Boiler thermal efficiency soft measurement method and system based on LTC and electronic equipment

Abstract

The invention provides a boiler thermal efficiency soft measurement method and system based on LTC and electronic equipment, and belongs to the technical field of boiler thermal efficiency soft measurement. Constructing a boiler thermal efficiency online soft measurement model of a four-layer network based on a liquid state time constant network (LTC); training the model by using the training set data, and determining an optimal network structure and hyper-parameters through verification of the verification set; inputting the test set into the optimized model to output a predicted value, and evaluating the performance after reverse normalization; and inputting real-time operation data into the trained model, and outputting a thermal efficiency prediction value. According to the method, the characteristic that the weight is adjusted in real time through the dynamic differential equation by means of the LTC model is utilized, the defect that parameters of a traditional neural network model are fixed is overcome, the method adapts to the time-varying characteristic of a measured object, high-precision dynamic soft measurement of the boiler thermal efficiency is achieved, and the measurement precision and the response real-time performance are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of soft measurement of boiler thermal efficiency, and particularly relates to a soft measurement method, system and electronic device for boiler thermal efficiency based on LTC. Background Technique

[0002] During the long-term operation of coal-fired boilers, due to the variable quality of the coal fed into the furnace (changes in parameters such as ash content, sulfur content, volatile matter, etc.) and the gradual and slow deterioration of the heating surface, the energy conversion performance shows a progressive decline. High-ash flue gas causes fouling on the radiation / convection heating surface, and the increase in heat transfer resistance leads to an increase in the flue gas temperature; the coupled deposition of the fouling layer and oxide scale reduces the heat transfer coefficient. The above factors together cause the boiler thermal efficiency to slowly decrease with the increase in operation time. At the initial stage of the design of coal-fired power units as stable power sources for the power system, with the rapid development of new energy, coal-fired power units have to participate in the deep peak regulation of the power grid and transform from basic power sources to regulating power sources. Under the background of deep peak regulation, coal-fired power units are forced to operate at low load conditions deviating from the design conditions for a long time, thereby causing the boiler thermal efficiency to have strong time-varying characteristics and a significant decline. Therefore, establishing an accurate boiler thermal efficiency prediction model is the key to improving the boiler thermal efficiency under complex conditions such as deep peak regulation and transient load change.

[0003] Regarding the research on the modeling of the boiler combustion process, it is currently mainly divided into two major technical branches: mechanism modeling methods and data-driven modeling methods. The mechanism modeling method is based on basic theories such as the kinetic model of combustion and related physical and chemical processes and the law of conservation of energy to construct a mathematical model reflecting the internal mechanism of the process. In existing research, some researchers have established a mathematical model describing the boiler thermal efficiency and NO x emission characteristics based on the combustion mechanism of the coal distribution characteristics in the furnace and over-fire air (OFA); a discrete-time linear state space model has been constructed through reasonable simplification of the combustion system. Constructing a mechanism model of the boiler combustion process with the help of computational fluid dynamics (CFD) technology is also a commonly used technical means, such as establishing a CFD model coupling the air and flue gas side and the steam and water side of the boiler.

[0004] However, when using a 3 GHz Dell 3600 workstation for 4-thread parallel computing, the solution time for a single operating condition is about one week, which is difficult to meet the engineering requirements of real-time operation control of coal-fired power plants. In the field of boiler thermal efficiency prediction, factors such as coking, wear, and component replacement of the heating surface will affect the characteristics of the thermal process. As the operating time increases, the thermal efficiency of the boiler shows a slow downward trend, resulting in a gradual decrease in the performance matching degree between the mechanism model and the actual operating conditions of the boiler. In addition, the combustion process of coal-fired boilers involves multidisciplinary complex physical processes such as aerodynamics, two-phase flow, heat and mass transfer. The online prediction method of thermal efficiency based on the mechanism model is difficult to meet the requirements of engineering applications in terms of prediction accuracy and real-time performance. In summary, the engineering application scope of the mechanism modeling method has significant limitations.

[0005] Compared with the mechanism modeling method, the data-driven modeling method uses machine learning algorithms to directly mine the mapping relationship between system parameters from the operating data, and has the technical advantages of fast modeling speed and high prediction accuracy. For industrial processes whose mechanisms are not yet fully understood, machine learning algorithms represented by neural networks are more suitable for data-driven modeling methods.

