam temperature prediction system based on an adaptive wavelet neural network, comprising a data acquisition module, a data preprocessing module, a time series decomposition module, a trend temperature data prediction module, an adaptive wavelet decomposition module, a channel attention mechanism capture module, a periodic temperature data prediction module, and a thermal power unit main steam temperature prediction module; The data acquisition module is used to obtain the main steam temperature data of the thermal power unit; The data preprocessing module is used to preprocess the obtained main steam temperature data of the thermal power unit, remove noise and fill in missing values, and obtain preprocessed main steam temperature data of the thermal power unit; The time series decomposition module is used to perform time series decomposition on the pre-processed main steam temperature data of the thermal power unit to obtain trend temperature data and periodic temperature data; The trend temperature data prediction module is used to cluster the trend temperature data to obtain a plurality of clusters, and perform linear transformation alignment on the plurality of clusters to obtain a trend temperature data prediction result; The adaptive wavelet decomposition module is used to perform adaptive wavelet decomposition on the periodic temperature data to obtain a plurality of original scale features, wherein the scale features include detail coefficients and approximation coefficients; The channel attention mechanism capture module is used to capture the attention relationship between multiple scale features in different channels using the channel attention mechanism to obtain multiple scale features after processing; The periodic temperature data prediction module is used to reconstruct the processed multiple scale features into periodic temperature data to obtain periodic temperature data prediction results; The thermal power unit main steam temperature prediction module is used to fuse the trend temperature data prediction result and the periodic temperature data prediction result to obtain the final thermal power unit main steam temperature prediction result.
[0020] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned method for predicting the main steam temperature of a thermal power unit based on an adaptive wavelet neural network are implemented.
[0021] In a fourth aspect, the present invention provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned method for predicting the main steam temperature of a thermal power unit based on an adaptive wavelet neural network.
[0022] Compared with the prior art, the present invention has the following beneficial effects: The present invention is an improved invention. Compared with existing methods for predicting the main steam temperature of thermal power plants, the present invention simultaneously uses adaptive wavelet decomposition and a channel attention mechanism to predict the main steam temperature of thermal power plants. Specifically, the main steam temperature data of the thermal power plants is decomposed into trend temperature data and periodic temperature data. Different prediction methods are used for the trend temperature data and the periodic temperature data, respectively. Linear transformation alignment is used for prediction of the trend temperature data, while adaptive wavelet decomposition and a channel attention mechanism are simultaneously used for prediction of the periodic temperature data. It can be seen that the present invention adopts a hybrid prediction strategy, which can adaptively segment data according to data characteristics and use different prediction methods for different data, thereby improving the accuracy of the final thermal power plant main steam temperature prediction. Furthermore, the present invention fuses the prediction results of the trend temperature data and the periodic temperature data to obtain the final thermal power plant main steam temperature prediction result. The prediction results of various data may have certain errors and limitations. By fusing the two prediction results, the errors can be reduced, thereby improving the accuracy of the final thermal power plant main steam temperature prediction, thereby solving the problem of low accuracy of thermal power plant main steam temperature prediction in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 This is a flow chart of a method for predicting the main steam temperature of a thermal power unit based on an adaptive wavelet neural network according to the present invention; Figure 2 Schematic diagram of a main steam temperature prediction system for a thermal power plant based on an adaptive wavelet neural network according to the present invention; Figure 3 This is a flow chart of a method for predicting the main steam temperature of a thermal power unit based on an adaptive wavelet neural network in Example 3 of the present invention; Figure 4 Schematic diagram of the structure of the electronic device of the present invention. DETAILED DESCRIPTION
[0024] In order to further understand the content of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the embodiments are only for explaining the present invention and are not intended to limit it.
