Thermal power plant boiler efficiency prediction system and method based on neural network

The BP neural network model predicts the boiler efficiency of thermal power plants, solves the wear and energy loss problems caused by long-term boiler operation, and achieves more efficient energy utilization and environmental protection.

CN117875510BActive Publication Date: 2024-11-29NORTH CHINA ELECTRIC POWER UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202410065804.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-01-17
Publication Date
2024-11-29
Estimated Expiration
2044-01-17

Smart Images

  • Figure CN117875510B_ABST
    Figure CN117875510B_ABST
Patent Text Reader

Abstract

The present invention discloses a thermal power plant boiler efficiency prediction system and method based on a neural network, relates to the technical field of thermal power plants, and includes the following steps: S100, collecting historical data during the operation of the thermal power plant boiler, cleaning the collected data, and selecting features related to the boiler efficiency. The present invention improves the prediction accuracy by using deep learning models such as BP neural networks, and by calculating the Spearman correlation coefficients between coal-fired indicators and boiler efficiency, and between operating indicators and boiler efficiency, the influence of each feature on the boiler efficiency can be understood. According to the results of the correlation coefficients, the weights related to these features in the neural network are adjusted, the model is further optimized, and the prediction accuracy is further improved. By predicting the boiler efficiency, the thermal power plant can adjust the operating parameters more accurately, achieve more efficient energy conversion, improve energy utilization, and reduce energy waste.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of thermal power plants, and in particular to a system and method for predicting boiler efficiency in thermal power plants based on neural networks. Background Art

[0002] With the rapid development of the economy, the process of urbanization is also constantly improving. At the same time, people have higher demands for electricity both at work and in life, so they have higher requirements for power plants. As far as the current situation is concerned, my country's power generation industry mainly chooses to use coal and fuel oil for thermal power generation, and in the process of conversion, boilers play an important role. In thermal power plants, boilers play a very important role, that is, converting the chemical energy of fuel into thermal energy, thereby providing corresponding electrical energy and thermal energy, that is, resources such as coal or oil, through conversion, converting them into thermal energy. Some of these thermal energies will be directly used to provide thermal energy resources, while others will be converted into electrical energy through some means.

[0003] Predicting the efficiency of thermal power plant boilers has the following (but not limited to) benefits:

[0004] Long-term operation of thermal power plant boilers can lead to problems such as wear and dust accumulation, which affect boiler performance. By predicting boiler efficiency, problems that may affect efficiency can be discovered in advance, so that maintenance measures can be taken in time to keep the equipment in the best condition.

[0005] There are certain losses in energy conversion in thermal power plants. By understanding the boiler efficiency in advance, targeted energy-saving and emission reduction measures can be taken, including optimizing the combustion process, improving thermal energy conversion equipment, reducing energy waste, etc. Predicting boiler efficiency can help thermal power plants better implement energy management strategies and reduce environmental impact.

[0006] By predicting the efficiency of thermal power plant boilers, thermal power plants can operate more efficiently and sustainably, meeting the growing demand for electricity while reducing environmental impact. Therefore, there is an urgent need for a thermal power plant boiler efficiency prediction system to solve the above-mentioned problem of thermal power plant boiler efficiency prediction.

[0007] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0008] The purpose of the present invention is to provide a system and method for predicting the efficiency of a thermal power plant boiler based on a neural network. By using deep learning models such as a BP neural network, the complex nonlinear relationship between the efficiency of a thermal power plant boiler and coal burning indicators and operating indicators can be better captured, and the prediction accuracy can be improved. By calculating the Spearman correlation coefficient between the coal burning indicator and the boiler efficiency, and the operating indicator and the boiler efficiency, the influence of each feature on the boiler efficiency can be understood. According to the result of the correlation coefficient, the weights related to these features in the neural network are adjusted, the model is further optimized, and the prediction accuracy is further improved. By predicting the boiler efficiency, the thermal power plant can adjust the operating parameters more accurately, achieve more efficient energy conversion, improve energy utilization, and reduce energy waste, so as to solve the problems in the above-mentioned background technology.

[0009] In order to achieve the above object, the present invention provides the following technical solution: A method for predicting boiler efficiency in a thermal power plant based on a neural network, comprising the following steps:

[0010] S100, collecting historical data during the operation of the thermal power plant boiler, cleaning the collected data, and selecting features related to boiler efficiency;

[0011] S200, after the data is standardized, a BP neural network model is constructed through the standardized data, and the efficiency of the boiler of the thermal power plant is predicted through the BP neural network model;

[0012] S300, evaluating the performance of the BP neural network model in predicting the efficiency of thermal power plant boilers, and increasing the amount of training data according to the evaluation results to enable the BP neural network model to generalize new data or improve the model complexity;

[0013] S400, calculating the Spearman correlation coefficients between the coal burning index and the boiler efficiency, and between the operation index and the boiler efficiency, and adjusting the weights of the features related to the coal burning index and the operation index in the neural network according to the results of the Spearman correlation coefficients.

