Fly ash carbon content prediction method and device, computer equipment and storage medium

Abstract

The invention relates to a fly ash carbon content prediction method and device, computer equipment and a storage medium. Comprising the following steps: determining a plurality of key influence factors influencing the coal-fired boiler fly ash carbon content; at the target moment, multiple pieces of target operation data, corresponding to the key influence factors, of the coal-fired boiler are obtained; and inputting the plurality of target operation data into a pre-trained target fly ash carbon content prediction model to obtain a target fly ash carbon content corresponding to the target moment. In this way, redundant influence factors can be eliminated, and then the subsequent calculation complexity of key influence factors is reduced; and the prediction accuracy of the carbon content of the fly ash is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of thermal automation control, and in particular to a method, device, computer equipment, and storage medium for predicting the carbon content of fly ash. Background Art

[0002] Since the carbon content of fly ash during boiler combustion is directly related to the power generation cost and economic benefits of thermal power plants, the carbon content of fly ash can be used as a key indicator to measure boiler combustion efficiency.

[0003] In related technologies, machine learning can be used to construct data-driven prediction models for boiler operating parameters, which can then be used to predict the carbon content in fly ash. However, these technologies often directly use the full set of operating parameters as model inputs. This consumes significant computational resources, potentially leading to poor prediction accuracy for the carbon content in fly ash. Summary of the Invention

[0004] In view of this, the embodiments of the present disclosure provide a method, apparatus, computer equipment and storage medium for predicting the carbon content of fly ash to solve the problems existing in the related art.

[0005] In a first aspect of an embodiment of the present disclosure, a method for predicting the carbon content in fly ash is provided, the method comprising: determining a plurality of key influencing factors affecting the carbon content in fly ash of a coal-fired boiler; obtaining a plurality of target operating data corresponding to the plurality of key influencing factors of the coal-fired boiler at a target moment; and inputting the plurality of target operating data into a pre-trained target fly ash carbon content prediction model to obtain a target fly ash carbon content corresponding to the target moment.

[0006] A second aspect of an embodiment of the present disclosure provides a fly ash carbon content prediction device, which is applied to the fly ash carbon content prediction selection method as in the first aspect. The device includes: a determination module for determining multiple key influencing factors affecting the fly ash carbon content of a coal-fired boiler; an acquisition module for acquiring multiple target operating data of the coal-fired boiler corresponding to the multiple key influencing factors at a target time; and an acquisition module for inputting the multiple target operating data into a pre-trained target fly ash carbon content prediction model to obtain the target fly ash carbon content corresponding to the target time.

[0007] According to a third aspect of the embodiments of the present disclosure, a computer device is provided, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-mentioned fly ash carbon content prediction method.

[0008] According to a fourth aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, on which a computer program / instruction is stored. When the computer program / instruction is executed by a processor, the steps of the above-mentioned fly ash carbon content prediction method are implemented.

[0009] According to a fifth aspect of the present disclosure, a computer program product is provided, which implements the steps of the above-mentioned fly ash carbon content prediction method when the computer program / instructions are executed by a processor.

[0010] At least one of the above-mentioned technical solutions adopted in the embodiments of the present disclosure can achieve the following beneficial effects: by determining multiple key influencing factors affecting the fly ash carbon content of a coal-fired boiler; at the target time, obtaining multiple target operating data corresponding to the multiple key influencing factors of the coal-fired boiler; inputting the multiple target operating data into a pre-trained target fly ash carbon content prediction model to obtain the target fly ash carbon content corresponding to the target time. In this way, redundant influencing factors can be eliminated, thereby reducing the subsequent calculation complexity of the key influencing factors; at the same time, obtaining multiple target operating data of multiple key influencing factors at the target time can accurately capture the temporal dynamic characteristics of the combustion process and avoid interference from irrelevant time data; finally, inputting the multiple target operating data into the pre-trained target fly ash carbon content prediction model can enable the target fly ash carbon content prediction model to adapt to different working conditions, thereby improving the prediction accuracy and generalization ability of the fly ash carbon content, thereby helping thermal power plants optimize combustion parameters, reduce carbon losses and power generation costs, and achieve a dual improvement in combustion efficiency and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other purposes, features, and advantages of the present disclosure will become more apparent through a more detailed description of the embodiments of the present disclosure in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and are not intended to limit the present disclosure. In the drawings, the same reference numerals generally represent the same components or steps.

[0012] Figure 1 A schematic diagram of a process for constructing a target fly ash carbon content prediction model provided by an exemplary embodiment of the present disclosure;

[0013] Figure 2 A schematic structural diagram of a preset fly ash carbon content prediction model provided by an exemplary embodiment of the present disclosure;

[0014] Figure 3 A schematic diagram of the structure of a VSN provided by an exemplary embodiment of the present disclosure;

[0015] Figure 4 A schematic structural diagram of an MHA provided as an exemplary embodiment of the present disclosure;

[0016] Figure 5 A schematic flow chart of a method for predicting carbon content in fly ash provided by an exemplary embodiment of the present disclosure;

[0017] Figure 6 A schematic structural diagram of a device for predicting carbon content in fly ash provided in one embodiment of the present disclosure;

[0018] Figure 7 A schematic structural diagram of an electronic device provided in one embodiment of the present disclosure;

[0019] Figure 8 A schematic diagram of the structure of a computer system provided in one embodiment of the present disclosure;

[0020] Figure 9 A schematic diagram of a computer program product provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0021] The following describes embodiments of the present disclosure in more detail with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of protection of the present disclosure.

