Main steam flow measurement method, system and medium based on FWSN and RF

Abstract

The present invention discloses a main steam flow measurement method, system, and medium based on FWSN and RF, belonging to the technical field of soft measurement of thermal parameters of energy power systems. The method comprises the following steps: 1, collecting raw data of operating parameters related to main steam flow from the unit's distributed control system; 2, filling missing values ​​and outliers in the raw data, and then normalizing the data; 3, performing dimensionality reduction on the data using a kernel principal component analysis algorithm; 4, constructing a main steam flow prediction model based on FWSN and RF; and 5, inputting the measured data into the main steam flow prediction model based on FWSN and RF, and outputting a predicted value for the main steam flow of the thermal power unit. The present invention can achieve high-precision prediction of main steam flow.

Description

Technical Field

[0001] The present invention belongs to the technical field of soft measurement of thermal parameters of energy power systems, and in particular relates to a main steam flow measurement method, system and medium based on FWSN (fractional wavelet scattering network) and RF (random forest), which is suitable for steam flow prediction and monitoring of large-capacity ultra-supercritical power generation units. Background Art

[0002] Main steam flow, as one of the core parameters of a thermal power unit's thermal system, has a significant impact on the unit's operation and performance. Accurately predicting main steam flow can achieve objectives such as unit operation monitoring, energy efficiency assessment, load control, and operational optimization, thereby improving the unit's operating efficiency and reliability. Currently, main steam flow measurement technologies for generator sets can be categorized into two main methods, direct and indirect, based on the measurement principle. Direct measurement typically uses the differential pressure method, achieved by installing throttling devices such as flow orifice plates or nozzle flowmeters. While this method provides intuitive and concise measurement results and is easy to operate, during actual generator operation, steam passing through the throttling devices inevitably causes a certain amount of pressure loss, which in turn increases system energy losses. Furthermore, with the recent commissioning of a large number of ultra-supercritical units, existing standard throttling devices have become less accurate and even unable to meet the high flow measurement requirements of large-capacity units, while also resulting in significant throttling losses. Therefore, modern large-scale units often use indirect measurement methods instead of throttling devices for main steam flow measurement.

[0003] Indirect measurements based on the Flugel formula usually use thermal parameters such as pressure and temperature correction before and after the turbine stage group to convert and calculate the value of the main steam flow. However, when the passage of the turbine changes due to scaling or corrosion, it will cause a large deviation in the measurement results of the main steam flow. Especially in the context of deep peak regulation, thermal power units need to be in variable load and low load operation conditions for a long time, the main steam pressure and main steam temperature fluctuate more frequently, and the pressure after the regulation stage and the main steam flow are not an ideal linear relationship. Although researchers have made multiple revisions to the Flugel formula, the model calculation results still have large deviations and their scope of use is limited. Summary of the Invention

[0004] In order to solve the problem that it is difficult to accurately monitor the main steam flow of thermal power units, the present invention provides a main steam flow measurement method, system and medium based on FWSN and RF.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A main steam flow measurement method based on FWSN and RF includes the following steps: step 1, collecting original data of operating parameters related to the main steam flow from the unit distributed control system; step 2, filling missing values ​​and outliers in the original data obtained in step 1, and then normalizing the data; step 3, performing dimensionality reduction processing on the normalized data in step 2 by using a kernel principal component analysis algorithm; step 4, constructing a main steam flow prediction model based on FWSN and RF, FWSN extracts the feature vector of the main steam flow, and then inputs the extracted feature vector into the RF model, and the RF model predicts based on a trained decision tree and outputs the predicted value of the main steam flow; FWSN represents fractional wavelet scattering network, and RF represents random forest;

[0007] Based on the data after dimensionality reduction processing in step 3, the main steam flow prediction model based on FWSN and RF is trained; in step 5, the data to be measured is input into the main steam flow prediction model based on FWSN and RF trained in step 4, and the main steam flow measurement value of the thermal power unit is output.

[0008] The present invention also provides a main steam flow measurement system for implementing the above-mentioned main steam flow measurement method based on FWSN and RF, including: a data acquisition module, which collects original data of operating parameters related to the main steam flow from the unit distributed control system; a preprocessing module, which fills missing values ​​and abnormal values ​​in the original data, and then performs normalization and dimensionality reduction processing; a model operation module, which deploys a main steam flow prediction model based on FWSN and RF and performs prediction calculations; a visualization terminal, which displays real-time main steam flow measurement values ​​and historical trend curves.