[0014] Preferably, in step S200, the specific steps of data standardization are as follows:

[0015] a. Calculate the minimum and maximum values ​​of each feature through the Min-Max standardization method;

[0016] b. Apply the normalization formula to each feature and calculate the normalized value. The formula is: , where represents the normalized value, is the original value, is the minimum value of the data, is the maximum value of the data.

[0017] Preferably, in step S200, the steps of constructing the BP neural network model are as follows:

[0018] S201, determine the number of layers of the neural network and the number of nodes in each layer, and select an activation function for each layer;

[0019] S202, initializing the weights of the neural network;

[0020] S203, select mean square error as the loss function and gradient descent as the optimization algorithm.

[0021] Preferably, in step S300, the information entropy, KL divergence and model complexity of the BP neural network model for thermal power plant boiler efficiency prediction are obtained, and a comprehensive analysis model is established for the information entropy, KL divergence and model complexity to generate a fitting index , by fitting the index The performance of BP neural network model in predicting boiler efficiency of thermal power plant was evaluated according to the fitting index. Increasing the amount of training data allows the BP neural network model to generalize new data or increase the complexity of the model, based on the formula: , where , , They represent information entropy, KL divergence and model complexity respectively. , , are the preset proportional coefficients of information entropy, KL divergence and model complexity, respectively, and , , Both are greater than 0.

[0022] Preferably, the calculation expression of information entropy is: ,in, represents the information entropy of random variable X, n represents the number of possible values ​​of the random variable, n is a positive integer, The random variable X takes the value The probability of

[0023] The calculation expression of KL divergence is: ,in, Indicates an event i The probability of being in distribution P, event i The probability of being in the distribution Q, represents the KL divergence of P relative to Q;

[0024] The calculation formula of model complexity is: , where represents the model complexity, k Indicates the number of parameters of the model.

[0025] Preferably, the fitting index generated by the BP neural network model when predicting the efficiency of the thermal power plant boiler is compared and analyzed with the pre-set fitting index reference threshold. If the fitting index is greater than or equal to the fitting index reference threshold, the amount of training data is increased to enable the BP neural network model to generalize the new data and improve the generalization ability of the new data; if the fitting index is less than the fitting index reference threshold, the complexity of the model is increased.

[0026] Preferably, in step S400, the Spearman correlation coefficient between the coal combustion index and the boiler efficiency is calculated by the following steps:

[0027] S401, obtaining information related to coal combustion index and boiler efficiency from historical data during operation of the thermal power plant boiler;

[0028] S402, sorting the data of coal burning index and boiler efficiency in ascending order respectively, recording the original serial number of each data, and assigning a rank to each sorted data;

[0029] S403. Calculate the Spearman correlation coefficient between the coal burning index and the boiler efficiency. The calculation expression is: , where d represents the rank difference of each pair of coal combustion index and boiler efficiency, ,in, x It is the coal burning index. y is the boiler efficiency, n Indicates the sample size.

[0030] Preferably, the Spearman correlation coefficient between the operating index and the boiler efficiency is calculated by the following steps:

[0031] S411, obtaining information related to operating indicators and boiler efficiency from historical data during the operation of the thermal power plant boiler;

[0032] S412, sorting the data of the operating index and boiler efficiency in ascending order, recording the original serial number of each data, and assigning a rank to each sorted data;

[0033] S413. Calculate the Spearman correlation coefficient between the operating index and the boiler efficiency. The calculation expression is: , where h represents the rank difference between each pair of operating index and boiler efficiency, ,in, k is the operating indicator, v is the boiler efficiency, m Indicates the sample size.

[0034] Preferably, in the process of adjusting the weights of the features related to the coal combustion index and the operation index in the neural network based on the result of the Spearman correlation coefficient, the gradient descent method is used to update the weights. The specific steps are as follows:

[0035] S421. For each characteristic related to coal combustion index and operation index i , calculate the gradient of the loss function with respect to the feature weight, and the calculation expression is: , the gradient represents the direction of change of the loss function with respect to the weight, where J is the mean square error loss function, Is with characteristics i The relevant weights, N is the sample size, is the true value, is the predicted value of the model;

[0036] S422, calculating the weight adjustment amount according to the gradient, the calculation formula is: , where the learning rate is a hyperparameter that controls the update step size of the weights in each iteration;

[0037] S423. For each characteristic related to coal combustion index and operation index i , use the following formula to update the weight, the calculation formula is: , where is the updated weight;

[0038] S424. Repeat the above steps S421-S423 until the model converges or reaches the set number of training times.