[0022] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0023] The term "including" and its variations used in this document are open inclusions, that is, "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the description below. It should be noted that the concepts of "first", "second", etc. mentioned in this disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0024] It should be noted that the modifications of "one" and "multiple" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, they should be understood as "one or more".

[0025] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0026] Generally speaking, since the carbon content of fly ash during boiler combustion directly impacts the power generation costs and economic benefits of thermal power plants, fly ash carbon content can be used as a key indicator for measuring boiler combustion efficiency. Here, the carbon content of boiler fly ash refers to the mass fraction of carbon in the fly ash after pulverized coal is burned in the boiler furnace.

[0027] Currently, the main methods for measuring fly ash carbon content include loss on ignition, microwave absorption, and infrared measurement. However, in actual measurement scenarios, due to the diversity of coal types and the fact that the particle size composition of fly ash produced during the coal combustion process varies with equipment operating conditions, measuring fly ash carbon content using any of these methods may result in low measurement accuracy and insufficient stability. However, with the development of distributed control systems and artificial energy technologies, data-driven soft sensor modeling methods have been widely used to measure fly ash carbon content. This method does not require the establishment of a precise model of the fly ash formation mechanism, but instead predicts fly ash content by monitoring boiler operating parameters.

[0028] In the related art, a data-driven prediction model can be constructed for the operating parameters of the boiler based on machine learning. For example, a soft measurement method for fly ash carbon content determined based on gradient boosting decision tree (Light Gradient Boosting Machine, LightGBM) and extreme gradient boosting tree (eXtreme Gradient Boosting, XGBoost) can integrate multiple machine learning algorithms to analyze the various operating parameters of the boiler. At the same time, the correlation matrix and packaging method are used to remove redundant features in two steps and extract important features; finally, the LightGBM and XGBoost combination model is used to establish a fly ash carbon content soft measurement model. It should be understood that the above-mentioned LightGBM is a fast and efficient gradient boosting framework based on the decision tree algorithm, which is particularly suitable for processing large-scale data sets and high-dimensional data. XGBoost is a machine learning algorithm based on decision tree integration. The core idea is to gradually optimize the model prediction ability by iteratively training multiple decision trees. It is often used for tasks such as classification and regression.

[0029] For example, a fly ash carbon content prediction method based on genetic simulated annealing parameter optimization includes the following steps: first, the principal component analysis method is used to screen the operating parameters in the historical data of fly ash carbon content; then, a fly ash carbon content prediction model is established based on a multi-layer feedforward neural network based on the error back propagation algorithm; finally, the genetic simulated annealing algorithm is used to optimize the weights and thresholds of the fly ash carbon content prediction model to obtain the optimal weights and thresholds.

[0030] In practical applications, the prediction models built based on machine learning in related technologies have the following two problems:

[0031] Question 1: The data-driven "black box" model lacks transparency and interpretability, which restricts the ability of relevant technical personnel to understand the internal mechanisms of the model and implement precise control.

[0032] Problem 2: Due to the inertia and hysteresis of the boiler's combustion system, the carbon content in fly ash is closely related not only to the current boiler operating status and operating parameters, but also to the cumulative effects and dynamic changes of various parameters at multiple past moments. However, existing models struggle to effectively capture this complex temporal dependency, which can easily lead to missing information about the coupling between historical and current data, resulting in significant deviations between the predicted and actual fly ash carbon content.

[0033] To address the above issues, the disclosed embodiments provide a method for predicting fly ash carbon content. First, a time series feature set is constructed, and operating data corresponding to multiple key influencing factors at multiple moments are introduced to capture the delay and inertia dynamic characteristics of the boiler combustion system. Second, a variable selection network is used to calculate the factor weights corresponding to the key influencing factors, and the moment weights corresponding to each moment are extracted through a multi-head attention mechanism. In this way, the degree of influence of different key influencing factors and moments on the fly ash carbon content can be quantified, and weight analysis of key influencing factors and time series can be achieved. Then, a clustering algorithm and elbow analysis method are combined to cluster the factor weights and moment weights, and target key influencing factors and target moments with weight values ​​higher than a preset threshold are screened out, thereby optimizing the key nodes of the feature space and time series. Finally, the target operating data corresponding to the target key influencing factors at the target moment can be used to train the preset fly ash carbon content prediction model to obtain the target fly ash carbon content prediction model. Based on this, high-precision real-time prediction of fly ash carbon content and identification of key influencing factors can be achieved, solving the prediction problems caused by the lack of interpretability of traditional "black box" models and system inertia lag.

[0034] Figure 1 A schematic diagram of a process for constructing a target fly ash carbon content prediction model is provided in accordance with an exemplary embodiment of the present disclosure. Figure 1 As shown, specifically including:

[0035] S101: Determine key influencing factors and obtain operating data corresponding to these key influencing factors at different times during the boiler combustion process. Key influencing factors may include coal feeder rate, secondary air damper openings at each layer, burnout air damper openings, economizer outlet oxygen setting, unit operating load, and fly ash carbon content.