[0009] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the above-mentioned main steam flow measurement method based on FWSN and RF are implemented.

[0010] Beneficial effects:

[0011] This paper combines the powerful feature extraction capabilities of fractional wavelet scattering networks with the excellent generalization and anti-interference properties of the random forest algorithm to construct a fractional wavelet scattering network and random forest model for main steam flow prediction. Compared with existing prediction models, this paper avoids the gradient vanishing and gradient exploding problems that occur in time series prediction. Furthermore, the introduced random forest algorithm optimizes and integrates the features extracted by the fractional wavelet scattering network, further improving the accuracy of prediction results. Furthermore, the fractional wavelet scattering network effectively reduces computational complexity, improves model performance and generalization, and facilitates model training. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 is a flow chart of the main steam flow measurement method based on FWSN and RF of the present invention;

[0013] Figure 2 1 is a load curve diagram of a unit in an embodiment of the present invention;

[0014] Figure 3 This is a diagram showing the kernel principal component analysis results in an embodiment of the present invention;

[0015] Figure 4 Schematic diagram of a three-layer fractional-order wavelet scattering network in an embodiment of the present invention;

[0016] Figure 5 Schematic diagram of the random forest algorithm in an embodiment of the present invention;

[0017] Figure 6 4 is a structural diagram of a main steam flow prediction model based on FWSN and RF in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0019] like Figure 1 As shown, the present invention provides a main steam flow measurement method based on FWSN and RF, comprising the following steps: step 1, collecting original data of operating parameters related to the main steam flow from the unit distributed control system (DCS); step 2, filling missing values ​​and abnormal values ​​in the original data obtained in step 1, and then normalizing the data; step 3, performing dimensionality reduction processing on the data after normalization in step 2 by using the kernel principal component analysis (KPCA) algorithm; step 4, constructing a main steam flow prediction model based on FWSN and RF, FWSN extracts features of the feature vector of the main steam flow, and then inputs the extracted feature vector into the random forest algorithm model, and the main steam flow prediction model based on FWSN and RF makes a prediction according to the trained decision tree, and outputs the predicted value of the main steam flow;

[0020] The data processed in step 3 are divided into a training set and a test set in an 8:2 ratio, where the data in the training set is used to train the main steam flow prediction model based on FWSN and RF. In step 5, the data to be predicted in the test set is input into the main steam flow prediction model based on FWSN and RF trained in step 4, and the main steam flow measurement value of the thermal power unit is output.

[0021] Furthermore, in step 2, the missing values ​​and outliers in the original data are replaced by linear interpolation results according to the time tags, and the normalization formula is:

[0022] ;

[0023] in, , Represent the minimum and maximum values ​​of the i-th variable, respectively. represents the normalized value, Represents the original data.

[0024] Furthermore, in step 3, the dimensionality reduction process of the kernel principal component analysis (KPCA) algorithm includes: using the kernel function to obtain random high-order correlations between input variables, and mapping the original data set into a high-dimensional feature space, so that the originally linearly inseparable data becomes linearly separable in the new feature space.

[0025] Furthermore, in step 4, the fractional-order wavelet scattering network can effectively extract the complex nonlinear features in the main steam flow-related parameters, and can better capture the inherent laws of the data compared to traditional feature extraction methods.

[0026] Furthermore, in step 4, the random forest algorithm has good generalization and anti-interference capabilities, and can improve the stability and accuracy of the main steam flow prediction.

[0027] Furthermore, in step 4, the training of the main steam flow prediction model based on FWSN and RF includes a hyperparameter tuning process: a grid search technique is used to further optimize the hyperparameters of the main steam flow prediction model based on FWSN and RF.

[0028] Furthermore, the hyperparameter tuning process includes:

[0029] (1) Determine the optimal value of 1.1 in the fractional order candidate set [0.1, 0.6, 1.1, 1.6];

[0030] (2) Select 4 as the optimal number of layers from the candidate set of decomposition layers [3, 4, 5];

[0031] (3) Determine the optimal value of 100 in the candidate set of random forest tree numbers [50, 100, 200];

[0032] (4) Select 10 as the optimal depth from the maximum depth candidate set of the tree [None, 10, 20].