[0006] Existing modeling methods do not consider the time scale factor in the design of input dimensions, and all belong to the category of static models, which are only applicable to steady-state boiler operating conditions. Specifically, when the boiler is in the cold start process or the load switching stage during the deep peak regulation process, the historical operating conditions will have a significant impact on the current thermal efficiency. This is because during the change of boiler load, the heat storage of the working fluid will be released or rebalanced, and under stable operating conditions, the heat storage effect of the working fluid can usually effectively improve the combustion efficiency and thermal efficiency and accelerate the load response of the unit. Therefore, in view of the time series dynamic characteristics of the boiler thermal efficiency, it is necessary to construct a dynamic prediction model and incorporate the historical operating condition parameters into the model input system. Summary of the invention

[0007] To solve the above technical problems, the present invention provides a soft measurement method, system and electronic device for boiler thermal efficiency based on LTC. The Liquid Time-constant Networks (LTC) is inspired by the nervous system of Caenorhabditis elegans (C. elegans). The present invention constructs a soft measurement model for boiler thermal efficiency based on LTC, which adopts the architecture of Neural Circuit Policies (NCP) and is composed of four layers of neurons, namely sensory neurons, inter-neurons, command neurons and motor neurons. Each neuron in each layer is constructed using the LTC model. The LTC model adjusts the weights in real time using dynamic differential equations, which can effectively overcome the deficiency that the parameters of the previous soft measurement model remain fixed after training, realize high-precision dynamic soft measurement of boiler thermal efficiency, and improve the accuracy and real-time performance of boiler thermal efficiency measurement.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] A soft measurement method for boiler thermal efficiency based on LTC, comprising the following steps:

[0010] Step 1, collect the historical operation data of the unit's distributed control system;

[0011] Step 2, preprocess the historical operation data obtained in Step 1, including: missing data processing, data dimensionality reduction, data reconstruction, so as to form a data set, and divide the data set into a training set, a validation set and a test set;

[0012] Step 3, construct a soft measurement model for boiler thermal efficiency based on LTC, including four layers of neuron layers connected in sequence, namely: a sensory neuron layer as the input layer, an inter-neuron layer, a command neuron layer and a motor neuron layer as the output layer; LTC represents the Liquid Time-constant Networks;

[0013] Step 4, use the training set to train the soft measurement model for boiler thermal efficiency based on LTC constructed in Step 3, optimize the model hyperparameters, and use the validation set to determine the optimal hyperparameter settings to obtain an optimized soft measurement model for boiler thermal efficiency based on LTC;

[0014] Step 5, input the data of the test set into the optimized soft measurement model for boiler thermal efficiency based on LTC obtained in Step 4, obtain the prediction results, after anti-normalization processing, use the evaluation function to analyze and evaluate the prediction performance of the soft measurement model for boiler thermal efficiency based on LTC under different working conditions;

[0015] Step 6: Input the real-time collected boiler operation data into the optimized LTC-based soft sensor model of boiler thermal efficiency obtained in Step 4, and output the predicted value of boiler thermal efficiency.

[0016] Further, in the above Step 1, the historical operation data includes boiler thermal efficiency and auxiliary variables; wherein, the boiler thermal efficiency is the percentage of the effectively utilized heat of the boiler to the heat input of the fuel; the auxiliary variables include fuel parameters, operation condition parameters, flue gas parameters, and environmental parameters.

[0017] Further, in the above Step 2, the missing data processing includes: detecting abnormal data by using the 3σ-rule, and replacing the abnormal data based on the change rate of the auxiliary variables at adjacent moments.

[0018] Further, in the above Step 2, the data dimensionality reduction includes: sorting the importance of the auxiliary variables by using the random forest algorithm, and screening the top 10 auxiliary variables with the highest importance as the input variables.

[0019] Further, in the above Step 2, the data reconstruction includes: converting the time series data into a two-dimensional tensor by using the sliding window method, where the time dimension contains 20 consecutive sampling moments, and the feature dimension is the top 10 auxiliary variables with the highest importance screened.

[0020] Further, in the above Step 3, each neuron in the four neuron layers of the LTC-based soft sensor model of boiler thermal efficiency is constructed by using an LTC model.

[0021] The present invention also provides a soft sensor system for boiler thermal efficiency for implementing the above LTC-based soft measurement method for boiler thermal efficiency, including:

[0022] A data acquisition module, configured to obtain the historical operation data of the unit distributed control system;

[0023] A preprocessing module, configured to perform missing data processing, data dimensionality reduction, and data reconstruction on the historical operation data;

[0024] A model operation module, configured to deploy the LTC-based soft sensor model of boiler thermal efficiency and perform prediction calculations to obtain the predicted result of boiler thermal efficiency;

[0025] A visualization terminal, configured to online display the predicted result of boiler thermal efficiency.

[0026] The present invention also provides an electronic device, including a processor and a memory, where the memory stores a computer program, and the processor implements the steps of the above LTC-based soft measurement method for boiler thermal efficiency when executing the program.

[0027] Furthermore, the electronic device is integrated into the unit distributed control system or the edge computing terminal, receives the boiler operation data in real time, and outputs the prediction result of the boiler thermal efficiency.