[0025] The present invention proposes a method for predicting the main steam temperature of a thermal power plant based on an adaptive wavelet neural network. This method performs adaptive wavelet decomposition on periodic temperature data to obtain several original scale features. A channel attention mechanism is then used to capture the attention relationships between these scale features across different channels, resulting in several processed scale features. The processed scale features are then reconstructed into periodic temperature data to obtain a periodic temperature data prediction result. The trend temperature data prediction result and the periodic temperature data prediction result are then fused to obtain the final thermal power plant main steam temperature prediction result. Compared to existing technologies, this method effectively addresses the low accuracy of thermal power plant main steam temperature prediction in existing technologies.
[0026] Example 1: The flow chart of the method for predicting the main steam temperature of a thermal power plant based on an adaptive wavelet neural network is shown in FIG. Figure 1 As shown, the method for predicting the main steam temperature of a thermal power unit based on an adaptive wavelet neural network of the present invention comprises the following steps: S1. Obtain the main steam temperature data of the thermal power unit.
[0027] S2. Preprocess the main steam temperature data of the thermal power unit obtained, remove noise and fill in missing values, and obtain preprocessed main steam temperature data of the thermal power unit.
[0028] S3. Perform time series decomposition on the preprocessed main steam temperature data of the thermal power unit to obtain trend temperature data and periodic temperature data.
[0029] S4. Cluster the trend temperature data to obtain several clusters, and perform linear transformation and alignment on the clusters to obtain trend temperature data prediction results.
[0030] S5. Perform adaptive wavelet decomposition on the periodic temperature data to obtain several original scale features, wherein the scale features include detail coefficients and approximation coefficients.
[0031] S6. Use the channel attention mechanism to capture the attention relationship between several scale features in different channels and obtain several processed scale features.
[0032] S7. Reconstruct the processed scale features into periodic temperature data to obtain a periodic temperature data prediction result.
[0033] S8. The trend temperature data prediction results and the periodic temperature data prediction results are integrated to obtain the final prediction results of the main steam temperature of the thermal power unit.
[0034] Example 2: The schematic diagram of the main steam temperature prediction system of thermal power unit based on adaptive wavelet neural network of the present invention is as follows: Figure 2 As shown, the main steam temperature prediction system of a thermal power unit based on an adaptive wavelet neural network of the present invention includes a data acquisition module, a data preprocessing module, a time series decomposition module, a trend temperature data prediction module, an adaptive wavelet decomposition module, a channel attention mechanism capture module, a periodic temperature data prediction module and a thermal power unit main steam temperature prediction module.
[0035] The data acquisition module is used to obtain the main steam temperature data of the thermal power unit.
[0036] The data preprocessing module is used to preprocess the main steam temperature data of the thermal power unit, remove noise and fill missing values, and obtain the preprocessed main steam temperature data of the thermal power unit.
[0037] The time series decomposition module is used to perform time series decomposition on the preprocessed main steam temperature data of the thermal power unit to obtain trend temperature data and periodic temperature data.
[0038] The trend temperature data prediction module is used to cluster the trend temperature data to obtain several clusters, and perform linear transformation alignment on the several clusters to obtain the trend temperature data prediction results.
[0039] The adaptive wavelet decomposition module is used to perform adaptive wavelet decomposition on the periodic temperature data to obtain a plurality of original scale features, wherein the scale features include detail coefficients and approximation coefficients.
[0040] The channel attention mechanism capture module is used to capture the attention relationship between several scale features in different channels using the channel attention mechanism to obtain several processed scale features; The periodic temperature data prediction module is used to reconstruct the processed scale features into periodic temperature data to obtain the periodic temperature data prediction results; The main steam temperature prediction module of the thermal power unit is used to fuse the trend temperature data prediction results and the periodic temperature data prediction results to obtain the final main steam temperature prediction results of the thermal power unit.
[0041] Example 3: The flow chart of the method for predicting the main steam temperature of a thermal power plant based on an adaptive wavelet neural network is shown in FIG. Figure 3 As shown, the method for predicting the main steam temperature of a thermal power unit based on an adaptive wavelet neural network of the present invention comprises the following steps: S1. Obtain the main steam temperature data of the thermal power unit.