[0039] The thermal power plant boiler efficiency prediction system based on neural network includes data preprocessing module, neural network modeling module, generalization improvement module and feature weight adjustment module;

[0040] The data preprocessing module collects historical data during the operation of thermal power plant boilers, cleans the collected data, and selects features related to boiler efficiency;

[0041] The neural network modeling module standardizes the data, builds a BP neural network model through the standardized data, and predicts the efficiency of the thermal power plant boiler through the BP neural network model;

[0042] The generalization improvement module evaluates the performance of the BP neural network model in predicting the efficiency of thermal power plant boilers, and increases the amount of training data based on the evaluation results to enable the BP neural network model to generalize new data or improve the model complexity;

[0043] The feature weight adjustment module calculates the Spearman correlation coefficients between the coal burning index and the boiler efficiency, and between the operation index and the boiler efficiency, and adjusts the weights of the features related to the coal burning index and the operation index in the neural network according to the results of the Spearman correlation coefficients.

[0044] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0045] By using deep learning models such as BP neural networks, the present invention can better capture the complex nonlinear relationship between the boiler efficiency of a thermal power plant and the coal burning index and the operating index, and improve the prediction accuracy. By calculating the Spearman correlation coefficient between the coal burning index and the boiler efficiency, and the operating index and the boiler efficiency, the influence of each feature on the boiler efficiency can be understood. According to the result of the correlation coefficient, the weights related to these features in the neural network are adjusted to further optimize the model and further improve the prediction accuracy. By predicting the boiler efficiency, the thermal power plant can adjust the operating parameters more accurately, achieve more efficient energy conversion, improve energy utilization, and reduce energy waste.

[0046] The performance of the BP neural network model of the present invention when predicting the efficiency of a thermal power plant boiler is monitored, and a comprehensive analysis model is established for information entropy, KL divergence and model complexity to generate a fitting index. The fitting of the prediction process is evaluated by the fitting index. During the evaluation, if it is found that the model has overfitting or underfitting problems, the generalization ability of the model is improved by increasing the amount of training data or adjusting the model complexity, so that it can better adapt to new data. By more accurately predicting the efficiency of thermal power plant boilers, the operating parameters can be better adjusted, the effective use of energy can be improved, the energy cost can be reduced, and the overall energy efficiency can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0048] Figure 1 The present invention is a method flow chart of the system and method for predicting boiler efficiency in a thermal power plant based on a neural network.

[0049] Figure 2 It is a module schematic diagram of the thermal power plant boiler efficiency prediction system and method based on neural network of the present invention.

[0050] Figure 3 Schematic diagram of a multi-layer feed-forward network. DETAILED DESCRIPTION

[0051] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0052] The present invention provides Figure 1 The method for predicting boiler efficiency of a thermal power plant based on a neural network includes the following steps:

[0053] S100, collecting historical data during the operation of the thermal power plant boiler, cleaning the collected data, and selecting features related to boiler efficiency;

[0054] The historical data of the thermal power plant boiler during operation, including fuel information, environmental conditions, boiler operating parameters, etc., are collected and cleaned to ensure data quality and integrity. Through data analysis and domain knowledge, key features related to boiler efficiency are selected. This may include coal combustion indicators (such as coal quality, sulfur content), operation indicators (such as boiler load, steam temperature), etc.

[0055] Here are some possible data cleaning methods:

[0056] Missing value processing: Detect and process missing values ​​in the data. You can delete missing values, fill in missing values ​​(such as using mean, median or other statistical values), etc. to ensure data integrity.

[0057] Outlier handling: Identify and handle outliers, which may be caused by sensor failure or other abnormal conditions. Statistical methods or domain knowledge can be used to detect outliers and handle them as appropriate, such as removing or repairing them.

[0058] Data deduplication: Check and delete duplicate data records to avoid affecting the model. Especially during the data collection process, there may be data redundancy caused by duplicate records.

[0059] Label outliers: For supervised learning problems, detect and handle outliers in the target variable (boiler efficiency) to ensure the accuracy of the target value.

[0060] Data type conversion: Ensure that all data is of the correct type and convert it as needed, such as converting text data to numeric data.

[0061] Time series processing: If the data contains time series information, ensure the continuity and consistency of the time series. It may be necessary to adjust the format of the timestamp, fill in missing time points, etc.

[0062] Feature engineering: Select features related to boiler efficiency based on domain knowledge and perform necessary feature engineering, such as creating new features and performing data transformation, to improve model performance.