[0028] Beneficial effects:

[0029] In the soft measurement model of boiler thermal efficiency based on LTC constructed by the present invention, each neuron in each layer is constructed by using the LTC model. The LTC model adjusts the weights in real time by using the dynamic differential equation, which can effectively overcome the deficiency that the parameters of the previous soft measurement model remain unchanged after training, realize the high-precision dynamic soft measurement of the boiler thermal efficiency, and improve the accuracy and real-time performance of the measurement. In addition, the present invention designs a dynamic soft measurement method suitable for the operation data of different working conditions of the boiler, which can adapt to the multi-variable and non-linear characteristics under complex working conditions, and has strong robustness, interpretability and generalization ability and high calculation efficiency. Specifically, the present invention uses the random forest algorithm to preprocess the boiler operation data, especially for dimensionality reduction; subsequently, the dimensionality-reduced eigenvalue is input into the soft measurement model of boiler thermal efficiency based on LTC for rapid modeling and prediction, so as to realize the real-time and dynamic measurement of the boiler thermal efficiency; finally, the soft measurement model of boiler thermal efficiency based on LTC is evaluated by using the real data under the variable load condition of a certain coal-fired boiler. The electronic device provided by the present invention can realize the automatic operation of the above method, and further improve the soft measurement efficiency and practicability. Description of the drawings

[0030] Figure 1 It is the flowchart of the soft measurement method of boiler thermal efficiency based on LTC of the present invention;

[0031] Figure 2 It is the schematic diagram of a single burner in the embodiment: (a) is the schematic diagram of the rich fuel nozzle and the lean gas nozzle; (b) is the layout schematic diagram of the secondary air nozzles and air ducts at different levels in the boiler furnace;

[0032] Figure 3 It is the schematic diagram of data reconstruction in the embodiment: (a) time series variable, (b) tensor diagram obtained through the sliding window;

[0033] Figure 4 It is the schematic diagram of dataset segmentation in the embodiment;

[0034] Figure 5 It is the schematic diagram of the principle of the closed-loop continuous time (CFC) neuron;

[0035] Figure 6 It is the structural diagram of the soft measurement model of boiler thermal efficiency based on LTC in the embodiment;

[0036] Figure 7Iterative loss value diagrams of the boiler thermal efficiency soft measurement model for different activation functions in the embodiments; among them, (a) is the tanh hyperbolic tangent activation function, (b) is the Sigmoid activation function, and (c) is the rectified linear unit ReLU activation function;

[0037] Figure 8 Prediction result diagrams of LSTM, GRU, and LTC on the test set in the embodiments; among them, (a) is the load reduction interval, (b) is the load increase interval, and (c) is the steady state interval;

[0038] Figure 9 Correlation diagram between unit load and boiler thermal efficiency in the embodiments. Detailed implementation manners

[0039] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the following various embodiments can be combined with each other without conflict.

[0040] As Figure 1 shown, the present invention provides a dynamic soft measurement method for boiler thermal efficiency based on LTC, including the following steps:

[0041] Step 1, collect historical operation data;

[0042] Step 2, preprocess the historical operation data, including: missing data processing, data dimensionality reduction, data reconstruction, and divide the preprocessed data into a training set, a validation set, and a test set;

[0043] Step 3, construct a soft measurement model for boiler thermal efficiency based on LTC, and the soft measurement model for boiler thermal efficiency based on LTC includes four layers of neuron layers connected in sequence: a sensory neuron layer (input layer), an intermediate neuron layer, a command neuron layer, and a motor neuron layer (output layer); that is Figure 1 the four-layer network recorded in step 3: the sensory, intermediate, command, and motor neurons are connected in sequence;

[0044] Step 4, train the soft measurement model for boiler thermal efficiency based on LTC with the training set, and use the validation set to determine the optimal hyperparameter settings to obtain an optimized soft measurement model for boiler thermal efficiency based on LTC;

[0045] Step 5, input the test set into the soft measurement model for boiler thermal efficiency based on LTC, obtain the prediction result, and perform inverse normalization and evaluation analysis;

[0046] Step 6, input the real-time collected data into the soft measurement model for boiler thermal efficiency based on LTC, and output the dynamic thermal efficiency prediction value.

[0047] Further, in the step 1, the historical operation data includes the thermal efficiency and its related auxiliary variables (as shown in Table 2 below).

[0048] Further, in the step 2, the steps for missing data processing are as follows: detecting abnormal data by using the 3σ-rule, and replacing the abnormal data based on the following formula:

[0049] ;

[0050] where i represents the auxiliary variable index value (the i-th variable), t represents the current time point, k is the time window length, is the index of historical sampling moments within the window, with the range of [2, t] or [t - k, t - 1], represents the value of the i-th auxiliary variable at time point t, represents the value of the i-th auxiliary variable at time point t - 1, represents the auxiliary variable the change rate between adjacent time points t and t - 1, represents the average change rate before time point t, represents the standard deviation.

[0051] Further, in the step 2, the steps for data dimensionality reduction are as follows: using the random forest algorithm (Random forest, RF) to rank the importance of auxiliary variables, and screening the top 10 variables with the highest importance as input variables.