[0033] The present invention also provides a main steam flow measurement system based on FWSN and RF, including: a data acquisition module, which collects original data of operating parameters related to the main steam flow from the unit's distributed control system; a preprocessing module, which fills missing values ​​and abnormal values ​​in the original data and then normalizes the data; a model calculation module, which deploys a main steam flow prediction model based on FWSN and RF and performs prediction calculations; and a visualization terminal, which displays real-time main steam flow measurement values ​​and historical trend curves.

[0034] The present invention will be further explained below with reference to the accompanying drawings and embodiments.

[0035] Example

[0036] The main steam flow measurement method based on FWSN and RF in this embodiment includes the following steps: Step 1: Collecting raw data of operating parameters related to the main steam flow from the unit's distributed control system DCS. Modern large-scale power generation units are generally reheat units. After pretreatment, the feed water is sent into the boiler through a feed water pump to exchange heat with the fuel, generating high-temperature and high-pressure steam. After entering the superheater through the steam pipe, it is heated again to form superheated steam that meets specific pressure and temperature, and drives the turbine to rotate and perform work, thereby driving the generator to convert mechanical energy into electrical energy. After the energy is released, part of the steam is extracted for reheating cycle or heat supply, while the remaining steam (exhaust steam) is condensed into water through the condenser and sent back to the feed water pump, forming a closed cycle.

[0037] The above analysis of the steam power cycle reveals that the generation and consumption of main steam for thermal power units involves complex processes and is influenced by multiple factors, including boiler fuel and combustion efficiency, boiler parameters, water circulation system, power generation load, auxiliary equipment requirements, and steam extraction. Based on the recommendations of mechanism and operation experts, 18 variables were initially selected as auxiliary variables for this simulation experiment, as shown in Table 1:

[0038] Table 1 List of auxiliary variables

[0039]

[0040] Taking a 1000MW ultra-supercritical single-reheat power generation unit as the research object, the historical data of the unit from 4:13 on March 11 to 14:33 on March 13, 2024 were extracted from the unit's distributed control system DCS, totaling 7000 groups of samples with a sampling time of 30s. Figure 2 It can be seen from the load curve shown that this set of data basically covers various operating conditions of the target unit under stable operation and variable load (including load increase and load decrease operation), and the unit load range is [330MW, 1000MW].

[0041] Step 2: Fill in the missing values ​​and outliers in the original data obtained in step 1, and then normalize the data.

[0042] Faced with a large amount of raw data collected, some data may be missing or abnormal due to various factors such as the complex production environment of thermal power plants and signal noise. In this case, we first replaced missing and abnormal values ​​in the data using linear interpolation results based on time tags to ensure data integrity and accuracy. In addition, to avoid gradient explosion in the model output and improve model accuracy and training efficiency, the raw input data was normalized. The calculation formula is shown above.

[0043] Step 3: The normalized data is subjected to dimensionality reduction through kernel principal component analysis (KPCA).

[0044] Since there are both linear and nonlinear relationships between the 18 candidate variables shown in Table 1, in order to prevent the curse of dimensionality, the KPCA method is used to further reduce the dimensionality of the feature data. The main principles are as follows:

[0045] Principal Component Analysis (PCA) is a data analysis method widely used for dimensionality reduction. It projects the original data into a new low-dimensional space through linear transformation to achieve the purpose of converting high-dimensional data sets into low-dimensional representations while retaining the key information in the data set. However, due to the complexity of thermal power unit systems, the large amount of operating data generated contains complex nonlinear and strongly coupled relationships, resulting in PCA possibly failing to fully capture the nonlinear characteristics of thermal power unit data. In addition, abnormal operations or measurement errors may exist during unit operation, and these outliers may further adversely affect the results of PCA. In contrast, Kernel Principal Component Analysis (KPCA), as a nonlinear multivariate statistical technique, has better applicability when processing unit data. KPCA uses kernel functions to obtain random high-order correlations between input variables and maps the original data set into a high-dimensional feature space, thereby making the originally linearly inseparable data linearly separable in the new feature space. The algorithm principle is as follows:

[0046] Assume that for the N sample points after standardization , where each sample point has M features, then the original data set A can be expressed as:

[0047] ;

[0048] In the formula, M is the data dimension, also the feature number, R MRepresents the input space, and i represents the sequence number of the sample point.