[0036] In some embodiments, the boiler's distributed control system can directly acquire real-time operating data from the boiler during the combustion process. Specifically, the boiler's distributed control system can acquire real-time operating data generated by the boiler during the combustion process through sensors distributed across key locations of the boiler, such as flow sensors, temperature sensors, and damper opening sensors. The data acquired at different times is stored in the boiler's distributed control system's historical database for subsequent use.

[0037] S102: Build a preset fly ash carbon content prediction model based on a variable selection network (VSN) and a multi-head attention mechanism (MHA). VSN is a component used in deep learning and data analysis to dynamically select important input features, aiming to improve model performance, enhance interpretability, and reduce computational complexity. The core goal of MHA is to enable the model to capture the correlation between input data from multiple dimensions and different subspaces, thereby gaining a more comprehensive and flexible understanding of key influencing factors.

[0038] Figure 2 This is a schematic diagram of a preset fly ash carbon content prediction model provided by an exemplary embodiment of the present disclosure. Figure 2 As shown, in some embodiments, the data input to the preset fly ash carbon content prediction model may include multiple operating data corresponding to key influencing factors at different times. Therefore, in order to simplify the subsequent data processing, VSN can be used to screen the key influencing factors and select the target key influencing factors that meet specific conditions; then the target key influencing factors are input into MHA, and MHA can determine the target moment from multiple moments, and then the operating data corresponding to the target key influencing factors at the target moment can be analyzed. By fusing the operating data corresponding to each moment, the dependency between each moment can be captured. In this process, the dependency between the key operating data corresponding to each moment can be captured; finally, the weights calculated by MHA can be input into the linear layer (Linear), and the linear layer performs linear transformation to obtain the predicted value of fly ash carbon content.

[0039] S103 , using the operating data corresponding to the key influencing factors at different times to train the preset fly ash carbon content prediction model, to obtain a target fly ash carbon content prediction model, specifically including steps S1031 to S1035 .

[0040] S1031, considering the dynamic characteristics of delay and inertia during boiler combustion, a feature vector corresponding to a certain moment can be constructed based on the operating data of the key influencing factors at that moment, and then a time series feature set can be constructed based on the feature vectors at different moments.

[0041] In some embodiments, since the boiler is usually burning continuously, the operating data generated by the boiler during the combustion process can be obtained, specifically the operating data of the key influencing factors at different times. Based on this, the feature vector corresponding to each moment can be constructed based on the operating data corresponding to that moment, and then the time series feature set [X t-d+1 ,X t-d+2 ,…,X t ]. Where t represents the time; d represents the time step; X t represents the eigenvector corresponding to time t, X t Including multiple operating data corresponding to time t; X t-d+1 represents the eigenvector corresponding to time t-d+1, X t-d+1 Including multiple operating data corresponding to time t-d+1; X t-d+2 represents the eigenvector corresponding to time t-d+2, X t-d+2 Including multiple operating data corresponding to time t-d+2. Here, Figure 2 As shown, there may be multiple VSNs for the preset fly ash carbon content prediction model, wherein one VSN may correspond to one feature vector.

[0042] It should be understood that the historical time step d represents the past time span considered by the model. For example, if d = 5, the model will use data from time t and the four previous time steps (t-1, t-2, t-3, and t-4), a total of five time steps, for prediction. Furthermore, when obtaining operating data corresponding to different time points, it can be obtained at regular time intervals, for example, at 20-second intervals.

[0043] In some embodiments, in view of the dynamic characteristics of delay and inertia in the boiler combustion process, it is difficult to accurately predict the fly ash carbon content by simply relying on the current moment characteristics. Therefore, in order to capture the dynamic behavior of the boiler combustion system, the embodiment of the present disclosure sets the input feature space of the preset fly ash carbon content prediction model as a time series feature set, and introduces the fly ash carbon content at historical moments into the feature space. In this way, through this autoregressive method, the inertia and hysteresis effects of the system under different operating conditions can be better reflected, and the prediction accuracy of the model can be effectively improved. Based on this, when the operating data corresponding to the historical moments are introduced to train the preset fly ash carbon content prediction model, the inertia and hysteresis of the boiler combustion system can be captured, and then by fusing the operating data at multiple moments, the time dependence of the combustion state can be more comprehensively characterized.

[0044] S1032: Based on the time series feature set, a preset fly ash carbon content prediction model is trained and tested in the original high-dimensional feature space, and a grid search algorithm is used to optimize the structure of the preset fly ash carbon content prediction model.

[0045] In some embodiments, the input feature space corresponding to the time series feature set can be expressed as in, represents the real number space, N represents the dimension of the input feature, and d represents the time step. Based on this, when using the grid search algorithm to optimize the structure of the preset fly ash carbon content prediction model, the optimization parameters can be determined first, and the range corresponding to each optimization parameter can be determined.