[0052] Further, in the step 2, the steps for data reconstruction are as follows: using the sliding window method to convert the time series data into a two-dimensional tensor, where the time dimension contains 20 consecutive sampling moments, and the feature dimension is the 10 screened variables.

[0053] Further, in the step 3, the soft sensor model of the boiler thermal efficiency based on LTC includes four neuron layers, and each neuron in each layer is constructed by using the LTC model.

[0054] Further, in the step 4, the hyperparameters of the soft sensor model of the boiler thermal efficiency based on LTC are set as follows: the learning rate is set to 0.005, the activation function uses ReLU (Rectified Linear Unit), the number of neurons in each layer is 128, the number of iterations is 200, and the optimizer uses Adam.

[0055] The present invention will be further explained below in conjunction with the accompanying drawings and embodiments.

[0056] Embodiment

[0057] A soft measurement method for boiler thermal efficiency based on LTC in this embodiment includes the following steps:

[0058] Step 1: Collect historical operation data from the unit distributed control system (DCS).

[0059] The research object of this embodiment is a 600MW supercritical thermal power unit W-flame boiler. The boiler is a supercritical parameter, W-type flame combustion, variable pressure operation once-through boiler. The main design parameters of the boiler are shown in Table 1. The power plant is located in the hinterland of southwest China, with a humid climate and burning poor quality anthracite with a volatile content of less than 12%, resulting in a large amount of coking in the furnace. The coking phenomenon seriously affects the combustion atmosphere and heat transfer in the furnace, making it more difficult to predict the boiler thermal efficiency through the mechanism model. Therefore, the data-driven modeling method is more suitable for this kind of industrial scenario.

[0060] Table 1 Main design parameters of the boiler

[0061]

[0062] The target boiler is equipped with 6 double-inlet and double-outlet ball mills and 24 double-whirlwind once-through pulverized coal burners specialized for burning low-volatile coal. As Figure 2 shown in the schematic diagram of a single burner, Figure 2 (a) of it shows the structural appearance of the rich fuel nozzle and the lean gas nozzle in the burner; Figure 2 (b) of it presents the position distribution of the secondary air dampers at different levels (A-layer, B-layer, C-layer) and the secondary air dampers (D-layer, F-layer) and OFA (over-fire air) on the burner. Figure 2 In it, OFA (Over - Fire Air) refers to over-fire air, which is injected in the later stage of combustion to provide additional air for the burnout stage of the fuel, promoting the further reaction of the incompletely burned substances, improving the combustion efficiency, and reducing the unburned loss. The rich fuel nozzle is specially used to inject a fuel-rich mixture to provide the main fuel source for combustion; the lean gas nozzle is used to inject the lean gas carrying a certain amount of energy and substances generated during the coal pulverization process, which can participate in the combustion reaction and also contribute to the air flow organization. The secondary air of each burner is controlled separately, and the air distribution unit consists of an upper air box and a lower air box. The F baffle has the largest air volume and controls the main secondary air volume required for combustion. The opening degrees of the F-layer secondary air dampers of the 24 burners are included in the auxiliary variables required for modeling.

[0063] Coal-fired power generation boilers mainly use the inverse balance method to determine the thermal efficiency, that is, by experimentally measuring the various thermal losses of the boiler, and then calculating the boiler thermal efficiency according to the following formula :

[0064] ;

[0065] In the formula, represents the percentage of each heat loss in the input heat, , m = 2, 3, …, 6, is the input heat, where is the percentage of heat loss due to flue gas in the input heat (%); is the percentage of heat loss due to incomplete combustion of gas in the input heat (%); is the percentage of heat loss due to incomplete combustion of solid in the input heat (%); is the percentage of heat loss due to boiler heat dissipation in the input heat (%); is the percentage of other heat losses in the input heat (%). is the heat loss of the boiler, m = 2, 3, …, 6, where is the heat loss due to flue gas (kJ / kg); is the heat loss due to incomplete combustion of gas (kJ / kg); is the heat loss due to incomplete combustion of solid (kJ / kg); is the heat loss due to boiler heat dissipation (kJ / kg); is the heat loss of other heat losses (kJ / kg). By calculating each heat loss, the operating conditions of the boiler can be grasped, and optimization measures to improve the boiler thermal efficiency can be found. Analyze the variables affecting each heat loss and then form a list of variables affecting the thermal efficiency. Heat loss due to flue gas is the largest item in the boiler heat loss, generally accounting for 5% - 6%. For every 15 - 20 °C increase in the flue gas temperature, will increase by about 1%. Therefore, the flue gas temperature is selected as an input variable for predicting the boiler thermal efficiency. For the heat loss due to incomplete combustion of gas , in a coal-fired boiler, only CO is considered as the product of incomplete combustion of gas. Since the target boiler is not equipped with a CO sensor, the CO content is not considered as an auxiliary variable. Heat loss due to incomplete combustion of solid is second only to the heat loss due to flue gas. The main factors affecting this part of the heat loss include coal quality, combustion mode, oxygen content in flue gas, coal blending, air distribution mode, etc. Therefore, auxiliary variables such as oxygen content in flue gas, mill capacity air flow (coal blending), and secondary air damper opening (air distribution) are considered.

