Solid waste incineration process furnace temperature and flue gas oxygen content multi-objective prediction method and system

Abstract

The application relates to a solid waste incineration process furnace temperature and flue gas oxygen content multi-target prediction method and system, and belongs to the technical field of solid waste incineration. The method comprises the following steps: collecting characteristic data of an incinerator, and collecting the furnace temperature and the flue gas oxygen content at the current moment, and constructing a training set and a test set; training a random configuration network prediction model based on the training set of the incinerator; inputting the test set into the trained prediction model, and recording the output value of the prediction model, so as to verify the prediction accuracy of the prediction model. The method can improve the prediction accuracy of the change trend of the furnace temperature and the flue gas oxygen content in the urban solid waste incineration process, so that the on-site operator can timely master the incineration working condition of the solid waste in the furnace, thereby guaranteeing the optimization control of the incineration process.

Description

Technical Field

[0001] This application relates to the field of solid waste incineration technology, and in particular to a multi-objective prediction method and system for furnace temperature and flue gas oxygen content in solid waste incineration process. Background Technology

[0002] Urban solid waste incineration technology boasts advantages such as small footprint, short processing time, and high resource recovery rate, making it the preferred solution for urban solid waste treatment in my country. For the solid waste incineration process, furnace temperature and flue gas oxygen content are key indicators reflecting the quality of incineration. To ensure complete combustion of solid waste and combustible flue gas within the furnace, the furnace temperature needs to be stably maintained above 850℃, and the oxygen content of the flue gas at the waste heat boiler outlet needs to be controlled between 6% and 9%. Therefore, establishing accurate predictive models for furnace temperature and flue gas oxygen content is beneficial for on-site operators to promptly grasp the incineration conditions of solid waste within the furnace, thereby ensuring the stable and efficient operation of the solid waste incineration process.

[0003] Currently, there are two main methods for predicting furnace temperature and flue gas oxygen content in urban solid waste incineration processes: mechanistic modeling and data-driven modeling. Mechanistic modeling generally starts from the perspective of material and energy conservation, constructing analytical models of furnace temperature and waste heat boilers based on physical and chemical equations, and possesses good interpretability. However, the solid waste incineration process is very complex, the mechanisms of change of some parameters are not clear, and it is affected by factors such as different incinerator types and frequent fluctuations in solid waste composition. Therefore, mechanistic modeling methods are difficult to widely apply in the field of parameter prediction modeling for urban solid waste incineration processes.

[0004] With the development of sensor and storage technologies, a large amount of operational data from solid waste incineration processes has been recorded. This online or offline data contains implicit information about changes in equipment and process parameters. Therefore, data-driven predictive modeling methods for furnace temperature and flue gas oxygen content have gained attention from researchers. Data-driven modeling methods treat the changing process of the research object as a black box model, with neural networks being a typical example. However, in terms of predictive modeling of solid waste incineration process parameters, most existing methods for predicting furnace temperature and flue gas oxygen content are single-objective models, meaning they establish prediction models for each objective separately. These single-objective models only consider the changes in a single indicator and cannot establish multi-objective prediction models based on multi-objective collaborative optimization. This makes it difficult to provide comprehensive predictive information for multi-objective collaborative optimization control of urban solid waste incineration processes, and it fails to effectively utilize the correlation between multiple objectives to improve modeling accuracy. Consequently, it cannot accurately predict furnace temperature and flue gas oxygen content based on the prediction model, preventing on-site operators from timely understanding the incineration conditions of solid waste in the furnace, failing to guarantee the optimized control of the incineration process, and reducing work efficiency. Summary of the Invention

[0005] This application provides a multi-objective prediction method and system for furnace temperature and flue gas oxygen content in solid waste incineration process. This method can improve the prediction accuracy of furnace temperature and flue gas oxygen content change trends in urban solid waste incineration process, enabling on-site operators to timely grasp the incineration conditions of solid waste in the furnace, thereby providing a guarantee for the optimized control of the incineration process.

[0006] In a first aspect, this application provides a multi-objective prediction method for furnace temperature and flue gas oxygen content in a solid waste incineration process. The method includes: collecting characteristic data of the incinerator and the furnace temperature and flue gas oxygen content at the current moment, and constructing a training set and a test set; training a randomized network prediction model based on the training set of the incinerator; inputting the test set into the trained prediction model, and recording the output value of the prediction model to verify the prediction accuracy of the prediction model.

[0007] Optionally, the characteristic data of the incinerator and the furnace temperature and flue gas oxygen content at the current moment are collected in the step described above, and a training set is constructed. The characteristic data of the incinerator includes, but is not limited to: feeder data, drying section furnace data, combustion section furnace data, burnout section furnace data, primary air volume, primary air pressure, and primary air temperature.