[0042] In this step, the main steam temperature data of the thermal power unit is obtained from the DCS (Distributed Control System) of a supercritical coal-fired power unit.
[0043] The DCS system acquires data with a 10-second cycle, covering a variety of typical operating conditions, including startup and shutdown, sudden load changes, and steady-state operation. This data exhibits strong non-stationarity, multi-scale superposition, and significant high-frequency disturbances, making it a core data source for subsequent validation of the performance of industrial-grade prediction models.
[0044] S2. Preprocess the main steam temperature data of the thermal power unit obtained, remove noise and fill in missing values, and obtain preprocessed main steam temperature data of the thermal power unit.
[0045] In this step, the wavelet coefficient regularization method is used to remove noise, and the boundary filling method based on moving average is used to fill in missing values.
[0046] S3. Perform time series decomposition on the preprocessed main steam temperature data of the thermal power unit to obtain trend temperature data and periodic temperature data.
[0047] In this step, the STL (Seasonal-Trend decomposition) method is used to decompose the preprocessed main steam temperature data of the thermal power unit into time series to obtain trend temperature data and periodic temperature data (also called seasonal temperature data).
[0048] The calculation formula of the STL decomposition method is:
[0049] in, is the main steam temperature data of the thermal power unit after preprocessing, ∈RC×T, C is the number of channels, T is the time length, is the periodic temperature data, This is trend temperature data.
[0050] In this step, the following formula is also used to decompose the periodic temperature data:
[0051] in, is the even phase component, is the odd phase component, The number of points representing periodic temperature data, represents the number of decomposition levels, for Layer periodic temperature data, For any one, is a natural number.
[0052] S4. Cluster the trend temperature data to obtain several clusters, and perform linear transformation and alignment on the clusters to obtain trend temperature data prediction results.
[0053] In this step, the K-means algorithm (also called K-means algorithm) is used to cluster the trend temperature data ( Figure 3 The clustering algorithm and K-means algorithm are represented separately, indicating that the clustering algorithm used is the K-means algorithm), and several clusters are obtained.
[0054] A linear mapping function is used to perform linear transformation alignment on several clusters (also called grouped linear alignment) to obtain trend temperature data prediction results.
[0055] The linear mapping function expression is:
[0056] in, X represents the original signal (several clusters), Y Represents the mapped signal (trend temperature data prediction result), A Represents the mapping matrix.
[0057] S5. Perform adaptive wavelet decomposition on the periodic temperature data to obtain several original scale features.
[0058] This step uses a lifting scheme to perform adaptive wavelet decomposition (also called adaptive wavelet processing) on the periodic temperature data, obtaining several original scale features. The main steps of adaptive wavelet decomposition include splitting, predicting, and updating. Scale features include detail coefficients and approximation coefficients.
[0059] The calculation formula of detail coefficient is:
[0060] in, is the detail coefficient, is the odd phase component, is a convolution-based prediction operator, is the approximate coefficient.
[0061] Approximation coefficient The calculation formula is:
[0062] in, is the approximate coefficient, is the even phase component, is the convolution-based update operator, is the detail coefficient.
[0063] S6. Use the channel attention mechanism to capture the attention relationship between several scale features in different channels and obtain several processed scale features.
[0064] The calculation formula of the channel attention mechanism is:
[0065] in, is processed by the channel attention mechanism l Periodic temperature data of the layer, is the channel attention mechanism, for Periodic temperature data of a layer.
[0066] S7. Reconstruct the processed scale features into periodic temperature data to obtain a periodic temperature data prediction result.
[0067] In this step, an inverse adaptive wavelet transform method is specifically used to reconstruct the processed scale features into periodic temperature data to obtain the periodic temperature data prediction results.
[0068] The calculation formula reconstructed in this step is:
[0069] in, Represents the result of reconstructing several scale features after processing, In order to reconstruct several scale features after processing, is the even phase component, is the odd phase component.