[0066] Based on the combustion mechanism, expert experience, and the suggestions of operating engineers, the present invention integrates the above variables to form an auxiliary variable table. The auxiliary variable table incorporates a total of 53 parameters related to variables of air, coal, and water. The detailed information of all auxiliary variables is shown in Table 2. The selection of auxiliary variables comprehensively considers influencing factors such as the air flow field in the furnace, the two-phase flow field of air and coal, and the heat absorption of the working medium. Among them, the secondary air controls the main air required for combustion in the furnace, the volumetric air flow at the inlet of the coal mill controls the distribution of the two-phase flow field of air and coal in the furnace, and variables such as the main steam pressure and the main steam temperature reflect the heat absorption level of the working medium in the furnace.

[0067] Table 2 Auxiliary Variable Table

[0068]

[0069] Step 2: Preprocess the collected historical operation data.

[0070] Data for approximately 10 consecutive days of operation were extracted from the DCS of the target boiler, totaling 14,419 samples. The original data was preprocessed to meet the requirements of model input. The data preprocessing process includes: missing data processing, data dimensionality reduction, data reconstruction, and the dataset was divided into a training set, a validation set, and a test set.

[0071] Step 2.1 Missing data processing: Due to the complex environment at the boiler production site and signal doping with noise, the original data stored in the DCS has outliers to varying degrees. In this embodiment, the 3σ - rule is used to detect abnormal data, and the formula is as shown above.

[0072] Step 2.2 Data dimensionality reduction: With the in - depth study of boiler combustion process modeling, researchers have found that the dimensionality reduction of input variables not only helps to improve the learning efficiency of the model but also can enhance the generalization performance. Yang et al. used principal component analysis to reduce 35 original variables of the boiler to 6 principal components for predicting NOx emissions; Tang et al. screened out 13 variables highly correlated with NOx emissions based on mutual information. The random forest (RF) algorithm was used to rank the importance of auxiliary variables, and the top 10 variables in terms of importance were selected as input variables.

[0073] Random forest is an extended variant of Bagging, which is particularly suitable for evaluating the importance among high - dimensional non - linear variables. RF uses several independent decision trees for parallel prediction. Its random sampling with replacement enables the out - of - bag data (OOB) to naturally have the function of evaluating the prediction error of the model, and the degree of deterioration of this error indirectly reflects the importance of variables. For each decision tree, the in - bag data is used to train the tree, and the out - of - bag data (OOB) is used to evaluate the tree. The deterioration of the prediction accuracy is achieved by imposing noise interference on important variables. Before and after imposing the interference, the importance of input variable j The measurement is calculated by the following formula:

[0074] ;

[0075] where, quantifies the influence degree of the input variable j on the prediction accuracy of the model. The larger the value, the more important the variable is to the model. and are the prediction errors of the out-of-bag data before and after interference on the decision tree b respectively. is the number of decision trees, and b represents the serial number of the decision tree. The sum of the importance of all input variables is 1, that is, , where j = 1, 2, …, n represents the serial number of the auxiliary variable, and n represents the total number of auxiliary variables.

[0076] In this embodiment, aiming at the industrial attributes of high-dimensional nonlinearity and large inertia of the boiler combustion system, the random forest algorithm is adopted for dimensionality reduction.

[0077] The RF algorithm is used to evaluate the feature importance of 53 comprehensive auxiliary variables affecting the thermal efficiency, and the number of decision trees is set to 500. For the variables ranked in the top 6 in terms of importance, the sum of their importance has exceeded 0.9. Combining with the knowledge of the boiler combustion mechanism in this embodiment, the F-layer secondary air damper that mainly affects the in-furnace combustion is also included in the input variables, and the load has a significant impact on the overall unit. Finally, 10 variables (x1~x 10 ) are selected. The total weight of the variables ranked in the top 10 in terms of importance reaches 0.933, and the re-ordering is shown in Table 3.

[0078] Table 3 Importance of variables related to thermal efficiency (mean ± standard value)

[0079]