[0046] Specifically, the optimization parameters may include VSN parameters, MHA parameters, and regularization and training parameters, wherein the VSN parameters may include the number of fully connected layer output units of the initial feature conversion, the corresponding range may be [4, 8, 16], and the number of hidden units of the GRN, the corresponding range may be {16, 32, 64}; the MHA parameters may include the number of multi-head attention heads, the corresponding range may be {1, 2, 3, 4}, and the number of hidden units, the corresponding range may be {16, 32, 64}; the number of fully connected layer output units may include the Dropout probability, the corresponding range may be {0.1, 0.2, 0.3, 0.4}; and the learning rate, the corresponding range may be {0.01, 0.001, 0.0001}.

[0047] S1033: Determine the factor weight corresponding to each key influencing factor using the VSN, and determine a target key influencing factor from the multiple key influencing factors according to the factor weight corresponding to each key influencing factor.

[0048] Figure 3 FIG. 1 is a structural diagram of a VSN provided by an exemplary embodiment of the present disclosure. Figure 3As shown in the figure, Sigmoid is an activation function; linear is a linear layer, which is used to perform linear transformation on the data in the input model; Gate is a gating unit, which is used to control the flow of data and decide whether information passes through; Dropout is a technology to prevent overfitting of the neural network, which randomly sets the output of some neurons to 0; ELU (Exponential Linear Unit) is a linear unit activation function, which is used to speed up the convergence of the model and alleviate the gradient disappearance problem; LayerNorm is a normalization operation, which normalizes the input of a certain layer of the neural network to stabilize the training of the model.

[0049] Among them, the input of a VSN can be all the running data corresponding to a moment, which can be expressed as in, Indicates the operating data corresponding to the key influencing factor marked as 1 at time t, It represents the operating data corresponding to the key influencing factor marked as i at time t, Represents the operating data corresponding to the key influencing factor identified as n at time t, i = 1, 2...n.

[0050] In some embodiments, the VSN can be composed of multiple Gated Residual Networks (GRNs). A GRN is a neural network structure that combines a gating mechanism (GLU) with residual connections to enhance the model's adaptability to diverse data distributions. By controlling the flow of information through a dynamic gating mechanism, valid information can be retained and redundancy reduced. GLUs are gated linear units, which enable the network to dynamically adapt to diverse datasets.

[0051] Specifically, the calculation process of the GLU layer is expressed as:

[0052] GLU(x)=σ(w1x+b1)⊙(w2x+b2)

[0053] Where x is the input running data; w1 represents the first weight matrix, which is used to perform a linear transformation on the input x to generate a gating signal; b1 represents the first bias term, which is used in conjunction with w1 to adjust the baseline of the gating signal; w2 represents the second weight matrix, which is used to perform a linear transformation on the input x to generate candidate information; b2 represents the bias term, which is used in conjunction with w2 to adjust the offset of the candidate information; σ represents the Sigmoid activation function; ⊙ represents the element-by-element dot product.

[0054] In practical applications, σ(w1x+b1) represents the gating signal, and (w2x+b2) represents the candidate information. The element-by-element dot product of the two is to dynamically weight the candidate information using the gating signal. Here, it can be assumed that if the gating information is 0, the corresponding x is completely removed; if the gating information is 1, the corresponding x is completely retained.

[0055] like Figure 3 As shown, determining the factor weight corresponding to each key influencing factor based on VSN includes: for the key influencing factor marked as 1, the corresponding operating data After the fully connected layer (Linear), it can be converted into features For the operating data corresponding to the key influencing factor identified as n After the fully connected layer (Linear), it can be converted into features Based on this, all transformed features can be concatenated into a vector It is used to represent the state of the boiler distributed control system at time t. From this, the factor weights corresponding to each key influencing factor can be generated through the GRN and Softmax layers:

[0056] W t =Softmax(GRN W (E t ))

[0057] Where, It is the factor weight set of n running data at time t, and w is an identifier with no specific meaning.

[0058] In some embodiments, a target key influencing factor is determined from multiple key influencing factors based on the factor weight corresponding to each key influencing factor. Specifically, the K-means clustering algorithm can be combined with the elbow analysis method to divide the factor weights corresponding to multiple key influencing factors into different modes; then, the key influencing factors under each mode can be sorted according to the factor weight; finally, based on the sorting results, the key influencing factors that have a key impact on the carbon content of fly ash are screened out. It should be understood that K-means clustering here is an unsupervised learning algorithm that divides the data into k clusters, and the data points are assigned based on minimizing the distance to the cluster center. It is used to analyze the importance pattern of features and time steps, thereby optimizing the input feature space. The elbow method is a method for selecting the number of clusters, which searches for the "elbow point" as the optimal number of clusters based on the rate of change of the squared error within the cluster.

[0059] In some embodiments, each key influencing factor has a corresponding factor weight, and a higher factor weight indicates a greater impact of the key influencing factor on the carbon content of fly ash. Based on this, the "K-means clustering algorithm" can be used to classify these key influencing factors according to the size of the factor weight. For example, they can be divided into two categories: "high impact mode" and "low impact mode". At the same time, a feature importance threshold δ1 is set. Then, by comparing the rankings of key influencing factors in different categories, the key influencing factors with a ranking greater than the feature importance threshold can be retained, and the retained key influencing factors can be determined as target key influencing factors.