[0070] The following is a detailed description of the process of reconstructing the processed scale features into periodic temperature data and obtaining the periodic temperature data prediction results: Feature upsampling is achieved through transposed convolution:
[0071] in, is the inverse reconstruction operator, is processed by the channel attention mechanism Periodic temperature data of a layer.
[0072] Approximate component decomposition:
[0073] in, is the approximate coefficient, is the detail coefficient, is processed by the channel attention mechanism Periodic temperature data of the layer, is a symmetric difference set.
[0074] Reverse update process:
[0075] in, for Layer trend temperature data is updated based on convolution operators, is the even phase component, is the detail coefficient, is the approximate coefficient.
[0076] Inverse prediction process:
[0077] in, for Layer trend temperature data is based on the convolution prediction operator, is the odd phase component, is the even phase component, is the detail coefficient.
[0078] Reconstruct several scale features after processing:
[0079] in, Represents the result of reconstructing several scale features after processing, In order to reconstruct several scale features after processing, is the even phase component, is the odd phase component.
[0080] S8. The trend temperature data prediction results and the periodic temperature data prediction results are integrated to obtain the final prediction results of the main steam temperature of the thermal power unit.
[0081] Among them, the adaptive wavelet neural network consists of adaptive wavelet decomposition and channel attention mechanism.
[0082] In order to verify the effectiveness of the method of the present invention (adaptive wavelet neural network), this embodiment uses current mainstream prediction models to conduct comparative experiments, including Transformer, Autoformer, Informer, FEDformer and iTransformer.
[0083] All models were trained for a uniform number of 10 rounds, using the Adam optimizer with an initial learning rate of 0.0005. Each model was trained on the same training and validation sets to ensure fair comparison.
[0084] This example uses real-world power plant operating data, BMST (Boiler Main Steam Temperature), to validate the performance of the proposed method in predicting the main steam temperature of thermal power units. This example also uses five standard public time series datasets to verify the performance of the proposed method in various time series prediction tasks. These datasets include electricity consumption, power transformer temperature, exchange rate, traffic flow, and meteorological data. These datasets cover multiple application scenarios in industry, energy, finance, transportation, and meteorology.
[0085] The following is a detailed description of electricity consumption data, power transformer temperature data, exchange rate data, traffic flow data, and meteorological data: The Electricity Consumption Dataset (electricity) contains electricity consumption records for 370 users in the United States, covering a period starting in 2011 with a 15-minute sampling period. This dataset exhibits significant daily periodicity and nonlinear fluctuations driven by user behavior and is widely used in evaluating power load forecasting models.
[0086] b. ETT (Electricity Transformer Temperature) data is sourced from the State Grid monitoring system and records parameters such as temperature, voltage, and current during transformer operation. It includes subsets such as ETTh1 and ETTh2 (hourly) and ETTm1 and ETTm2 (minutely). This data set is a typical test model's ability to analyze load response and cyclic patterns.
[0087] c. The Exchange Rate Dataset collects daily exchange rate data from eight countries. It has nonlinear fluctuation characteristics dominated by low-frequency trends and driven by policy events, making it suitable for evaluating the model's ability to model long-term dependencies and structural mutations.
[0088] The d. Traffic Dataset (traffic) collects traffic flow information from 48 sensors on the California freeway system, sampling every hour over a one-year period. This dataset incorporates periodic weekday / weekend variations and peak / valley traffic fluctuations, making it suitable for validating models' capabilities in scenarios with periodic disturbances and multi-peak distributions.
[0089] The e.Weather Dataset (weather) includes meteorological parameters such as temperature, humidity, wind speed, and precipitation for 26 cities, with a one-hour sampling period. This dataset is widely used to evaluate the effectiveness of time series modeling and forecasting models under multivariate coupling and complex natural cycle characteristics.
[0090] All datasets are divided into 70% training, 10% validation and 20% test ratios, with an input length of 48 and a prediction length of 24.
[0091] The experimental environment of this embodiment is described as follows: The operating system is Ubuntu 20.04, the programming language is Python 3.8, the deep learning framework is PyTorch2.4, the hardware platform is CPU: AMD 7500F, GPU: NVIDIA RTX 4070 SUPER, RAM: 32GB.