[0049] In order to conveniently map the original input space to the high-dimensional feature space F through nonlinearity, a nonlinear mapping is defined :

[0050] ;

[0051] Among them, Q represents the dimension of the high-dimensional feature space data sample, Q>M, Represents a space of dimension Q.

[0052] Further through Will Map to , It represents the result of mapping the i-th sample point to the high-dimensional space, which realizes the transformation of the original data set from the low-dimensional feature space to the high-dimensional feature space. Therefore, the covariance matrix C of the mapped data in the high-dimensional feature space F is F As shown below:

[0053] ;

[0054] Where, the superscript T represents the transpose of the matrix. Represents the result of mapping the i-th sample point to the high-dimensional space, the covariance matrix C F It is a Q-order square matrix, assuming that its eigenvalue is , then the eigenvector V satisfies:

[0055] ;

[0056] Furthermore, when When , the eigenvector V can be linearly expressed as:

[0057] ;

[0058] in, represents the linear combination coefficient.

[0059] Define the kernel function matrix K, let , where i, j represent the sequence numbers of the sample points, Represents the high-dimensional inner product of the sample pair. By combining the above three equations, we can simplify it to get:

[0060] Nλα=Kα,α=[ α 1 , α 2 ,…, α N ] T ;

[0061] in, Represents a vector of N linear combination coefficients, and the superscript T represents the transpose of the matrix.

[0062] At this point, the problem of finding the eigenvector becomes the problem of finding the corresponding α value. KPCA uses the kernel function matrix to find the eigenvalues ​​and eigenvectors, and selects the eigenvector with the highest cumulative contribution rate as the new principal component to form a new feature space based on the eigenvalue, thereby achieving data dimensionality reduction.

[0063] The results of kernel principal component analysis on candidate variables are as follows: Figure 3 As shown in Figure 2, the contribution of the 13th to 18th principal components is relatively low and not obvious. Figure 3 Only the specific contribution rates of the first 12 principal components are shown. Figure 3 It can be seen that the principal component contribution rate represented by the first four principal components is the largest, among which the variance contribution rate of the first core principal component KPCA1 is as high as 40%. From the cumulative variance contribution rate curve, it can be seen that the cumulative contribution rate of the first eight principal components is 98%, which basically covers the main information of the original data. The graph of the contribution rate of the ninth and subsequent principal components is close to the horizontal, and the graph of the cumulative variance contribution rate is also close to flat, indicating that the contribution rate starting from the ninth core principal component is very low, and the amount of original information contained is very small. In summary, the first eight principal component variables are selected as the final input of the model. Step 4: Construct a main steam flow prediction model based on FWSN and RF. FWSN extracts the feature vector of the main steam flow, and then inputs the extracted feature vector into the RF model. The model makes predictions based on the trained decision tree and outputs the predicted value of the main steam flow.

[0064] The main steam flow prediction model based on FWSN and RF is trained based on the data processed in step 3.

[0065] The Fractional Wavelet Scattering Network (FWSN) is implemented using a deep convolutional network. Similar to a standard deep convolutional network, it extracts signal features layer by layer through a linear transformation (convolution kernel) followed by a nonlinear transformation (modulo operator). The convolution kernel of this network is a predefined fractional wavelet filter, and features are extracted by rotating and scaling the mother wavelet. Compared to traditional scattering networks, the FWSN adds fractional-order parameters. By adjusting these parameters, fractional frequencies can be obtained to extract optimal signal features. The fractional wavelet scattering coefficients of each layer are calculated as the network output, and the corresponding fractional wavelet transmission signal of each layer serves as the input to the next layer. Figure 4 is a schematic diagram of a 3-layer fractional-order wavelet scattering network. Figure 4 The ○ in the middle represents the intermediate result of scattering propagation, and ● represents the output. is the input signal, S [.] is the fractional-order scattering output operator, U [.] is the fractional-order scattering propagation operator, Represents complex modulo. is an empty set; is a Gaussian low-pass filter, is a specific complex wavelet, where represents the total number of wavelet bases, represents the eigenvalue of the sth wavelet basis. , where m represents the number of layers of the scattering path, represents the eigenvalue of the mth layer of the scattering path, then all scattering coefficients can be calculated by the following steps:

[0066] First layer output:

[0067] f(t) ⇒ S[∅]f=f Θ α ϕ 2 J ;

[0068] in, represents the input signal at time t, represents the fractional-order transform operator, represents the order of the fraction, is the spatial scale variable.