[0008] Optionally, the process of collecting the incinerator's characteristic data and the current furnace temperature and flue gas oxygen content, and constructing a training set, includes: using the collected incinerator's characteristic data and the current furnace temperature and flue gas oxygen content as input variables X, and the furnace temperature and flue gas oxygen content as output variables Y, to construct a training set D with N samples, described by formula (1):

[0009]

[0010] In the formula, n represents the nth set of data in the training set D, R represents the real number field, and K represents the number of input features in the training set D;

[0011] The feature variables in the training set are normalized, and the result is described by formula (2).

[0012]

[0013] In the formula, k = 1, 2, ..., K+1; n = 1, 2, ..., N, where N represents the training sample size; Min(x) represents the value of the k-th feature variable in the n-th sample after normalization. 1，k ,…,x N,k ) represents the minimum value of the k-th feature variable in the training set D, max(x) 1,k ,…,x N,k () represents the maximum value of the k-th feature variable in the training set D, x n,kThis represents the value of the k-th feature variable of the n-th training sample in the training set D.

[0014] Optionally, training the randomized network prediction model based on the training set of the incinerator includes: initializing the model parameters of the randomized network prediction model; constructing the hidden layer of the randomized network prediction model; and applying sparse constraints to the model parameters of the randomized network prediction model based on a matrix elastic network.

[0015] Optionally, the parameter initialization of the model parameters of the random configuration network prediction model includes: setting the maximum number of hidden layer neurons Lmax, the maximum number of configurations Tmax, the expected error τ, the hidden layer parameter configuration range Y, and the regularization coefficients λ and α of the random configuration network algorithm.

[0016] Optionally, based on the construction of the hidden layer of the randomly configured network prediction model, the following can be included: setting the output of the Lth node of the hidden layer constructed using a standard feedforward method. Described using formula (3):

[0017]

[0018] In the formula, g L (·) represents the activation function of the neuron in the Lth node of the hidden layer, w L With b L These represent the input weights and biases of the Lth node in the hidden layer, respectively.

[0019] Based on the standard feedforward hidden layer, a cascaded hidden layer is constructed, wherein the output of the Lth node of the cascaded hidden layer is... Described by formula (4):

[0020]

[0021] In the formula, q j This represents the connection weight between the j-th node in the hidden layer and other hidden layer nodes;

[0022] Based on the cascaded hidden layers, the hidden layers of the stochastically configured network prediction model are constructed in parallel.

[0023] Optionally, the construction of the hidden layer of the randomly configured network prediction model using a parallel construction method based on the cascaded hidden layer includes: when using parallel construction, the output h of the Lth node of the randomly configured network prediction model's hidden layer... L (X), described by formula (5):

[0024]

[0025]

[0026] In the formula, e L-1 (X) represents the network residual when the number of hidden layer neurons is L-1, and the hidden layer parameter configuration range is γ=[0.9,0.99,0.999,0.9999,0.99999], {μ L} represents a sequence of non-negative real numbers, and μ L = (1-r) / (L+1).

[0027] Optionally, the sparse constraint on the model parameters of the randomized network prediction model based on the matrix elastic network includes: constructing a matrix elastic network to apply sparse constraints to the output weights β of the randomized network prediction model, wherein the objective function of the output weights β is described by formula (7):

[0028]

[0029] In the formula, λ and α represent regularization coefficients, which are taken as 2 respectively. -8 0.9; ||·|| 2,1 L represents 2,1 The norm and the expression for iteratively solving the output weight β are described by formula (8):

[0030]

[0031]

[0032] Where t represents the number of iterations, I represents the identity matrix, and the iteration termination condition is described by formula (10):

[0033] ||β *(t+1) -β *(t) ||<10 -3 (10).

[0034] Optionally, the step of inputting test samples into the trained prediction model and recording the output value of the prediction model to verify the prediction accuracy includes: inputting test samples into the trained prediction model to obtain the output predicted values ​​of furnace temperature and flue gas oxygen content; and calculating the average root mean square error between the predicted values ​​of furnace temperature and flue gas oxygen content and the actual values ​​according to the average root mean square error evaluation method to verify the prediction accuracy of the prediction model.

[0035] Secondly, this application provides a multi-objective prediction system for furnace temperature and flue gas oxygen content in solid waste incineration processes. The system includes: a data acquisition module for acquiring characteristic data of the incinerator and the furnace temperature and flue gas oxygen content at corresponding times, and constructing a training set and a test set; a prediction model construction module for training a randomly configured network prediction model based on the training set of the incinerator; and a result prediction module for inputting the test set into the trained prediction model and recording the output value of the prediction model to verify the prediction accuracy of the prediction model.