[0080] According to the evaluation results of the importance of each variable, analyze its necessity in the thermal efficiency modeling. As can be seen from Table 3, in the ranking of the variable importance calculated by RF, the importance of the flue gas temperature is the highest, and the importance of a single variable occupies an absolute advantage, which is consistent with the fact that the flue gas heat loss is the largest heat loss of the boiler. In addition, the excess air coefficient at the furnace outlet of the coal-fired boiler has a great influence on the thermal efficiency. The excess air coefficient at the furnace outlet is generally characterized by the oxygen content in the flue gas, which refers to the ratio of the actual air volume required for coal supply to the theoretical air volume. The larger the excess air coefficient, the more the actual air volume exceeds the theoretical required air volume, that is, the air supply is excessive during the boiler combustion process. The influence trend of the excess air coefficient on the thermal efficiency of the coal-fired boiler is relatively complex. When the value of the excess air coefficient is relatively small and cannot ensure the air volume demand for fuel combustion, it will obviously cause an increase in the mechanical incomplete combustion heat loss. At this time, increasing the excess air coefficient will reduce the heat loss ; However, if the excess air coefficient can already meet the combustion requirements in the furnace, increasing the excess air coefficient at this time will increase the flue gas flow rate, and the heat loss due to flue gas discharge will also increase. At the same time, the residence time of the fuel in the furnace will be shortened, and the carrying capacity of the flue gas for large particle fuels will be enhanced, all of which will increase the heat loss due to incomplete combustion of solids increase. Therefore, there is an optimal excess air coefficient that can not only ensure the air volume required for fuel combustion but also ensure the combustion time. At this time, the heat loss due to incomplete combustion of solids will reach the minimum value, and the boiler thermal efficiency will also remain at a high level. Therefore, in actual operation, it is necessary to find a suitable balance point by adjusting the excess air coefficient to achieve the best thermal efficiency of the coal-fired boiler. Through the above analysis, the variable importance calculated by RF conforms to the variation law of the boiler thermal efficiency. In this embodiment, based on this importance ranking and boiler combustion knowledge, the top 10 variables in the importance ranking are selected as the input variables of the model.

[0081] Step 2.3 Data reconstruction:

[0082] To construct an input structure suitable for time series modeling, the present invention uses the sliding window method to convert the data into a two-dimensional tensor (time dimension x feature dimension). The data is segmented according to the feature dimension and the time dimension. Among them, the time dimension contains 20 sampling moments, that is, H = 20; the feature dimension is 10 variable types, that is, W = 10; the window slides to the right along the time dimension, and the window sliding step is 1. The label corresponding to each two-dimensional tensor is the boiler thermal efficiency at the last time dimension of the tensor, Figure 3 The schematic diagram of data reconstruction is shown, where, Figure 3 in (a) represents the time series variable, where variable - 1 represents the load in Table 3, variable - 2 represents Mill - E - left in Table 3, and so on. In addition, the output - y represents the predicted target variable, corresponding to the thermal efficiency, and label - 1 represents the first label; Figure 3 in (b) represents the tensor diagram obtained through the sliding window. In the tensor diagram, x represents the variable, the superscript of x represents the variable sampling moment, and the subscript of x represents the variable types 1, 2,..., W; y H represents the thermal efficiency at the Hth sampling moment.

[0083] Step 2.4 Dataset segmentation:

[0084] As Figure 4The figure shows a schematic diagram of dataset segmentation. The feature dimension consists of 10 variables (Feature #1 - Feature #10) that are highly correlated with the thermal efficiency. t represents the sampling time. The data of the original 14,419 sampling times are segmented into 14,400 standard input matrices. The label represents the predicted target variable y, that is, the thermal efficiency. The original sampling serial numbers of the labels corresponding to each matrix are 20 - 14,419. In the present invention, they are renumbered as 1 - 14,400 to be consistent with the matrix serial numbers, that is, the serial numbers of the input matrices and the labels are all counted from 1. All the segmented data are divided into a training set, a validation set, and a test set according to a ratio of 6:2:2. Among them, the training set includes 8,640 sampling points, the validation set includes 2,880 sampling points, and the test set includes 2,880 sampling points.

[0085] Step 3: Construct a soft sensor model for boiler thermal efficiency based on LTC. Its network structure consists of four neuron layers connected in sequence, namely the sensory neuron layer (input layer), the intermediate neuron layer, the command neuron layer, and the motor neuron layer (output layer). There are feedforward connections between these connection layers, and there are recurrent connections between the command neurons. Among them, each neuron in each layer adopts the liquid time constant (LTC) model.

[0086] The liquid time constant (LTC) model is based on ordinary differential equations (ODEs). Its time behavior adaptively adjusts the weights in real time according to the input, making it a general approximator for modeling system dynamics. The LTC model and its closed-form approximation CFC model are inspired by the hierarchical information processing mechanism of the biological nervous system (nematode) in terms of their structural design, and include four-layer topologies of sensory neurons, intermediate neurons, command neurons, and motor neurons. The LTC model is represented by the following ordinary differential equations:

[0087] ;

[0088] ;

[0089] In the formula: represents the nonlinear synapse, ; represents the nonlinear synaptic release. A and τ represent the synaptic reversal potential and the time constant respectively. θ is a set of parameters used to adjust the behavior of the nonlinear synaptic release function. Its specific values and meanings depend on the function form, and t is the time.