[0060] Based on this, we used the K-means clustering algorithm combined with the elbow analysis method to divide the factor weight vector into different modes, set a preset factor weight δ1, and screened out key influencing factors with a weight greater than δ1 as target key influencing factors. In this way, redundant or irrelevant variables (such as coal quality and other parameters that are difficult to measure in real time) can be eliminated from the high-dimensional input features, achieving the goal of reducing the dimension of the feature space and improving model efficiency.

[0061] S1034: Determine the time weight corresponding to each time using MHA, and determine the target time from multiple time periods according to the time weight corresponding to each time period.

[0062] Figure 4 This is a schematic diagram of the structure of an MHA provided by an exemplary embodiment of the present disclosure. Figure 4 As shown, it includes Query (Q), which represents the query vector, which is used to initiate a query on the data in the attention mechanism, and determines the degree of attention to different value vectors by calculating the similarity with the key vector; Key (K) key vector, which calculates the correlation with the query vector to help determine the weight of the query vector to each value vector; Value (V) value vector, which is the vector that actually participates in the calculation and provides information. According to the weight calculated by the query vector and the key vector, the value vector is weighted and summed to obtain the output of the attention mechanism. Here, for time t, Q can be the query vector corresponding to the time series feature set at time t, K represents the time series feature set corresponding to time t, and Q represents the multiple running data included in K corresponding to time t, where t is any time.

[0063] like Figure 4As shown in the figure, Softmax represents an activation function, which is used to convert the input data into a probability distribution and is used to calculate the attention weight so that the weight value is between 0 and 1 and the sum is 1; Mask represents a mask operation, which is used to shield the calculation of certain positions, such as shielding the padding part when processing a sequence to prevent it from participating in the calculation; Scale represents a scaling operation, which prevents the dot product result from being too large and causing Softmax to enter the gradient vanishing region; ScaledDot-Product Attention represents a scaled dot product attention mechanism, which calculates the dot product of the query vector and the key vector and scales it, and then obtains the weight by Softmax, and weighted sums the value vector; Linear represents a linear transformation layer, which performs a linear transformation on the input through matrix multiplication and bias addition to change the data dimension; Concat represents a splicing operation, which connects multiple vectors in a certain way.

[0064] In some embodiments, at time t, the input Q, K, and V are linearly transformed and enter the scaled dot-product attention module. Then, the product of Q and K can be scaled first, and then the mask and softmax operations are performed to obtain the time weight corresponding to time t. i represents the index of the attention head; then, Multiply by V. Here, The i in represents the number of the attention head. Based on this, MHA sets h attention heads in parallel, each of which independently calculates the attention weight. In this case, i = 1, 2, ... h, and different i correspond to different attention heads. Each attention head can capture the correlation information between the key influencing factors of the input from different subspaces. Finally, the information of each head is integrated, thereby improving the model's ability to capture and express complex information.

[0065] Among them, MHA can use Q and K to calculate the time weight corresponding to each moment, and use it to weight V to obtain the output considering the entire input time domain. The calculation process is as follows:

[0066] Attention(Q,K,V)=A(Q,K)V

[0067] Where Attention is the attention mechanism, A(·) is the normalization function, and Scaled Dot-Product Attention is used. The calculation formula is as follows:

[0068]

[0069] Among them, d attnRepresents the attention dimension. Scaled Dot-Product Attention is a method for calculating attention weights. It calculates the correlation between the query vector Q and the key vector K through the dot product and scales and normalizes the result.

[0070] In some embodiments, after determining the time weights corresponding to different time moments, a preset time weight δ2 can be set, and then the time moment with a time weight greater than δ2 is selected as the target time moment. Based on this, the target time moment and the corresponding operating data of the key influencing factors can be selected using the factor weights and time weights to optimize and reduce the feature space.

[0071] Specifically, the factor weights generated by VSN can be used to measure the degree of influence of the input key influencing factors on the fly ash carbon content. By setting a threshold, target key influencing factors with weights greater than the threshold can be screened out. At the same time, the moment weights calculated based on MHA can be used to quantify the dynamic impact of the moment. Through similar clustering and threshold setting, target moments that contribute significantly to the current prediction can be screened out. The combination of the two optimizes the feature space from both the "variable dimension" and the "time dimension": retaining the operating data of the target moment and its corresponding target key influencing factor, eliminating low-weight redundant variables and irrelevant moments, thereby reducing the model input dimension, reducing the risk of overfitting, and improving the accuracy and interpretability of fly ash carbon content prediction.

[0072] S1035 , training a preset fly ash carbon content prediction model according to a plurality of key data corresponding to the target time to obtain a target fly ash carbon content prediction model.

[0073] In some embodiments, after filtering out the operating data of the target time and its corresponding target key influencing factors based on VSN and MHA, an optimized feature space can be formed based on the operating data of the target time and its corresponding target key influencing factors; then, the operating data in this feature space is used to train a preset fly ash carbon content prediction model. By adjusting the model parameters, the preset fly ash carbon content prediction model focuses on the key variables and their dynamic hysteresis effects that significantly affect the prediction results, and finally obtains the target fly ash carbon content prediction model. In this way, by double screening and eliminating redundant information, the ability to capture the dynamic characteristics of the boiler combustion system is improved, and a more accurate real-time prediction of the fly ash carbon content can be achieved.