[0092] In order to verify the performance of the method of the present invention (adaptive wavelet neural network), the mean absolute error (Mean Absolute Error) and mean squared error (Mean Squared Error) are used to evaluate the performance of the method of the present invention. The mean absolute error measures the average absolute error between the predicted value and the true value, reflecting the overall degree of deviation, while the mean squared error can better highlight the impact of large error points on model performance.
[0093] The formula for calculating the mean absolute error is:
[0094] in, is the mean absolute error, is the number of sample points, For the The true value of the sample, For the The predicted value of the sample.
[0095] The formula for calculating the mean square error is:
[0096] in, is the mean square error, is the number of sample points, For the The true value of the sample, For the The predicted value of the sample.
[0097] Table 1 shows the evaluation results of each model on different datasets. The lowest MAE and MSE values are highlighted in red, and the second lowest values are highlighted in blue.
[0098] Table 1 shows the evaluation results of each model on different datasets
[0099] Table 2 shows the training efficiency on different datasets. The lowest training time value is highlighted in red, and the second lowest value is highlighted in blue.
[0100] Table 2 Training efficiency of each model on different datasets
[0101] The evaluation results of various models on various datasets in Table 1 show that the proposed method (also known as the AWNN model, which stands for Adaptive Wavelet Neural Network) demonstrates excellent predictive capabilities across multiple datasets. While the MAE and MSE of the AWNN model on the BMST dataset are not the lowest, the difference is not significant. This demonstrates that even when processing boiler main steam temperature data characterized by strong nonstationarity, multi-scale superposition, and significant high-frequency disturbances, the AWNN model is able to effectively capture the complex patterns and trends in the data, providing a relatively accurate solution for main steam temperature prediction of thermal power units. Furthermore, the AWNN model performs even better on other datasets.
[0102] On the electricity dataset, the AWNN model achieved a MAE of 0.225 and an MSE of 0.136, both the lowest values among all models. This demonstrates that the AWNN model demonstrates excellent predictive capabilities when processing power load data with significant daily periodicity and nonlinear fluctuations driven by user behavior, accurately capturing the periodic and nonlinear characteristics in the data. On the ETTh1 dataset, the AWNN model achieved a MAE of 0.382 and an MSE of 0.355, both the lowest values. This demonstrates that the AWNN model can effectively analyze load responses and periodic patterns when processing data containing parameters such as temperature, voltage, and current during transformer operation. On the ETTm1 dataset, the AWNN model achieved a MAE of 0.362, the lowest value, and an MSE of 0.363, with a small difference from the lowest value. This demonstrates that the AWNN model can also achieve good prediction results when processing minute-level transformer operation data, demonstrating its good adaptability to data of different time scales. On the traffic dataset, the AWNN model achieved a MAE of 0.282 and an MSE of 0.435, both of which were among the lowest. This demonstrates that when processing traffic flow data with periodic weekday / weekend variations and peak / valley flow fluctuations, the AWNN model is able to effectively capture the periodic disturbances and multi-peak distribution characteristics in the data. On the weather dataset, the AWNN model achieved a MAE of 0.120 and an MSE of 0.099, both of which were the lowest among all models. This fully demonstrates its advantages in processing meteorological data with multivariate coupling and complex natural periodic characteristics, and its ability to accurately model and predict changes in meteorological parameters.
[0103] Table 2 shows the training efficiency of various models on different datasets. The AWNN model has relatively short training times on most datasets. On the BMST dataset, the AWNN model's training time was 350 seconds, second only to the iTransformer model's 323 seconds. On the Electricity dataset, the AWNN model's training time was 896 seconds. While not the lowest, it was still within a reasonable range compared to other models. On the Exchange dataset (a public time series dataset), the AWNN model's training time was 269 seconds, comparable to the iTransformer model's 265 seconds and significantly lower than other models. On the Traffic dataset, the AWNN model's training time was 1365 seconds. While slightly higher than some models, it remains within acceptable limits for practical applications. On the Weather dataset, the AWNN model's training time was 347 seconds, second only to the iTransformer model's 307 seconds.