[0090] Solving the ordinary differential equations of LTC usually involves a large amount of computational effort with iterative methods. A closed-form expression is introduced to approximately solve it with lower complexity. The closed-form continuous time (CFC) model significantly reduces the computational complexity through approximate solution. Figure 5Shows the structural schematic diagram of the CFC model, which realizes the closed-form approximation solution of the ordinary differential equation of the LTC model through the combination of specific neural network branches (f, g, h), activation functions (such as Sigmoid), and operations (such as Hadamard product, addition). Figure 6 It is the structural diagram of the soft sensor model for the boiler thermal efficiency based on LTC. The input data passes through the sensory neurons (input), the intermediate neurons, the command neurons, and the motor neurons (output), and finally outputs.

[0091] Compared with the LTC model that requires a numerical solver, CFC significantly improves the training and inference efficiency. This layer efficiently extracts time-related dynamic features, and its core operation can be expressed as:

[0092] x ( t ) = σ ( − f ( x ( t ), i ( t ); θ f ) t ) ⊙ g( x ( t ), i ( t ); θ g ) + [ 1- σ (- f ( x ( t ), i ( t ); θ f ) t ) ] ⊙ h( x ( t ), i ( t ); θ h ) ;

[0093] In the formula, f, g, and h are trainable neural layers (with parameters θ f , θ g , and θ h ), represents the input, σ represents the sigmoid function, ⊙ represents the Hadamard product (element-wise product), is the hidden state, and t represents time.

[0094] Step 4: Use the training set to train the soft sensor model for the boiler thermal efficiency based on LTC, optimize the network hyperparameters, and use the validation set to determine the optimal hyperparameter settings to obtain the optimized soft sensor model for the boiler thermal efficiency based on LTC.

[0095] The LTC model establishes a four-layer network structure according to the NCP architecture. The hyperparameters of the LTC model include: the learning rate is set to 0.005, the activation function is ReLU, the number of neurons in each layer is 128, the number of iterations is 200, and the optimizer uses Adam.

[0096] Figure 7 Compares the training loss values under different activation functions (tanh, Sigmoid, ReLU). The results show that the loss value is the lowest when using the ReLU activation function. Among them, Figure 7 (a) of is the tanh hyperbolic tangent activation function, Figure 7 (b) of is the Sigmoid activation function, Figure 7 (c) of is the rectified linear unit ReLU activation function.

[0097] Step 5: Input the data of the test set into the optimized soft sensor model for the boiler thermal efficiency, obtain the prediction results, and use the evaluation function to analyze and evaluate the prediction performance of the model under different working conditions after anti-normalization processing;

[0098] To evaluate the model performance, the LTC model was compared with LSTM and GRU. The mean and standard deviation (Std) of the model prediction errors were from twenty independent experiments.

[0099] The LTC network was compared with LSTM and GRU using the test set. Both the LSTM and GRU models adopted a single hidden layer structure, and the number of neurons in this hidden layer was set to 128 LSTM neuron units and GRU neuron units. The output of the last neuron unit of LSTM and GRU was used as the overall prediction result of the model.

[0100] The mean and standard deviation of the prediction errors of LSTM, GRU, and the LTC network are shown in Table 4.

[0101] Table 4 Prediction performance of LSTM, GRU, and LTC on the test set

[0102]

[0103] The RMSE of the LTC network was 0.091 ± 0.005%, the MAE was 0.051 ± 0.005%, and R 2 was 0.986 ± 0.002. The root mean square error RMSE, mean absolute error MAE, and coefficient of determination R 2 of the LTC network were all superior to those of the LSTM and GRU networks.

[0104] Randomly select one experiment from the twenty experiments, Figure 8 The boiler thermal efficiency prediction curve and load curve of a randomly selected experiment are shown, where Figure 8 in (a), Figure 8 in (b), Figure 8 in (c) respectively correspond to the load reduction, load increase, and steady-state conditions. In the load reduction interval, the deviation between the predicted value and the true value of the boiler thermal efficiency of the LTC network was the smallest. In the load increase interval, the prediction deviation of the LTC model was the smallest. In the steady-state interval, the prediction performances of the three models were close. Figure 9 The Pearson's coefficient r showing the load and thermal efficiency was 0.826, confirming a strong positive correlation between the two. In summary, the boiler thermal efficiency prediction model based on LTC showed low prediction errors both in transient (load increase and decrease) and steady-state conditions.

[0105] Step 6: Input the real-time collected boiler operation data into the trained soft sensor model of the boiler thermal efficiency, and output the predicted value of the boiler thermal efficiency.

[0106] The present invention also provides a system for implementing the dynamic soft sensing method of the boiler thermal efficiency based on LTC, including:

[0107] A data acquisition module that acquires historical operation data;

[0108] A preprocessing module for performing missing data processing, data dimensionality reduction, and data reconstruction on the historical operation data;

[0109] A model operation module for deploying a soft measurement model of boiler thermal efficiency based on LTC and performing prediction calculations to obtain a prediction result of boiler thermal efficiency;

[0110] A visualization terminal for online display of the prediction result of boiler thermal efficiency.