[0074] Based on this, the embodiment of the present disclosure provides a method for predicting the carbon content in fly ash. First, a time series feature set is constructed, and operating data corresponding to multiple key influencing factors at multiple moments are introduced to capture the delay and inertia dynamic characteristics of the boiler combustion system. Second, the factor weights corresponding to the key influencing factors are calculated using a variable selection network, and the moment weights corresponding to each moment are extracted through a multi-head attention mechanism. In this way, the degree of influence of different key influencing factors and moments on the carbon content in fly ash can be quantified, and weight analysis of key influencing factors and time series can be achieved. Then, combining clustering algorithms and elbow analysis methods, the factor weights and moment weights are clustered, and target key influencing factors and target moments with weight values ​​higher than a preset threshold are screened out, thereby optimizing the key nodes of the feature space and time series. Finally, the target operating data corresponding to the target key influencing factors at the target moment can be used to train the preset fly ash carbon content prediction model to obtain the target fly ash carbon content prediction model. Based on this, high-precision real-time prediction of fly ash carbon content and identification of key influencing factors can be achieved, solving the prediction problems caused by the lack of interpretability of traditional "black box" models and system inertia lag.

[0075] The method for predicting the carbon content in fly ash provided by the embodiment of the present disclosure can be executed by a terminal or by a chip applied to the terminal.

[0076] Exemplarily, the above-mentioned terminals may include one or more of mobile phones, tablet computers, wearable devices, vehicle-mounted devices, laptop computers, ultra-mobile personal computers (UMPCs), netbooks, personal digital assistants (PDAs), and wearable devices based on augmented reality (AR) and / or virtual reality (VR) technologies, and may also include but not be limited to smart terminals such as remote control devices, wearable devices, street lights, and home appliances. The embodiments of the present disclosure do not impose specific restrictions on this.

[0077] Figure 5 A schematic diagram of a method for predicting carbon content in fly ash provided by an exemplary embodiment of the present disclosure. Figure 5 As shown in Figure 2, the methods for predicting fly ash carbon content include:

[0078] S501, determine multiple key influencing factors that affect the carbon content of fly ash from coal-fired boilers.

[0079] In some embodiments, key influencing factors may include coal feeder coal rate, secondary air door opening of each layer, ember air door opening, economizer outlet oxygen setting value, unit operating load and fly ash carbon content.

[0080] S502: At a target time, a plurality of target operating data of the coal-fired boiler corresponding to a plurality of key influencing factors are obtained.

[0081] In some embodiments, real-time boiler operating data during the combustion process can be directly acquired through a boiler distributed control system. Specifically, the boiler distributed control system can acquire target operating data generated by the boiler during the combustion process at a target moment through sensors distributed across key locations of the boiler, such as flow sensors, temperature sensors, and damper opening sensors.

[0082] S503: Input multiple target operation data into a pre-trained target fly ash carbon content prediction model to obtain the target fly ash carbon content corresponding to the target time.

[0083] Based on this, the embodiment of the present disclosure determines multiple key influencing factors that affect the fly ash carbon content of a coal-fired boiler; at the target moment, multiple target operating data corresponding to the multiple key influencing factors of the coal-fired boiler are obtained; the multiple target operating data are input into a pre-trained target fly ash carbon content prediction model to obtain the target fly ash carbon content corresponding to the target moment. In this way, redundant influencing factors can be eliminated, thereby reducing the subsequent calculation complexity of the key influencing factors; at the same time, multiple target operating data of multiple key influencing factors are obtained at the target moment, which can accurately capture the temporal dynamic characteristics of the combustion process and avoid interference from irrelevant time data; finally, multiple target operating data are input into a pre-trained target fly ash carbon content prediction model, which can make the target fly ash carbon content prediction model adaptive to different working conditions, thereby improving the prediction accuracy and generalization ability of the fly ash carbon content, thereby helping thermal power plants optimize combustion parameters, reduce carbon losses and power generation costs, and achieve a dual improvement in combustion efficiency and economic benefits.

[0084] In some embodiments, multiple target operating data are input into a pre-trained target fly ash carbon content prediction model to obtain the target fly ash carbon content corresponding to the target moment. The method also includes: using the target fly ash carbon content prediction model to determine the factor weight corresponding to each key influencing factor; and according to the factor weight corresponding to each key influencing factor, determining the target key influencing factor from multiple key influencing factors; using the target fly ash carbon content prediction model, according to the target operating data corresponding to the target key influencing factor, determining the target fly ash carbon content corresponding to the target moment.

[0085] Specifically, the target fly ash carbon content prediction model may include a variable selection network (VSN), and then the VSN may be used to analyze the degree of influence of key influencing factors and fly ash carbon content, and obtain the factor weight of each key influencing factor, thereby quantifying the contribution of the key influencing factors to the prediction results. Based on this, a threshold can be set based on the factor weight to screen out the target key influencing factors that have a significant impact on the fly ash carbon content, eliminate redundant features, and reduce the input dimension of the target fly ash carbon content prediction model. In this way, only the operating data of the target key influencing factors is used for prediction, which can focus on the core influencing factors and improve the computational efficiency and prediction accuracy of the target fly ash carbon content prediction model. Here, for the specific calculation process of the factor weight, please refer to the relevant content of step S1033 above.