[0069] Second layer output:

[0070] f Θ α ψ j 1 , r 1 ⇒ S[ λ 1 ]f= f Θ α ψ j 1 , r 1 Θ α ϕ 2 J ;

[0071] in, represents the complex wavelet basis function, represents the size index of the first layer wavelet, Indicates the channel index of the first layer wavelet.

[0072] The third layer output:

[0073] ;

[0074] in, represents the size index of the second layer wavelet, Indicates the channel index of the second layer wavelet.

[0075] No. Layer output:

[0076] ;

[0077] in, Indicates the The size index of the layer wavelet, Indicates the Channel index of the layer wavelet.

[0078] like Figure 5 As shown in the figure, the Random Forest (RF) algorithm is an ensemble learning solution based on decision trees. It solves classification and regression problems by constructing multiple decision trees, and has high accuracy and good generalization ability. The Random Forest algorithm obtains multiple different samples (sample 1-sample H) through data input and forms multiple different decision trees ( Figure 5 Decision tree 1-decision tree H in the dataset) is trained independently on a randomly selected part. Since the model features selected by each decision tree are different, multiple results can be obtained. Based on the results of each decision tree, the final result is obtained by voting / averaging.

[0079] Due to the fact that thermal power units often run for a long time without stopping, the historical data they generate usually have a long time series, a large amount of data and complex relationships. Therefore, the present invention combines the powerful feature extraction ability of the fractional-order wavelet scattering network and the good generalization ability and anti-interference ability of the random forest algorithm to construct a fractional-order wavelet scattering network and a random forest model to predict the main steam flow. Compared with the existing prediction model, this model avoids the gradient vanishing and gradient explosion problems caused by time series prediction, and the random forest algorithm can optimize and integrate the features extracted by the fractional-order wavelet scattering network, further making the prediction results more accurate. At the same time, the fractional-order wavelet scattering network can effectively reduce the computational complexity, improve the performance and generalization ability of the model, and facilitate the training of the model. The main steam flow prediction model based on FWSN and RF is constructed as follows. Figure 6 As shown in Figure 1, the network's main structure consists of data preprocessing, model construction, model training, and result verification. The prediction process is as follows: first, data preprocessing, including operating data collection, data processing, normalization, KPCA dimensionality reduction, and partitioning into training and test sets; then, model construction, extracting scattering characteristic coefficients from a fractional-order scattering network and inputting them into the RF model; finally, the prediction data obtained from the trained model is used for result verification.

[0080] During the experiment, 5600 groups of the processed original data set were selected as training sets and 1400 groups of data were selected as test sets according to the ratio of 8:2 to train the constructed model. MAPE , root mean square error ERMSE and the coefficient of determination R 2 Three indicators are used to evaluate the model performance standards, which are defined as follows:

[0081] ;

[0082] ;

[0083] ;

[0084] Where, 、 、 They represent the actual value, predicted value and the mean of the actual value respectively, and n represents the total number of samples.

[0085] To further improve the model's prediction accuracy and fitting performance, a grid search technique was used to optimize its hyperparameters. The optimal value for the fractional order was 1.1; the optimal number of decomposition layers was 4; the optimal number of random forest trees was 100; and the optimal maximum tree depth was 10.

[0086] Step 5: Input the data to be predicted in the test set into the main steam prediction model based on FWSN and RF trained in step 4, and output the measured value of the main steam flow of the thermal power unit.

[0087] Based on the aforementioned hyperparameter optimization results, a main steam flow rate model based on FWSN and RF was used to predict the test data. The fractional wavelet scattering network and random forest model constructed by this invention accurately predicted its changing trends. The model prediction results were able to identify the turning points of the actual main steam flow rate changes, maintain stability during the period of stable flow rate, and maintain a good fit with the actual value.