[0104] Taking into account both prediction performance (evaluation metrics) and training efficiency, the AWNN model demonstrates significant advantages in thermal power plant main steam temperature prediction. Although the AWNN model's prediction accuracy is not the highest on the BMST dataset, its excellent performance on other datasets demonstrates its broad applicability and strong generalization capabilities. The AWNN model achieves excellent results in a variety of complex time series prediction scenarios, providing reliable and accurate main steam temperature predictions for thermal power plant operations, helping to optimize combustion processes, improve power generation efficiency, and reduce pollutant emissions. Therefore, the AWNN-based method for predicting the main steam temperature of thermal power units has broad application prospects and significant practical value, providing strong technical support for the safe, stable, and economical operation of thermal power plants.
[0105] Example 4: See also Figure 4 As shown, the present invention also provides an electronic device 100 for a main steam temperature prediction method for a thermal power unit based on AWNN; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and capable of running on the at least one processor 102, and at least one communication bus 104.
[0106] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the AWNN-based thermal power unit main steam temperature prediction method described in Example 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 can mainly include a program storage area and a data storage area. The program storage area can store an operating system, at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area can store data (such as audio data) created based on the use of the electronic device 100. In addition, the memory 101 can include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device.
[0107] The at least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100 and connects various parts of the entire electronic device 100 using various interfaces and lines.
[0108] The memory 101 in the electronic device 100 stores a plurality of instructions to implement a method for predicting the main steam temperature of a thermal power unit based on an AWNN. The processor 102 can execute the plurality of instructions to implement: Obtain main steam temperature data of thermal power units; Preprocessing the main steam temperature data of the thermal power unit obtained, removing noise and filling missing values, and obtaining the preprocessed main steam temperature data of the thermal power unit; Perform time series decomposition on the pre-processed main steam temperature data of thermal power units to obtain trend temperature data and periodic temperature data; Cluster the trend temperature data to obtain several clusters, and perform linear transformation alignment on the clusters to obtain the trend temperature data prediction results; Performing adaptive wavelet decomposition on the periodic temperature data to obtain several original scale features, wherein the scale features include detail coefficients and approximation coefficients; The channel attention mechanism is used to capture the attention relationship between several scale features in different channels, and several scale features are processed; Reconstructing the processed scale features into periodic temperature data to obtain the periodic temperature data prediction results; The trend temperature data prediction results and the periodic temperature data prediction results are integrated to obtain the final main steam temperature prediction results of the thermal power unit.
[0109] Example 5: If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory and read-only memory (ROM, Read-Only Memory).
[0110] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0111] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, 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 produce 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 flowcharts and / or block diagrams. 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.
[0112] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 The function specified in one or more boxes.
[0113] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device 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.
[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.
Claims
1. A method for predicting the main steam temperature of a thermal power unit based on an adaptive wavelet neural network, characterized in that: The following steps are involved: Obtain main steam temperature data of thermal power units; Preprocessing the main steam temperature data of the thermal power unit obtained, removing noise and filling missing values, and obtaining the preprocessed main steam temperature data of the thermal power unit; Perform time series decomposition on the pre-processed main steam temperature data of thermal power units to obtain trend temperature data and periodic temperature data; Cluster the trend temperature data to obtain several clusters, and perform linear transformation alignment on the clusters to obtain the trend temperature data prediction results; Performing adaptive wavelet decomposition on the periodic temperature data to obtain several original scale features, wherein the scale features include detail coefficients and approximation coefficients; The channel attention mechanism is used to capture the attention relationship between several scale features in different channels, and several scale features are processed; Reconstructing the processed scale features into periodic temperature data to obtain the periodic temperature data prediction results; The trend temperature data prediction results and the periodic temperature data prediction results are integrated to obtain the final main steam temperature prediction results of the thermal power unit.