[0086] In some embodiments, a plurality of target operating data are input into a pre-trained target fly ash carbon content prediction model to obtain the target fly ash carbon content corresponding to the target moment, including: constructing an operating database, wherein the operating database includes a plurality of data sets, and one data set includes a plurality of operating data corresponding to a plurality of key influencing factors at one moment; constructing a training set according to the operating database; and using the training set to train the preset fly ash carbon content prediction model to obtain the target fly ash carbon content prediction model.

[0087] Specifically, an operation database containing multiple operation data corresponding to multiple moments and multiple key influencing factors can be constructed. Then, a certain amount of operation data can be divided from the operation database as a training set to drive model training, so that the preset fly ash carbon content prediction model can learn the mapping relationship between the key influencing factors in the combustion process and the fly ash carbon content; finally, through the data-driven training process, the preset fly ash carbon content prediction model is given the ability to predict the fly ash carbon content at the target moment, so as to obtain the target fly ash carbon content prediction model. This process can enhance the generalization ability of the model by relying on large-scale historical data, improve training efficiency through structured data organization, lay a data foundation for accurate prediction, and is suitable for dynamic characteristic modeling and engineering applications of boiler combustion systems.

[0088] In some embodiments, the target fly ash carbon content prediction model includes a data screening module and a time series screening module, and constructs a training set based on an operating database, including: using the data screening module to determine the factor weight corresponding to each key influencing factor, and determining a target key influencing factor from multiple key influencing factors based on the factor weights corresponding to multiple key influencing factors; using the time series screening module to determine the time weights corresponding to multiple moments based on the moments corresponding to multiple data sets, and determining multiple target moments from multiple moments based on the time weights corresponding to the multiple moments; and constructing a training set based on the operating data corresponding to the multiple key influencing factors at the multiple target moments. Here, the data screening module can be a variable selection network, and the time series screening module can be a multi-head attention mechanism.

[0089] Specifically, the input data of the preset fly ash carbon content prediction model may include multiple operating data corresponding to key influencing factors at different times. Therefore, in order to simplify the subsequent data processing, VSN can be used to screen the key influencing factors and select the target key influencing factors that meet specific conditions.

[0090] like Figure 3 As shown, using the data screening module to determine the factor weight corresponding to each key influencing factor includes: for the key influencing factor marked as 1, the corresponding operating data After the fully connected layer (Linear), it can be converted into features For the operating data corresponding to the key influencing factor identified as n After the fully connected layer (Linear), it can be converted into features Based on this, all transformed features can be concatenated into a vector It is used to represent the state of the boiler distributed control system at time t. From this, the factor weights corresponding to each key influencing factor can be generated through the GRN and Softmax layers:

[0091] W t =Softmax(GRN W (E t ))

[0092] Where, It is the factor weight set of n running data at time t, and w is an identifier with no specific meaning.

[0093] In some embodiments, based on the factor weights corresponding to multiple key influencing factors, a target key influencing factor is determined from multiple key influencing factors. Specifically, the factor weights corresponding to multiple key influencing factors can be divided into different modes by using the K-means clustering algorithm combined with the elbow analysis method; then, the key influencing factors in each mode can be sorted according to the factor weights; finally, based on the sorting results, the key influencing factors that have a key impact on the carbon content of fly ash are screened out. It should be understood that K-means clustering here is an unsupervised learning algorithm that divides the data into k clusters, and the data points are assigned based on minimizing the distance to the cluster center. It is used to analyze the importance pattern of features and time steps, thereby optimizing the input feature space. The elbow method is a method for selecting the number of clusters, which searches for the "elbow point" as the optimal number of clusters based on the rate of change of the squared error within the cluster.

[0094] In some embodiments, key influencing factors whose factor weights are greater than or equal to preset factor weights can be determined as target key influencing factors. Specifically, each key influencing factor has a corresponding factor weight, and the higher the factor weight, the greater the impact of this key influencing factor on the carbon content of fly ash. Based on this, the "K-means clustering algorithm" can be used to classify these key influencing factors according to the size of the factor weights. For example, they can be divided into two categories: "high impact mode" and "low impact mode", and a preset factor weight δ1 can be set at the same time. Then, by comparing the rankings of key influencing factors in different categories, the key influencing factors in the rankings that are greater than the feature importance threshold can be retained, and the retained key influencing factors can be determined as target key influencing factors.

[0095] Based on this, we used the K-means clustering algorithm combined with the elbow analysis method to divide the factor weight vector into different modes. We set a threshold of δ1 and screened out key influencing factors with weights greater than δ1 as target key influencing factors. This way, we can eliminate redundant or irrelevant variables (such as parameters that are difficult to measure in real time, such as coal quality) from high-dimensional input features, thus reducing the dimension of the feature space and improving model efficiency.

[0096] Specifically, MHA can determine the target moment from multiple moments, and then analyze the operating data of the target key influencing factors corresponding to the target moment. By integrating the operating data corresponding to each moment, the dependency relationship between the key operating data corresponding to each moment can be captured.