[0036] This application has at least the following advantages:

[0037] According to the technical content provided in the embodiments of this application, historical data of characteristic variables generated during the incineration process of solid waste incineration power plants are collected to obtain training and test sets. The training set is normalized to eliminate the influence of different dimensions between variables, resulting in a unified training set and standardized data to facilitate subsequent training and prediction model acquisition. A parallel construction method is used to build the hidden layer of the randomized network prediction model, using the training set as input data for continuous learning and training to obtain the prediction model. The parallel construction method also enhances the model's nonlinear learning ability. A matrix elastic network is used to apply sparse constraints to the output weights of the randomized network prediction model, further improving modeling accuracy by utilizing the correlation between furnace temperature and flue gas oxygen content. Finally, a multi-objective prediction model for furnace temperature and flue gas oxygen content is obtained. This method can improve the prediction accuracy of furnace temperature and flue gas oxygen content trends during urban solid waste incineration, enabling on-site operators to promptly grasp the incineration conditions of solid waste in the furnace, thus ensuring optimized control of the incineration process. Attached Figure Description

[0038] Figure 1 This is a diagram illustrating the application environment of a multi-objective prediction method for furnace temperature and flue gas oxygen content in a solid waste incineration process, as shown in one embodiment.

[0039] Figure 2 This is a flowchart illustrating a multi-objective prediction method for furnace temperature and flue gas oxygen content during solid waste incineration in one embodiment.

[0040] Figure 3 This is a schematic diagram illustrating the process of training a randomly configured network prediction model in one embodiment;

[0041] Figure 4 This is a schematic diagram illustrating the process of constructing the hidden layer of a randomly configured network prediction model in one embodiment;

[0042] Figure 5 This is a flowchart illustrating the process of testing the accuracy of a prediction model in one embodiment;

[0043] Figure 6This is a line graph showing the comparison between the predicted and actual furnace temperature values ​​of a solid waste incineration process prediction model in one embodiment.

[0044] Figure 7 This is a line graph showing the comparison between the predicted and actual values ​​of the oxygen content in flue gas from a solid waste incineration process prediction model in one embodiment.

[0045] Figure 8 This is a block diagram illustrating the structure of a multi-objective prediction system for furnace temperature and flue gas oxygen content during solid waste incineration, as shown in one embodiment.

[0046] Figure 9 This is a schematic structural diagram of a computer device in one embodiment. Detailed Implementation

[0047] The present application 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 for explaining the present application and are not intended to limit the present application.

[0048] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when used in this specification, the words “comprising” and / or “including” indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0049] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details are presented in the various embodiments of this application to facilitate a better understanding of the application. However, the technical solutions claimed in this application can be implemented even without these technical details and various variations and modifications based on the following embodiments. The division of the various embodiments below is for ease of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0050] For ease of understanding, the system to which this application applies will first be described. This application provides a multi-objective prediction method for furnace temperature and flue gas oxygen content in solid waste incineration processes, which can be applied to, for example... Figure 1The system architecture shown includes a user-space file server 103 and a terminal device 101. The terminal device 101 communicates with the user-space file server 103 via a network. The user-space file server 103 can be a file server based on the NFSv3 / v4 protocol, running in a Linux environment. NFS (Network File System) is a network abstraction on top of a file system, allowing remote clients running on the terminal device 101 to access the file system over the network in a manner similar to a local file system. The terminal device 101 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc. The user-space file server 103 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0051] Figure 2 This application provides a flowchart illustrating a multi-objective prediction method for furnace temperature and flue gas oxygen content in a solid waste incineration process. The method can be implemented by, for example... Figure 1 The user-space file server in the system shown executes this. For example... Figure 2 As shown, the method may include the following steps:

[0052] S201. Collect characteristic data of the incinerator and the furnace temperature and flue gas oxygen content at the corresponding time, and construct training set and test set;

[0053] S202. Train a randomly configured network prediction model based on the training set of the incinerator;

[0054] S203. Input the test set into the trained prediction model and record the output value of the prediction model to verify the prediction accuracy of the prediction model.

[0055] The following is a detailed explanation of each step:

[0056] S201. Collect characteristic data of the incinerator and the furnace temperature and flue gas oxygen content at the corresponding time, and construct training set and test set;

[0057] Reference Figure 2 As shown in this embodiment, it should be noted that the characteristic data generated by the incinerator during the incineration process and the corresponding furnace temperature and flue gas oxygen content are collected. Specifically, 1000 sets of data can be collected as test samples to test the multi-objective prediction method for furnace temperature and flue gas oxygen content proposed in this invention, and the test samples are divided into training set and test set. The training set is used to train the constructed prediction model, and the test set is used to test whether the final prediction model can accurately predict and obtain its accuracy.

[0058] In some embodiments, in step S201, the characteristic data of the incinerator collected includes, but is not limited to, feeder data, drying section data, combustion section data, burnout section data, primary air volume, primary air pressure, and primary air temperature, etc. Specifically, 47 types of data can be collected as shown in Table 1:

[0059] Table 1 is a data table showing the characteristics of the collected data;

[0060]

[0061]

[0062] In this embodiment, it should also be noted that the above 47 variables are all process variables of the solid waste incineration process, and are also necessary conditions for predicting furnace temperature and flue gas oxygen content. By collecting historical data of these 47 characteristic variables, it is possible to train the prediction model.