[0111] The present invention also provides an electronic device, including a processor and a memory. The memory stores a computer program, and when the processor executes the program, the steps of the above-mentioned soft measurement method of boiler thermal efficiency based on LTC are implemented.

[0112] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned soft measurement method of boiler thermal efficiency based on LTC are implemented.

[0113] Further, the electronic device is integrated into a coal-fired power plant DCS system or an edge computing terminal, and receives boiler operation data in real time and outputs a prediction result of boiler thermal efficiency.

[0114] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented in various computer languages, for example, object-oriented programming languages such as Java and interpreted scripting languages such as JavaScript.

[0115] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate for implementing in the process Figure 1 One process or multiple processes and / or blocksFigure 1 means for the functions specified in one or more boxes.

[0116] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction means that implements the functions specified in one Figure 1 one or more processes and / or boxes Figure 1 or more boxes.

[0117] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one Figure 1 one or more processes and / or boxes Figure 1 or more boxes.

[0118] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn the basic creative concept. Therefore, the appended claims are intended to be construed to cover the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0119] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A soft measurement method for the boiler thermal efficiency based on LTC, characterized in that It includes the following steps: Step 1: Collect the historical operation data of the unit's distributed control system; Step 2: Preprocess the historical operation data obtained in Step 1, including: missing data processing, data dimensionality reduction, and data reconstruction, so as to form a data set, and divide the data set into a training set, a validation set, and a test set; Step 3: Construct a soft sensor model for boiler thermal efficiency based on LTC, including four neuron layers connected in sequence, namely: a sensory neuron layer as the input layer, an intermediate neuron layer, a command neuron layer, and a motor neuron layer as the output layer; LTC represents the Liquid Time Constant Network; Step 4: Use the training set to train the soft sensor model for boiler thermal efficiency based on LTC constructed in Step 3, optimize the model hyperparameters, and use the validation set to determine the optimal hyperparameter settings to obtain an optimized soft sensor model for boiler thermal efficiency based on LTC; Step 5: Input the data of the test set into the optimized soft sensor model for boiler thermal efficiency based on LTC obtained in Step 4 to obtain the prediction results. After inverse normalization processing, use the evaluation function to analyze and evaluate the prediction performance of the soft sensor model for boiler thermal efficiency based on LTC under different working conditions; Step 6: Input the real-time collected boiler operation data into the optimized soft sensor model for boiler thermal efficiency based on LTC obtained in Step 4, and output the predicted value of the boiler thermal efficiency.

2. The soft measurement method of boiler thermal efficiency based on LTC according to claim 1, characterized in that, In the said Step 1, the historical operation data includes boiler thermal efficiency and auxiliary variables; among them, the boiler thermal efficiency is the percentage of the effectively utilized heat of the boiler to the heat input by the fuel; the auxiliary variables include fuel parameters, operation condition parameters, flue gas parameters, and environmental parameters.

3. The soft measurement method for boiler thermal efficiency based on LTC according to claim 1, characterized in that, In the said Step 2, the missing data processing includes: using the 3σ-rule to detect abnormal data, and replacing the abnormal data based on the change rate of the auxiliary variables at adjacent moments.

4. The soft measurement method of boiler thermal efficiency based on LTC according to claim 1, characterized in that, In the said Step 2, the data dimensionality reduction includes: using the random forest algorithm to rank the importance of the auxiliary variables, and screening the top 10 auxiliary variables with importance as the input variables.

5. The soft measurement method of boiler thermal efficiency based on LTC according to claim 4, characterized in that In the said Step 2, the data reconstruction includes: using the sliding window method to convert the time series data into a two-dimensional tensor, where the time dimension contains 20 consecutive sampling moments, and the feature dimension is the top 10 auxiliary variables with screened importance.

6. The soft measurement method of boiler thermal efficiency based on LTC according to claim 1, characterized in that In the said Step 3, each neuron of the four neuron layers of the soft sensor model for boiler thermal efficiency based on LTC is constructed by using an LTC model.

7. A boiler thermal efficiency soft measurement system for implementing the boiler thermal efficiency soft measurement method based on LTC according to any one of claims 1-6, characterized in that, It includes: A data acquisition module for obtaining the historical operation data of the unit's distributed control system; A preprocessing module for performing missing data processing, data dimensionality reduction, and data reconstruction on the said historical operation data; A model operation module for deploying a soft sensor model for boiler thermal efficiency based on LTC and performing prediction calculations to obtain the predicted result of the boiler thermal efficiency; A visualization terminal for online displaying the predicted result of the boiler thermal efficiency.

8. An electronic device, comprising a processor and a memory, characterized in that, The said memory stores a computer program, and when the processor executes the program, it implements the steps of the soft sensor method for boiler thermal efficiency based on LTC as described in any one of claims 1-6.

9. The electronic device according to claim 8, wherein The said electronic device is integrated into the unit's distributed control system or the edge computing terminal, and receives the boiler operation data in real time and outputs the predicted result of the boiler thermal efficiency.

Images