[0063] Data normalization or standardization: Normalize or standardize the data to ensure that the scales of different features are consistent to avoid some features having too much influence on the model.

[0064] S200, after the data is standardized, a BP neural network model is constructed through the standardized data, and the efficiency of the boiler of the thermal power plant is predicted through the BP neural network model;

[0065] The purpose of standardizing the data is to ensure that the training process of the neural network is more stable;

[0066] Data standardization is to ensure that the difference in the numerical range between different features does not adversely affect the training of the neural network. The standardization process usually includes converting the data to a standard normal distribution with a mean of 0 and a standard deviation of 1, or scaling the data to a specific range. The specific steps are as follows:

[0067] a. Calculate the minimum and maximum values ​​of each feature through the Min-Max standardization method;

[0068] b. Apply the normalization formula to each feature and calculate the normalized value. The formula is: , where represents the normalized value, is the original value, is the minimum value of the data, is the maximum value of the data;

[0069] The steps to construct the BP neural network model are as follows:

[0070] S201, determine the number of layers of the neural network and the number of nodes in each layer. A typical BP neural network includes an input layer, several hidden layers and an output layer. An activation function is selected for each layer. Common ones include sigmoid, tanh, ReLU, etc. The output layer uses a linear activation function.

[0071] S202, initializing the weights of the neural network, which may be initialized by a random initialization method to ensure that the initial weights are neither too large nor too small;

[0072] S203, selecting mean square error as the loss function and gradient descent as the optimization algorithm;

[0073] BP (Back Propagation, BP) neural network is a multi-layer feedforward neural grid trained according to the error back propagation algorithm. It corrects the connection weights by adjusting the error between the target output and the actual to achieve the accuracy of network prediction. It is one of the most widely used neural grid models and the most classic and most used artificial neural network model. The basic idea of ​​BP neural network is the gradient descent method, which uses gradient search technology to minimize the mean square error between the actual output value and the expected output value of the network.

[0074] The calculation process of BP neural network consists of forward calculation process and reverse calculation process. In the forward propagation process, the input pattern is processed layer by layer from the input layer to the hidden layer and then transferred to the output layer. The state of each layer of neurons only affects the state of the next layer of neurons. If the expected output cannot be obtained in the output layer, it will enter the reverse propagation and return the error signal along the original connection path. By modifying the weights of each neuron, the error signal is minimized.

[0075] Multi-layer feed-forward network Figure 3 As shown, it contains d input neurons, q hidden neurons and l output neurons: the connection weights of each neuron between the input layer and the hidden layer, and between the hidden layer and the output layer are , , the inputs received by each neuron in the hidden layer and output layer are: , ; The thresholds of neurons in the hidden layer and output layer are , ; The hidden layer and output layer use activation functions , ; The outputs of each neuron in the hidden layer and output layer are: , .

[0076] BP neural network has strong nonlinear mapping ability and flexible network structure. The number of intermediate layers and the number of neurons in each layer of the network can be set arbitrarily according to the specific situation, and its performance varies with the difference of structure.

[0077] BP network is mainly used in the following four aspects.

[0078] 1) Function approximation: Use input vectors and corresponding output vectors to train a network to approximate a function.

[0079] 2) Pattern recognition: Use a pending output vector to link the output vector to the input vector.

[0080] 3) Classification: Classify the input vector in a suitable way as defined by it.

[0081] 4) Data compression: reducing the dimension of the output vector for easier transmission or storage.

[0082] S300, evaluating the performance of the BP neural network model in predicting the efficiency of thermal power plant boilers, and increasing the amount of training data according to the evaluation results to enable the BP neural network model to generalize new data or improve the model complexity;

[0083] Information entropy is a concept used to measure the uncertainty or amount of information in data. In the BP neural network model, information entropy is usually used to evaluate the uncertainty of the model prediction results. A large or small information entropy will cause the BP neural network model to have overfitting or underfitting problems when predicting the efficiency of thermal power plant boilers. The following is a specific explanation:

[0084] Information entropy is large:

[0085] May lead to overfitting: If the model's prediction results for thermal power plant boiler efficiency show a large information entropy, this may indicate that the model has a large uncertainty or confusion in the prediction. In extreme cases, a large information entropy may indicate that the model is too flexible and tries to adapt to the details of the training data, resulting in overfitting.

[0086] Information entropy is small:

[0087] May lead to underfitting: If the model is too simple, the information entropy may be small, indicating that the model cannot capture the complex relationship of the training data well, which may lead to insufficient fitting on the training data, that is, underfitting. The expressive ability of the model is limited and cannot capture the true pattern of the data.