[0063] Reference Figure 2 As shown, in some embodiments, in S201, characteristic data of the incinerator and the corresponding furnace temperature and flue gas oxygen content are collected, and a training set is constructed, including:

[0064] The collected characteristic data of the incinerator and the corresponding furnace temperature and flue gas oxygen content are used as input variables X, and the furnace temperature and flue gas oxygen content are used as output variables Y. A training set D with a sample size of N is constructed, which is described by formula (1):

[0065]

[0066] In the formula, n represents the nth data set in the training set D, R represents the real number field, and K represents the number of input features in the training set D, where K = 47.

[0067] The feature variables in the training set are normalized and described by formula (2).

[0068]

[0069] In the formula, k = 1, 2, ..., K+1; n = 1, 2, ..., N, where N represents the training sample size; Min(x) represents the value of the k-th feature variable in the n-th sample after normalization. 1，k ,…,x N,k ) represents the minimum value of the k-th feature variable in the training set D, max(x) 1,k ,…,x N,k () represents the maximum value of the k-th feature variable in the training set D, x n,k This represents the value of the k-th feature variable of the n-th training sample in the training set D.

[0070] In this embodiment, it should be noted that characteristic data of the incinerator and the corresponding furnace temperature and flue gas oxygen content will be collected and a training set will be constructed. The collected data will be normalized to eliminate the influence of different dimensions between variables. Here, all characteristic variables in the training set D will be normalized so that they can be directly analyzed and processed later.

[0071] Reference Figure 2 , Figure 3 As shown, in some embodiments, in S202, training a randomly configured network prediction model based on the incinerator's training set includes:

[0072] S2021. Initialize the model parameters of the randomly configured network prediction model;

[0073] S2022, Construct the hidden layer of the randomized network prediction model;

[0074] S2023. Sparse constraints are applied to the model parameters of a randomized network prediction model based on matrix elastic networks.

[0075] In this embodiment, it should be noted that the Randomly Configured Network (SCN) algorithm is a multi-layered neural network consisting of an input layer, hidden layers, and an output layer. It offers advantages such as fast training speed and general approximation properties. Based on historical data generated during the incineration process, a prediction model built using SCN technology can accurately predict multiple objectives, including furnace temperature and flue gas oxygen content.

[0076] The training set of the incinerator is used to train a randomized network prediction model. Specifically, the prediction model is constructed based on the randomized network algorithm. The learning process of the randomized network algorithm is mainly divided into two stages: randomized configuration of hidden layer neuron parameters and evaluation of hidden layer neuron output weights. Here, the prediction model is constructed based on the SCN structure. First, its parameters are configured and initialized, and the parameter settings of the algorithm used in the prediction model are configured. Then, the hidden layers are constructed, and the output weights of the hidden layer neurons are evaluated for subsequent calculations.

[0077] In some embodiments, refer to Figure 3As shown in S2021, the model parameters of the randomly configured network prediction model are initialized, including: setting the maximum number of hidden layer neurons Lmax, the maximum number of configurations Tmax, the expected error τ, the hidden layer parameter configuration range Y, and the regularization coefficients λ and α, to obtain the final prediction model. Specifically, the settings are: maximum number of hidden layer nodes Lmax = 60, maximum number of configurations Tmax = 50, and the termination condition is set to τ, i.e., the expected training error τ = 0.001, the hidden layer parameter configuration range Y = [1:1:10], and the regularization coefficient λ = 2. -8 α = 0.9, so that a prediction model can be built later.

[0078] In some embodiments, refer to Figure 3 , Figure 4 As shown in S2022, the hidden layer of the randomly configured network prediction model is constructed, including:

[0079] S20221. Set the output of the Lth node of the hidden layer in the standard feedforward construction. Described using formula (3):

[0080]

[0081] In the formula, g L (·) represents the activation function of the neuron in the Lth node of the hidden layer, w L With b L These represent the input weights and biases of the Lth node in the hidden layer, respectively.

[0082] S20222. Based on the standard feedforward construction of the hidden layer, a cascaded hidden layer is set up, wherein the output of the Lth node of the cascaded hidden layer is... Described by formula (4):

[0083]

[0084] In the formula, q j This represents the connection weight between the j-th node in the hidden layer and other hidden layer nodes;

[0085] S20223. Based on the cascaded hidden layers, construct the hidden layers of the randomly configured network prediction model in a parallel manner.