[0088] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the above-mentioned main steam flow measurement method based on FWSN and RF.

[0089] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the above-mentioned main steam flow measurement method based on FWSN and RF are implemented.

[0090] 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 disk drives, CD-ROMs, optical storage devices, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention may be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0091] 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.

[0092] 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.

[0093] 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 The steps for the function specified in one or more boxes.

[0094] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0095] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A main steam flow measurement method based on FWSN and RF, characterized in that: The following steps are involved: Step 1: Collecting raw data of operating parameters related to main steam flow from the unit distributed control system; Step 2: Fill in the missing values ​​and outliers in the original data obtained in step 1, and then normalize the data; Step 3: Perform dimensionality reduction on the normalized data in step 2 using the kernel principal component analysis algorithm; Step 4: Construct a main steam flow prediction model based on FWSN and RF. FWSN extracts the feature vector of the main steam flow and then inputs the extracted feature vector into the RF model. The RF model makes predictions based on the trained decision tree and outputs the predicted value of the main steam flow. FWSN stands for fractional wavelet scattering network, and RF stands for random forest. The fractional wavelet scattering network is implemented through a deep convolutional network, which first extracts the signal features layer by layer through linear transformation and then nonlinear transformation. The convolution kernel of the fractional wavelet scattering network is a pre-given fractional wavelet filter, and the features are extracted by rotating and scaling the mother wavelet. The fractional-order wavelet scattering network adds fractional-order parameters, and the fractional-order frequency is obtained by adjusting the fractional-order parameters, thereby extracting the optimal signal characteristics. The fractional-order wavelet scattering coefficient of each layer is calculated as the network output of each layer, and the fractional-order wavelet transmission signal corresponding to each layer is used as the input of the next layer network. The training of the main steam flow prediction model based on FWSN and RF includes a hyperparameter tuning process, that is, the grid search technology is used to further optimize the hyperparameters of the main steam flow prediction model based on FWSN and RF; The hyperparameter optimization process includes: (1) Determine the optimal value of 1.1 in the fractional order candidate set [0.1, 0.6, 1.1, 1.6]; (2) Select 4 as the optimal number of layers from the candidate set of decomposition layers [3, 4, 5]; (3) Determine the optimal value of 100 in the candidate set of random forest tree numbers [50, 100, 200]; (4) Select 10 as the optimal depth from the maximum depth candidate set of the tree [None, 10, 20]; Based on the data after dimensionality reduction in step 3, the main steam flow prediction model based on FWSN and RF is trained; Step 5: Input the data to be measured into the main steam flow prediction model based on FWSN and RF that has been trained in step 4, and output the main steam flow measurement value of the thermal power unit.

2. The main steam flow measurement method based on FWSN and RF according to claim 1 is characterized in that: In step 2, the missing values ​​and abnormal values ​​in the original data are replaced by linear interpolation results according to the time labels.

3. The main steam flow measurement method based on FWSN and RF according to claim 1 is characterized in that: In step 3, the dimensionality reduction process of the kernel principal component analysis algorithm includes: using the kernel function to obtain random high-order correlations between input variables, and mapping the set of original data into a high-dimensional feature space, so that the original linearly inseparable original data becomes linearly separable in the new high-dimensional feature space.

4. The main steam flow measurement method based on FWSN and RF according to claim 1 is characterized in that: In step 4, the FWSN extracts complex nonlinear features in the operating parameters related to the main steam flow.

5. The main steam flow measurement method based on FWSN and RF according to claim 1 is characterized in that: In step 4, the RF model improves the stability and accuracy of the main steam flow prediction.

6. A main steam flow measurement system for implementing the main steam flow measurement method based on FWSN and RF according to any one of claims 1 to 5, characterized in that: include: Data acquisition module, which collects raw data of operating parameters related to main steam flow from the unit's distributed control system; The preprocessing module fills the missing values ​​and outliers in the original data, and then performs normalization and dimensionality reduction; Model operation module, deploys the main steam flow prediction model based on FWSN and RF, and performs prediction calculations; Visual terminal displays real-time main steam flow measurement values ​​and historical trend curves.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the main steam flow measurement method based on FWSN and RF are implemented as described in any one of claims 1 to 5.

Images