[0088] KL (Kullback-Leibler) divergence is an indicator used to measure the difference between two probability distributions. The size of KL divergence may be related to the fit of the model. The following is a specific explanation:

[0089] KL divergence is large:

[0090] May lead to overfitting: When the KL divergence is large, it means that there is a large difference between the predicted distribution of the model and the actual training data distribution. This may indicate that the model is too sensitive to the training data, capturing the noise or details of the training data and failing to generalize well to new data. An overfitted model may be too complex for the training data and cannot generalize well to unseen data.

[0091] KL divergence is small:

[0092] May lead to underfitting: When the KL divergence is small, it means that the predicted distribution of the model is similar to the actual training data distribution, but this does not necessarily mean that the model performance is good. A smaller KL divergence may indicate that the model is too simple and fails to capture the complex relationships in the training data, resulting in underfitting.

[0093] Model complexity refers to the learning ability of the model, which is usually related to the number of parameters in the model. The complexity of the BP neural network can be measured by the number of layers of the neural network and the number of neurons in each layer. A large or small model complexity will cause the BP neural network model to have overfitting or underfitting problems when predicting the efficiency of thermal power plant boilers. The following is a specific explanation:

[0094] The model is more complex:

[0095] May lead to overfitting: When the model complexity is large, the network may be too flexible and able to remember the details and noise of the training data, but fail to capture the true data generation distribution, which may lead to poor performance on unseen data, that is, overfitting. In the prediction of boiler efficiency in thermal power plants, overfitting may manifest as being too sensitive to specific training data, but performing poorly in other cases.

[0096] The model complexity is small:

[0097] May lead to underfitting: When the model complexity is small, the model may not be able to capture complex relationships and make full use of the information in the training data, which may cause the model to perform well on the training set but poor generalization performance on the test set or new data, that is, underfitting.

[0098] Obtain the information entropy, KL divergence and model complexity of the BP neural network model for thermal power plant boiler efficiency prediction, establish a comprehensive analysis model based on the information entropy, KL divergence and model complexity, and generate a fitting index , based on the formula: , where , , They represent information entropy, KL divergence and model complexity respectively. , , are the preset proportional coefficients of information entropy, KL divergence and model complexity, respectively, and , , All are greater than 0;

[0099] The calculation expression of information entropy is: ,in, represents the information entropy of random variable X, n represents the number of possible values ​​of the random variable, n is a positive integer, The random variable X takes the value The probability of

[0100] The calculation expression of KL divergence is: ,in, Indicates an event i The probability of being in distribution P, event i The probability of being in the distribution Q, represents the KL divergence of P relative to Q;

[0101] The calculation formula of model complexity is: , where represents the model complexity, k Indicates the number of parameters of the model;

[0102] It can be seen from the formula that the greater the information entropy, the greater the KL divergence, and the greater the model complexity when the BP neural network model is used to predict the efficiency of thermal power plant boilers, that is, the fitting index generated by the BP neural network model when predicting the efficiency of thermal power plant boilers is The larger the performance value, the higher the risk of overfitting when the BP neural network model is used to predict the efficiency of thermal power plant boilers. The smaller the information entropy, KL divergence and model complexity of the BP neural network model are, that is, the fitting index generated by the BP neural network model when predicting the efficiency of thermal power plant boilers. The smaller the performance value, the lower the risk of overfitting when the BP neural network model is used to predict the efficiency of thermal power plant boilers.

[0103] The fitting index generated by the BP neural network model when predicting the efficiency of thermal power plant boilers is compared and analyzed with the pre-set fitting index reference threshold. If the fitting index is greater than or equal to the fitting index reference threshold, the amount of training data is increased to enable the BP neural network model to generalize new data and improve its generalization ability for new data. More data helps to alleviate overfitting, especially when the amount of data is small, the model tends to remember the training set too much and cannot generalize to new data; if the fitting index is less than the fitting index reference threshold, the complexity of the model is increased. Increasing the complexity of the model can improve its fitting ability for the training data and make it better adapt to the characteristics of the data. Appropriately increasing the complexity of the model can alleviate the underfitting problem and enable it to better capture the complex relationships in the data.

[0104] The performance of the BP neural network model of the present invention when predicting the efficiency of a thermal power plant boiler is monitored, and a comprehensive analysis model is established for information entropy, KL divergence and model complexity to generate a fitting index. The fitting of the prediction process is evaluated by the fitting index. During the evaluation, if it is found that the model has overfitting or underfitting problems, the generalization ability of the model is improved by increasing the amount of training data or adjusting the model complexity, so that it can better adapt to new data. By more accurately predicting the efficiency of thermal power plant boilers, the operating parameters can be better adjusted, the effective use of energy can be improved, the energy cost can be reduced, and the overall energy efficiency can be improved.