[0086] In this embodiment, it should be noted that the neural network includes multiple neuron "layers," namely the input layer, hidden layers, and the output layer. The hidden layers are responsible for the necessary calculations and outputting the results to the output layer. Based on the standard feedforward neural network, for multi-objective prediction, a randomly configured network of hidden layers is constructed in parallel according to the cascaded hidden layers. This enhances the nonlinear learning ability of the prediction model, thereby constructing a multi-objective prediction model for furnace temperature and flue gas oxygen content.

[0087] The activation function of the neuron in the Lth node of the hidden layer primarily performs a nonlinear transformation on the computation result. Initially, a multi-layered neural network is constructed, where each layer performs a linear weighted summation operation. Therefore, this neural network can only perform linear classification at this stage. To perform a nonlinear transformation on each node in each layer, a value is calculated for each activation in the next layer based on the activations, weights, and biases of the previous layers. However, before sending this value to the next layer, the output is scaled using an activation function to achieve the nonlinear classification task. Commonly used activation functions include the Sigmoid activation function, the tanh activation function, and the ReLU activation function. In this embodiment, the Sigmoid activation function is used. The Sigmoid function is between 0 and 1, and each input neuron, node, or activation is scaled to a value between 0 and 1.

[0088] In some embodiments, refer to Figure 4 As shown in S20223, based on the cascaded hidden layer, the hidden layer of the stochastically configured network prediction model is constructed in parallel, including: when constructing in parallel, the output h of the Lth node of the stochastically configured network prediction model's hidden layer. L (X), described by formula (5):

[0089]

[0090]

[0091] In the formula, e L-1 (X) represents the network residual when the number of hidden layer neurons is L-1, and the hidden layer parameter configuration range is Υ=[0.9,0.99,0.999,0.9999,0.99999], {μ L} represents a sequence of non-negative real numbers, and μ L = (1-r) / (L+1).

[0092] In this embodiment, it should be noted that constructing the randomly configured hidden layers of the network using a parallel construction method can enhance the nonlinear learning ability of the model, thereby achieving multi-objective prediction of furnace temperature and flue gas oxygen content. First, the model parameters are configured, and the training set is used as input to the SCN algorithm to train the SCN. The output results are compared with preset conditions, and the learning process continues until the preset conditions are met, resulting in a prediction model based on the SCN algorithm.

[0093] Specifically, based on the output of the Lth node of the hidden layer constructed according to the standard feedforward method... Add the connection weight q between the j-th node in the hidden layer and other hidden layer nodes. j The output of the Lth node of the cascaded hidden layer is obtained. Then, using parallel construction, the output of the Lth node of the hidden layer constructed through the standard feedforward method is... and the output of the Lth node of the cascaded hidden layer The output h of the Lth node in the hidden layer of the randomly configured network is obtained. L (X) is used to construct a prediction model. The training set is used as input to the SCN algorithm to train the SCN. At this time, the output residual of the randomly configured network is e. L-1 (X) represents the network residual when the number of hidden layer neurons is L-1, according to equation (6) in the expected error ε L When the value is greater than or equal to 0, the preset conditions are met, and a prediction model based on the SCN algorithm is obtained.

[0094] In some embodiments, refer to Figure 3 As shown in S2023, sparse constraints are applied to the model parameters of the randomized network prediction model based on the matrix elastic network, including:

[0095] A matrix elastic network is constructed to impose sparse constraints on the output weights β of the stochastically configured network prediction model. The objective function of the output weights β is described by formula (7):

[0096]

[0097] In the formula, λ and α represent regularization coefficients, which are taken as 2 respectively. -8 0.9, ||·|| 2,1 L represents 2,1 The norm and the expression for iteratively solving the output weight β are described by formula (8):

[0098]

[0099]

[0100] Where t represents the number of iterations, I represents the identity matrix, and the iteration termination condition is described by formula (10):

[0101] ||β *(t+1) -β *(t) ||<10 -3 (10).

[0102] In this embodiment, it should be noted that the cost function of the elastic network regression algorithm combines the regularization methods of Lasso regression and ridge regression, using two parameters λ and α to control the magnitude of the penalty term. The L1 paradigm can generate relatively sparse solutions and possesses a certain feature selection capability, which is useful when solving high-dimensional feature spaces; the L2 paradigm is mainly used to prevent overfitting. Here, the L1 and L2 paradigms are combined to form new constraints, and sparsity is used to represent the relationship between the L1 and L2 paradigms, thereby determining the correlation between furnace temperature and flue gas oxygen content, further improving the modeling accuracy.

[0103] Specifically, based on the supervision mechanism of the sparse constraint of the matrix elastic network algorithm in equation (7), the Lth hidden layer neuron is randomly generated within the hidden layer parameter configuration range Y = [0.9, 0.99, 0.999, 0.9999, 0.99999]. At this time, the input weights and biases of the Lth node of the hidden layer are w L With b L The hidden layer output weights β of the randomly configured network are redefined using equation (8) until the residual is less than the expected error τ. The value of τ is set according to the allowable error of the model, and here τ = 0.001. Through the supervision mechanism of sparse constraints in the matrix elastic network algorithm, the general approximation property of the randomly configured network is guaranteed, the correlation between furnace temperature and flue gas oxygen content is determined, and the modeling accuracy is further improved.