2. The method for predicting main steam temperature of a thermal power plant based on an adaptive wavelet neural network according to claim 1, characterized in that: In the step of performing time series decomposition on the preprocessed main steam temperature data of the thermal power unit to obtain trend temperature data and periodic temperature data, the STL decomposition method is specifically used to perform time series decomposition on the preprocessed main steam temperature data of the thermal power unit to obtain trend temperature data and periodic temperature data.
3. The method for predicting main steam temperature of a thermal power plant based on an adaptive wavelet neural network according to claim 2, characterized in that: The following formula is also used to decompose the periodic temperature data: in, is the even phase component, is the odd phase component, The number of points representing periodic temperature data, represents the number of decomposition levels, for Layer periodic temperature data, For any one, is a natural number.
4. The method for predicting main steam temperature of a thermal power plant based on an adaptive wavelet neural network according to claim 2, characterized in that: The calculation formula of the STL decomposition method is: in, is the main steam temperature data of the thermal power unit after preprocessing, is the periodic temperature data, This is trend temperature data.
5. The method for predicting main steam temperature of a thermal power plant based on an adaptive wavelet neural network according to claim 1, characterized in that: The calculation formula of detail coefficient is: in, is the detail coefficient, is the odd phase component, is a convolution-based prediction operator, is the approximate coefficient; The calculation formula of the approximate coefficient is: in, is the approximate coefficient, is the even phase component, is the convolution-based update operator, is the detail coefficient.
6. The method for predicting main steam temperature of a thermal power plant based on an adaptive wavelet neural network according to claim 1, characterized in that: The calculation formula of the channel attention mechanism is: in, is processed by the channel attention mechanism Periodic temperature data of the layer, is the channel attention mechanism, for Periodic temperature data of a layer.
7. The method for predicting main steam temperature of a thermal power plant based on an adaptive wavelet neural network according to claim 1, characterized in that: The calculation formula for reconstruction is: in, Represents the result of reconstructing several scale features after processing, In order to reconstruct several scale features after processing, is the even phase component, is the odd phase component.
8. A main steam temperature prediction system for thermal power units based on adaptive wavelet neural network, characterized in that: It includes data acquisition module, data preprocessing module, time series decomposition module, trend temperature data prediction module, adaptive wavelet decomposition module, channel attention mechanism capture module, periodic temperature data prediction module and thermal power unit main steam temperature prediction module; The data acquisition module is used to obtain the main steam temperature data of the thermal power unit; The data preprocessing module is used to preprocess the obtained main steam temperature data of the thermal power unit, remove noise and fill in missing values, and obtain preprocessed main steam temperature data of the thermal power unit; The time series decomposition module is used to perform time series decomposition on the pre-processed main steam temperature data of the thermal power unit to obtain trend temperature data and periodic temperature data; The trend temperature data prediction module is used to cluster the trend temperature data to obtain a plurality of clusters, and perform linear transformation alignment on the plurality of clusters to obtain a trend temperature data prediction result; The adaptive wavelet decomposition module is used to perform adaptive wavelet decomposition on the periodic temperature data to obtain a plurality of original scale features, wherein the scale features include detail coefficients and approximation coefficients; The channel attention mechanism capture module is used to capture the attention relationship between multiple scale features in different channels using the channel attention mechanism to obtain multiple scale features after processing; The periodic temperature data prediction module is used to reconstruct the processed multiple scale features into periodic temperature data to obtain periodic temperature data prediction results; The thermal power unit main steam temperature prediction module is used to fuse the trend temperature data prediction result and the periodic temperature data prediction result to obtain the final thermal power unit main steam temperature prediction result.
9. An electronic 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 for predicting the main steam temperature of a thermal power unit based on an adaptive wavelet neural network according to any one of claims 1 to 7 are implemented.
10. A 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 for predicting the main steam temperature of a thermal power unit based on an adaptive wavelet neural network according to any one of claims 1 to 7 are implemented.