[0105] S400, calculating the Spearman correlation coefficient between the coal combustion index (such as coal quality, sulfur content) and boiler efficiency, and between the operation index (such as boiler load, steam temperature) and boiler efficiency, and adjusting the weights of the features related to the coal combustion index and the operation index in the neural network according to the result of the Spearman correlation coefficient;

[0106] Spearman correlation coefficient between coal burning index and boiler efficiency:

[0107] Spearman correlation coefficient: Measures the monotonic relationship between two variables and does not require the variables to be normally distributed.

[0108] Effect: If there is a strong monotonic relationship between the coal burning index and boiler efficiency, the Spearman correlation coefficient will be close to 1 or -1.

[0109] If there is a correlation, it means that these coal combustion indicators have a high amount of information for the prediction of boiler efficiency.

[0110] By adjusting the weights of the features related to these indicators in the neural network, the influence of these features in the model can be enhanced, thereby improving the model's prediction accuracy for boiler efficiency.

[0111] The Spearman correlation coefficient between coal burning index and boiler efficiency is calculated by the following steps:

[0112] S401. Obtain information related to coal combustion indicators (such as coal quality and sulfur content) and boiler efficiency from historical data during the operation of thermal power plant boilers;

[0113] S402, sorting the coal burning index and boiler efficiency data in ascending order, recording the original serial number of each data, and assigning a rank to each sorted data (i.e., its position in the sorted data set);

[0114] S403. Calculate the Spearman correlation coefficient between the coal burning index and the boiler efficiency. The calculation expression is: , where d represents the rank difference of each pair of coal combustion index and boiler efficiency, ,in, x It is the coal burning index. y is the boiler efficiency, n represents the number of samples;

[0115] if It is close to 1, indicating that there is a strong positive correlation between coal combustion index and boiler efficiency.

[0116] if A value close to -1 indicates a strong negative correlation.

[0117] if A value close to 0 indicates that the relationship between the two is weak or irrelevant.

[0118] Spearman correlation coefficient between operating indicators and boiler efficiency:

[0119] Spearman correlation coefficient: measures the monotonic relationship between two variables, unaffected by outliers.

[0120] Effect: If there is a strong monotonic relationship between the operating indicator and boiler efficiency, the Spearman correlation coefficient will be close to 1 or -1.

[0121] If there is a correlation, it means that these operating indicators have a high amount of information for the prediction of boiler efficiency.

[0122] By adjusting the weights of the features related to these indicators in the neural network, the influence of these features in the model can be enhanced, thereby improving the model's prediction accuracy for boiler efficiency.

[0123] The Spearman correlation coefficient between the operating index and boiler efficiency is calculated by the following steps:

[0124] S411. Obtaining information related to operating indicators (such as boiler load, steam temperature) and boiler efficiency from historical data during the operation of a thermal power plant boiler;

[0125] S412, sorting the data of the operating index and boiler efficiency in ascending order respectively, recording the original serial number of each data, and assigning a rank to each sorted data (i.e., its position in the sorted data set);

[0126] S413. Calculate the Spearman correlation coefficient between the operating index and the boiler efficiency. The calculation expression is: , where h represents the rank difference between each pair of operating index and boiler efficiency, ,in, k is the operating indicator, v is the boiler efficiency, m represents the number of samples;

[0127] if It is close to 1, indicating that there is a strong positive correlation between the operating index and boiler efficiency.

[0128] if A value close to -1 indicates a strong negative correlation.

[0129] if A value close to 0 indicates that the relationship between the two is weak or irrelevant.

[0130] Adjust the weights of relevant features in the neural network:

[0131] Function: According to the result of Spearman correlation coefficient, the weights of the features related to coal combustion index and operation index in the neural network are adjusted, so that the features with higher correlation are more important in the model.

[0132] The weight adjustment can be achieved by increasing or decreasing the weight value of the relevant features, so as to better capture the impact of coal burning and operating indicators on boiler efficiency. Such adjustment helps to improve the flexibility of the model, enabling it to better adapt to different operating conditions and data distributions, and improve the generalization ability of the model.