[0104] In some embodiments, refer to Figure 2 , Figure 5 As shown in step S203, the test samples are input into the trained prediction model, and the output value of the prediction model is recorded to verify the prediction accuracy of the prediction model, including:

[0105] S2031. Input the test samples into the trained prediction model to obtain the predicted values ​​of furnace temperature and flue gas oxygen content.

[0106] S2032. Based on the average root mean square error evaluation method, calculate the average root mean square error between the predicted values ​​and the actual values ​​of furnace temperature and flue gas oxygen content to verify the prediction accuracy of the prediction model.

[0107] In this embodiment, it should be noted that, based on the established prediction model, the collected test set is used as the input data of the prediction model, and the output data obtained through the prediction model is the predicted value of furnace temperature and flue gas oxygen content. This enables accurate prediction of multiple objectives of furnace temperature and flue gas oxygen content, so as to provide a guarantee for the optimized control of the incineration process, and enable on-site operators to keep abreast of the incineration conditions of solid waste in the furnace and improve work efficiency.

[0108] Specifically, the prediction model is tested by collecting multiple sets of historical data of the incineration process as a test set. The furnace temperature and flue gas oxygen content are obtained through the prediction model. The root mean square error (RMSE) is used to evaluate the overall prediction performance of the model by comparing the actual values ​​of furnace temperature and flue gas oxygen content measured by instruments with the predicted values ​​output by the prediction model. The smaller the RMSE, the higher the prediction accuracy.

[0109] Specifically, in one example, refer to Figure 6 , Figure 7 As shown, 200 sets of historical incineration process data were collected as input data for the prediction model. The predicted values ​​were compared with the actual furnace temperature and flue gas oxygen content corresponding to the 200 sets of historical incineration process data. By using the average root mean square error (RMSE) evaluation method, the average root mean square errors of the predicted values ​​and the actual values ​​of furnace temperature and flue gas oxygen content were calculated to be 2.38 (°C) and 0.23 (%), respectively, which meet the target errors. This indicates that the method can achieve accurate estimation of furnace temperature and flue gas oxygen content in the urban solid waste incineration process. The multi-objective prediction of furnace temperature and flue gas oxygen content based on SCN can accurately predict the furnace temperature and flue gas oxygen content in the urban solid waste incineration process, providing a guarantee for the optimized control of the incineration process, enabling on-site operators to keep abreast of the incineration conditions of solid waste in the furnace and improve work efficiency.

[0110] The implementation principle of this embodiment is as follows: The above steps mainly involve collecting historical data of characteristic variables generated during the incineration process of solid waste incineration power plants to obtain training and test sets. The training set is then normalized to eliminate the influence of different variables on their dimensions, resulting in a unified training set and standardized data to facilitate subsequent training and the acquisition of a prediction model. A parallel construction method is used to build the hidden layer of the stochastically configured network prediction model. The training set is used as input data, and the model is continuously trained to obtain the prediction model. This parallel construction method also enhances the model's nonlinear learning ability. A matrix elastic network is used to apply sparse constraints to the output weights of the stochastically configured network prediction model, further improving the modeling accuracy by utilizing the correlation between furnace temperature and flue gas oxygen content. Finally, a multi-objective prediction model for furnace temperature and flue gas oxygen content is obtained. This method can improve the prediction accuracy of the changing trends of furnace temperature and flue gas oxygen content during urban solid waste incineration, enabling on-site operators to promptly grasp the incineration conditions of solid waste in the furnace, thus ensuring optimal control of the incineration process.

[0111] Reference Figure 8 As shown in the illustration, this application also provides a system for multi-objective prediction of furnace temperature and flue gas oxygen content in solid waste incineration processes. This system may include: a data acquisition module, a prediction model construction module, and a result prediction module. The main functions of each component module are as follows:

[0112] The data acquisition module 301 is used to collect characteristic data of the incinerator and the furnace temperature and flue gas oxygen content at the corresponding time, and to build training and testing sets.

[0113] The prediction model construction module 302 is used to train a randomly configured network prediction model based on the training set of the incinerator;

[0114] The result prediction module 303 is used to input the test set into the trained prediction model and record the output value of the prediction model to verify the prediction accuracy of the prediction model.

[0115] According to embodiments of this application, this application also provides a computer device and a computer-readable storage medium.

[0116] like Figure 9 The diagram shown is a block diagram of a computer device according to an embodiment of this application. The term "computer device" is intended to represent various forms of digital computers or mobile devices. The digital computer may include a desktop computer, a portable computer, a workbench, a personal digital assistant, a server, a mainframe computer, and other suitable computers. The mobile device may include a tablet computer, a smartphone, a wearable device, etc.