[0133] In the process of adjusting the weights of the features related to the coal combustion index and the operation index in the neural network based on the results of the Spearman correlation coefficient, the gradient descent method is used to update the weights. The specific steps are as follows:

[0134] S421. For each characteristic related to coal combustion index and operation index i , calculate the gradient of the loss function with respect to the feature weight, and the calculation expression is: , the gradient represents the direction of change of the loss function with respect to the weight, where J is the mean square error loss function, Is with characteristics i The relevant weights, N is the sample size, is the true value, is the predicted value of the model;

[0135] S422, calculating the weight adjustment amount according to the gradient, the calculation formula is: , where the learning rate is a hyperparameter used to control the update step size of the weights in each iteration, the learning rate It is one of the hyperparameters that needs to be manually adjusted. Through experiments and the performance of the validation set, the optimal learning rate can be determined to ensure good results in training;

[0136] S423. For each characteristic related to coal combustion index and operation index i , use the following formula to update the weight, the calculation formula is: , where is the updated weight;

[0137] S424. Repeat the above steps S421-S423 until the model converges or reaches the set number of training times.

[0138] By using deep learning models such as BP neural networks, the present invention can better capture the complex nonlinear relationship between the boiler efficiency of a thermal power plant and the coal burning index and the operating index, and improve the prediction accuracy. By calculating the Spearman correlation coefficient between the coal burning index and the boiler efficiency, and the operating index and the boiler efficiency, the influence of each feature on the boiler efficiency can be understood. According to the result of the correlation coefficient, the weights related to these features in the neural network are adjusted to further optimize the model. By predicting the boiler efficiency, the thermal power plant can adjust the operating parameters more accurately, achieve more efficient energy conversion, improve energy utilization, and reduce energy waste.

[0139] The present invention provides Figure 2 The thermal power plant boiler efficiency prediction system based on neural network shown includes a data preprocessing module, a neural network modeling module, a generalization improvement module and a feature weight adjustment module;

[0140] The data preprocessing module collects historical data during the operation of thermal power plant boilers, cleans the collected data, and selects features related to boiler efficiency;

[0141] The neural network modeling module standardizes the data, builds a BP neural network model through the standardized data, and predicts the efficiency of the thermal power plant boiler through the BP neural network model;

[0142] The generalization improvement module evaluates the performance of the BP neural network model in predicting the efficiency of thermal power plant boilers, and increases the amount of training data based on the evaluation results to enable the BP neural network model to generalize new data or improve the model complexity;

[0143] The feature weight adjustment module calculates the Spearman correlation coefficients between the coal burning index and the boiler efficiency, and between the operation index and the boiler efficiency, and adjusts the weights of the features related to the coal burning index and the operation index in the neural network according to the results of the Spearman correlation coefficients;

[0144] The method for predicting boiler efficiency of a thermal power plant based on a neural network provided in an embodiment of the present invention is realized by the above-mentioned thermal power plant boiler efficiency prediction system based on a neural network. The specific method and process of the thermal power plant boiler efficiency prediction system based on a neural network are detailed in the embodiment of the method for predicting boiler efficiency of a thermal power plant based on a neural network, which will not be repeated here.

[0145] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0146] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0147] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0148] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0149] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0150] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0151] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0152] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for predicting boiler efficiency in thermal power plants based on neural networks, characterized in that: The following steps are involved: S100, collecting historical data during the operation of the thermal power plant boiler, cleaning the collected data, and selecting features related to boiler efficiency; S200, after standardizing the data, constructing a BP neural network model based on the standardized data, and predicting the efficiency of the thermal power plant boiler using the BP neural network model; S300, evaluating the performance of the BP neural network model in predicting the efficiency of thermal power plant boilers, and increasing the amount of training data based on the evaluation results to enable the BP neural network model to generalize new data or to improve the model complexity; In step S300, the information entropy, KL divergence and model complexity of the BP neural network model for thermal power plant boiler efficiency prediction are obtained, and a comprehensive analysis model is established based on the information entropy, KL divergence and model complexity to generate a fitting index. , by fitting the index The performance of BP neural network model in predicting boiler efficiency of thermal power plant was evaluated according to the fitting index. Increasing the amount of training data allows the BP neural network model to generalize new data or increase the complexity of the model, based on the formula: , where 、 、 Represent information entropy, KL divergence and model complexity respectively, 、 、 are the preset proportional coefficients of information entropy, KL divergence and model complexity, respectively, and 、 、 All greater than 0; The calculation expression of information entropy is: ,in, represents the information entropy of random variable X, n represents the number of possible values ​​of the random variable, n is a positive integer, The random variable X takes the value probability; The calculation expression of KL divergence is: ,in, Representing an event i The probability in distribution P, event i The probability of being in the distribution Q, represents the KL divergence of P relative to Q; The calculation formula for model complexity is: , where represents the model complexity, k Indicates the number of parameters of the model; The fitting index generated by the BP neural network model when predicting the efficiency of thermal power plant boilers is compared with the pre-set fitting index reference threshold. If the fitting index is greater than or equal to the fitting index reference threshold, the amount of training data is increased to enable the BP neural network model to generalize the new data and improve the generalization ability of the new data; if the fitting index is less than the fitting index reference threshold, the complexity of the model is increased. S400, calculating the Spearman correlation coefficients between the coal combustion index and the boiler efficiency, and between the operation index and the boiler efficiency, and adjusting the weights of the features related to the coal combustion index and the operation index in the neural network according to the results of the Spearman correlation coefficients; In step S400, the Spearman correlation coefficient between the coal combustion index and the boiler efficiency is calculated by the following steps: S401, obtaining information related to coal combustion index and boiler efficiency from historical data during operation of a thermal power plant boiler; S402, sorting the coal combustion index and boiler efficiency data in ascending order, recording the original serial number of each data, and assigning a rank to each sorted data; S403. Calculate the Spearman correlation coefficient between the coal combustion index and the boiler efficiency. The calculation expression is: , where d represents the rank difference of each pair of coal combustion index and boiler efficiency, ,in, x It is the coal burning index. y is the boiler efficiency, n represents the number of samples; The Spearman correlation coefficient between the operating index and boiler efficiency is calculated by the following steps: S411. Obtaining information related to operating indicators and boiler efficiency from historical data during operation of a thermal power plant boiler; S412. Sort the data of the operating index and boiler efficiency in ascending order, record the original serial number of each data, and assign a rank to each sorted data; S413. Calculate the Spearman correlation coefficient between the operating index and boiler efficiency. The calculation expression is: , where h represents the rank difference of each pair of operating index and boiler efficiency, ,in, k is the operating indicator, v is the boiler efficiency, m represents the number of samples; In the process of adjusting the weights of the features related to the coal combustion index and the operation index in the neural network based on the results of the Spearman correlation coefficient, the gradient descent method is used to update the weights. The specific steps are as follows: S421, for each characteristic related to coal combustion index and operation index i , calculate the gradient of the loss function with respect to the feature weight, and the calculation expression is: , the gradient represents the direction of change of the loss function with respect to the weight, where J is the mean square error loss function, Is and characteristics i The relevant weights, N is the sample size, is the true value, is the predicted value of the model; S422. Calculate the weight adjustment amount based on the gradient. The calculation formula is: , where the learning rate Is a hyperparameter that controls the update step size of the weights in each iteration; S423, for each characteristic related to coal combustion index and operation index i , use the following formula to update the weight, the calculation formula is: , where is the updated weight; S424. Repeat the above steps S421-S423 until the model converges or reaches the set number of training times.

2. The method for predicting boiler efficiency in thermal power plants based on neural networks according to claim 1, characterized in that: In step S200, the specific steps of data standardization are as follows: a. Calculate the minimum and maximum values ​​of each feature using the Min-Max normalization method; b. Apply the normalization formula to each feature and calculate the normalized value. The formula is: , where represents the normalized value, is the original value, is the minimum value of the data, The maximum value of the data.

3. The method for predicting boiler efficiency of a thermal power plant based on a neural network according to claim 1, characterized in that: In step S200, the steps of constructing the BP neural network model are as follows: S201, determine the number of layers of the neural network and the number of nodes in each layer, and select an activation function for each layer; S202, initializing the weights of the neural network; S203, mean square error is used as the loss function, and gradient descent is used as the optimization algorithm.

4. A thermal power plant boiler efficiency prediction system based on a neural network, used to implement the thermal power plant boiler efficiency prediction method based on a neural network as described in any one of claims 1 to 3, characterized in that: It includes data preprocessing module, neural network modeling module, generalization improvement module and feature weight adjustment module; The data preprocessing module collects historical data during the operation of thermal power plant boilers, cleans the collected data, and selects features related to boiler efficiency; The neural network modeling module normalizes the data and then constructs a BP neural network model based on the normalized data. The BP neural network model is used to predict the efficiency of thermal power plant boilers. The generalization improvement module evaluates the performance of the BP neural network model in predicting the efficiency of thermal power plant boilers. Based on the evaluation results, the amount of training data is increased to enable the BP neural network model to generalize to new data or improve the model complexity. The feature weight adjustment module calculates the Spearman correlation coefficients between the coal burning index and the boiler efficiency, and between the operation index and the boiler efficiency, and adjusts the weights of the features related to the coal burning index and the operation index in the neural network according to the results of the Spearman correlation coefficients.

Citation Information

Patent Citations

  • Multi-target boiler combustion optimization control method based on embedded algorithm fusion under variable loads

    CN115111601A

  • Gasoline octane number loss prediction and optimization method based on BP neural network and conditional generative adversarial network

    CN115115102A