[0117] like Figure 9 As shown, device 600 includes a computing unit 601, a ROM 602, a RAM 603, a bus 604, and an input / output (I / O) interface 605. The computing unit 601, ROM 602, and RAM 603 are interconnected via the bus 604. The input / output (I / O) interface 605 is also connected to the bus 604.

[0118] The computing unit 601 can execute various processes in the method embodiments of this application according to computer instructions stored in the read-only memory (ROM) 602 or computer instructions loaded from the storage unit 608 into the random access memory (RAM) 603. The computing unit 601 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. The computing unit 601 can include, but is not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. In some embodiments, the methods provided in the embodiments of this application can be implemented as computer software programs, which are tangibly contained in a computer-readable storage medium, such as the storage unit 608.

[0119] RAM 603 may also store various programs and data required for the operation of device 600. Part or all of the computer program may be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609.

[0120] The input unit 606, output unit 607, storage unit 608, and communication unit 609 in device 600 can be connected to I / O interface 605. The input unit 606 can be, for example, a keyboard, mouse, touchscreen, or microphone; the output unit 607 can be, for example, a display, speaker, or indicator light. Device 600 can exchange information and data with other devices through the communication unit 609.

[0121] It should be noted that the device may also include other components necessary for normal operation. It may also include only the components necessary for implementing the solution of this application, without necessarily including all the components shown in the figures.

[0122] Various implementations of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SOCs), payload programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof.

[0123] The computer instructions used to implement the methods of this application may be written in any combination of one or more programming languages. These computer instructions may be provided to the computing unit 601 such that when executed by the computing unit 601, such as a processor, the computer instructions cause the execution of the steps involved in the embodiments of the methods of this application.

[0124] The computer-readable storage medium provided in this application can be a tangible medium that can contain or store computer instructions for performing the steps involved in the method embodiments of this application. The computer-readable storage medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, and other forms of storage media.

[0125] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A multi-objective prediction method for furnace temperature and flue gas oxygen content in solid waste incineration process, characterized in that, Includes the following steps: Collect characteristic data of the incinerator and the corresponding furnace temperature and flue gas oxygen content at the corresponding time, and construct training and testing sets; A randomized network prediction model is trained based on the training set of the incinerator. The test set is input into the trained prediction model, and the output value of the prediction model is recorded to verify the prediction accuracy of the prediction model. The training of the randomly configured network prediction model based on the training set of the incinerator includes: Initialize the model parameters of the randomly configured network prediction model; Constructing the hidden layers of a randomly configured network prediction model: Sparse constraints are applied to the model parameters of a randomized network prediction model based on matrix elasticity networks. The hidden layer of the constructed randomized network prediction model includes: The hidden layer with standard feedforward construction is set up. L Output of each node It can be described by formula (3): （3）， In the formula, g L (·) indicates the hidden layer. L The activation functions of neurons in each node. w L and b L They represent the hidden layers respectively. L The input weights and biases of each node; Based on the standard feedforward construction of the hidden layer, a cascaded hidden layer is set up, wherein the cascaded hidden layer is the first... L Output of each node Described by formula (4): （4）， In the formula, q j Indicates the hidden layer number 1 j The connection weights between each node and other hidden layer nodes; Based on the cascaded hidden layers, the hidden layers of the randomly configured network prediction model are constructed in parallel. The construction of hidden layers for a randomly configured network prediction model based on cascaded hidden layers, using a parallel construction approach, includes: When using parallel construction, the hidden layers of the network prediction model are randomly configured. L Output of each node It can be described by formula (5): （5）， （6）， In the formula, This represents the network residual when the number of hidden layer neurons is L-1, and the range of hidden layer parameter configurations. =[0.9, 0.99, 0.999, 0.9999, 0.99999], {μ L } represents a sequence of non-negative real numbers, and μ L =(1-r) / (L+1); The sparse constraint on the model parameters of the prediction model for a randomly configured network based on a matrix elastic network includes: Constructing the output weights of a matrix elastic network for a randomly configured network prediction model β Apply sparsity constraints and output weights. β The objective function is described by formula (7): （7）， In the formula, λ and α This represents the regularization coefficient, which is 2. -8 0.9; ||·|| 2,1 express L 2,1 Norm, output weights β The expression for the iterative solution is described by formula (8): （8） （9） in, t Indicates the number of iterations. I Let represent the identity matrix, and the iteration termination condition is described by formula (10): （10）。 2. The multi-objective prediction method for furnace temperature and flue gas oxygen content in solid waste incineration process according to claim 1, characterized in that, The steps involve collecting characteristic data of the incinerator and the corresponding furnace temperature and flue gas oxygen content at the time of the incinerator, and constructing a training set. The characteristic data of the incinerator includes, but is not limited to: feeder data, drying section furnace data, combustion section furnace data, burnout section furnace data, primary air volume, primary air pressure, and primary air temperature.

3. The multi-objective prediction method for furnace temperature and flue gas oxygen content in solid waste incineration process according to claim 2, characterized in that, The process involves collecting characteristic data of the incinerator and corresponding furnace temperature and flue gas oxygen content at specific times, and constructing a training set, including: The collected characteristic data of the incinerator and the corresponding furnace temperature and flue gas oxygen content at the corresponding time are used as input variables. X Furnace temperature and flue gas oxygen content are used as output variables. Y The number of samples constructed is N training set D It can be described by formula (1): （1）， In the formula, n Indicates training set D The nth set of data, R Represents the real number field. K Indicates training set D The number of input features; The feature variables in the training set are normalized, and the result is described by formula (2). (2)， In the formula, k =1,2,…, K +1; n =1,2,..., N , here N Indicates the training sample size; Indicates the normalized i-th n The first sample k The value of each feature variable is min( x 1，k ,…,x N,k ) represents the training set D The Middle k The minimum value of each characteristic variable, max( x 1, k ,…,x N,k ) represents the training set D The Middle k The maximum value of each characteristic variable. x n,k Indicates training set D The first in n The training sample of the th training sample k The values ​​of each characteristic variable.

4. The multi-objective prediction method for furnace temperature and flue gas oxygen content in solid waste incineration process according to claim 1, characterized in that, The parameter initialization of the model parameters for the randomly configured network prediction model includes: Set the maximum number of hidden layer neurons (Lmax), the maximum number of configuration attempts (Tmax), the expected error (τ), and the configuration range of hidden layer parameters for the randomized network algorithm. The regularization coefficients λ and α.

5. The multi-objective prediction method for furnace temperature and flue gas oxygen content in solid waste incineration process according to claim 1, characterized in that, The step of inputting the test set into the trained prediction model and recording the output value of the prediction model to verify the prediction accuracy includes: The test set is input into the trained prediction model to obtain the predicted values ​​of furnace temperature and flue gas oxygen content. Based on the average root mean square error (RMSE) evaluation method, the average root mean square error between the predicted and actual values ​​of furnace temperature and flue gas oxygen content was calculated to verify the prediction accuracy of the prediction model.

6. A system for multi-objective prediction of furnace temperature and flue gas oxygen content in solid waste incineration processes, characterized in that, The system includes: The data acquisition module is used to collect characteristic data of the incinerator and the furnace temperature and flue gas oxygen content at the corresponding time, and to build training and testing sets. A prediction model construction module is used to train a randomly configured network prediction model based on the training set of the incinerator; the training of the randomly configured network prediction model based on the training set of the incinerator includes: Initialize the model parameters of the randomly configured network prediction model; Constructing the hidden layers of a randomly configured network prediction model: Sparse constraints are applied to the model parameters of a randomized network prediction model based on matrix elasticity networks. The hidden layer of the constructed randomized network prediction model includes: The hidden layer with standard feedforward construction is set up. L Output of each node It can be described by formula (3): （3）， In the formula, gL(·) represents the first hidden layer. L The activation function of the neurons in each node, w L With b L They represent the hidden layers respectively. L The input weights and biases of each node; Based on the standard feedforward construction of the hidden layer, a cascaded hidden layer is set up, wherein the cascaded hidden layer is the first... L Output of each node Described by formula (4): （4）， In the formula, q j Indicates the hidden layer number 1 j The connection weights between each node and other hidden layer nodes; Based on the cascaded hidden layers, the hidden layers of the randomly configured network prediction model are constructed in parallel. The construction of hidden layers for a randomly configured network prediction model based on cascaded hidden layers, using a parallel construction approach, includes: When using parallel construction, the hidden layers of the network prediction model are randomly configured. L Output of each node It can be described by formula (5): （5）， （6）， In the formula, This represents the network residual when the number of hidden layer neurons is L-1, and the range of hidden layer parameter configurations. =[0.9, 0.99, 0.999, 0.9999, 0.99999], {μ L } represents a sequence of non-negative real numbers, and μ L =(1-r) / (L+1); The sparse constraint on the model parameters of the prediction model for a randomly configured network based on a matrix elastic network includes: A matrix elastic network is constructed to impose sparse constraints on the output weights β of the randomly configured network prediction model. The objective function of the output weights β is described by formula (7): （7）， In the formula, λ and α represent regularization coefficients, which are taken as 2. -8 0.9; ||·||2,1 indicates L 2,1 Norm, output weights β The expression for the iterative solution is described by formula (8): （8） （9） in, t Indicates the number of iterations. I Let represent the identity matrix, and the iteration termination condition is described by formula (10): （10）； The result prediction module is used to input the test set into the trained prediction model and record the output value of the prediction model to verify the prediction accuracy of the prediction model.

Images