[0097] like Figure 4 As shown in Figure 2, at time t, the input Q, K, and V are linearly transformed and enter the scaled dot-product attention module. Then, the product of Q and K can be scaled first, and then the mask and softmax operations are performed to obtain the time weight corresponding to time t. i represents the index of the attention head; then, Multiply by V. Here, The i in represents the number of the attention head. Based on this, MHA sets h attention heads in parallel, each of which independently calculates the attention weight. In this case, i = 1, 2, ... h, and different i correspond to different attention heads. Each attention head can capture the correlation information between the key influencing factors of the input from different subspaces. Finally, the information of each head is integrated, thereby improving the model's ability to capture and express complex information.

[0098] Among them, MHA can use Q and K to calculate the time weight corresponding to each moment, and use it to weight V to obtain the output considering the entire input time domain. The calculation process is as follows:

[0099] Attention(Q,K,V)=A(Q,K)V

[0100] Where Attention is the attention mechanism, A(·) is the normalization function, and Scaled Dot-Product Attention is used. The calculation formula is as follows:

[0101]

[0102] Among them, d attn Represents the attention dimension. Scaled Dot-Product Attention is a method for calculating attention weights. It calculates the correlation between the query vector Q and the key vector K through the dot product and scales and normalizes the result.

[0103] In some embodiments, multiple target moments are determined from the multiple moments according to moment weights corresponding to the multiple moments, including: determining a moment whose moment weight is greater than or equal to a preset moment weight as the target moment.

[0104] Specifically, after determining the time weights corresponding to different moments, a preset time weight δ2 can be set, and then the moment with a time weight greater than δ2 is selected as the target moment. Based on this, the target moment and the operating data of the corresponding key influencing factors can be used to optimize and reduce the feature space.

[0105] Specifically, the factor weights generated by VSN can be used to measure the degree of influence of the input key influencing factors on the fly ash carbon content. By setting a threshold, target key influencing factors with weights greater than the threshold can be screened out. At the same time, the moment weights calculated based on MHA can be used to quantify the dynamic impact of the moment. Through similar clustering and threshold setting, target moments that contribute significantly to the current prediction can be screened out. The combination of the two optimizes the feature space from both the "variable dimension" and the "time dimension": retaining the operating data of the target moment and its corresponding target key influencing factor, eliminating low-weight redundant variables and irrelevant moments, thereby reducing the model input dimension, reducing the risk of overfitting, and improving the accuracy and interpretability of fly ash carbon content prediction.

[0106] In some embodiments, after filtering out the operating data of the target time and its corresponding target key influencing factors based on VSN and MHA, an optimized feature space can be formed based on the operating data of the target time and its corresponding target key influencing factors; then, the operating data in this feature space is used to train a preset fly ash carbon content prediction model. By adjusting the model parameters, the preset fly ash carbon content prediction model focuses on the key variables and their dynamic hysteresis effects that significantly affect the prediction results, and finally obtains the target fly ash carbon content prediction model. In this way, by double screening and eliminating redundant information, the ability to capture the dynamic characteristics of the boiler combustion system is improved, and a more accurate real-time prediction of the fly ash carbon content can be achieved.

[0107] Based on this, the embodiment of the present disclosure provides a method for predicting the carbon content in fly ash. First, a time series feature set is constructed, and operating data corresponding to multiple key influencing factors at multiple moments are introduced to capture the delay and inertia dynamic characteristics of the boiler combustion system. Second, the factor weights corresponding to the key influencing factors are calculated using a variable selection network, and the moment weights corresponding to each moment are extracted through a multi-head attention mechanism. In this way, the degree of influence of different key influencing factors and moments on the carbon content in fly ash can be quantified, and weight analysis of key influencing factors and time series can be achieved. Then, combining clustering algorithms and elbow analysis methods, the factor weights and moment weights are clustered, and target key influencing factors and target moments with weight values ​​higher than a preset threshold are screened out, thereby optimizing the key nodes of the feature space and time series. Finally, the target operating data corresponding to the target key influencing factors at the target moment can be used to train the preset fly ash carbon content prediction model to obtain the target fly ash carbon content prediction model. Based on this, high-precision real-time prediction of fly ash carbon content and identification of key influencing factors can be achieved, solving the prediction problems caused by the lack of interpretability of traditional "black box" models and system inertia lag.

[0108] The above mainly introduces the solution provided by the embodiment of the present disclosure from the perspective of the server. It can be understood that in order to realize the above functions, the server includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present disclosure can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present disclosure.

[0109] The embodiments of the present disclosure can divide the server into functional units according to the above method examples. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one management module. The above integrated modules can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiments of the present disclosure is schematic and is only a logical functional division. In actual implementation, there may be other division methods.

[0110] In the case of dividing each functional module according to each function, an exemplary embodiment of the present disclosure provides a fly ash carbon content prediction device, which may be a server or a chip applied to a server. Figure 6 This is a schematic diagram of the structure of a fly ash carbon content prediction device provided by one embodiment of the present disclosure. Figure 6 As shown, the fly ash carbon content prediction device 600 includes:

[0111] Determination module 601, for determining a plurality of key influencing factors affecting the carbon content of fly ash of a coal-fired boiler;

[0112] An acquisition module 602 is configured to acquire, at a target time, a plurality of target operating data of the coal-fired boiler corresponding to the plurality of key influencing factors;

[0113] The acquisition module 603 is used to input the plurality of target operation data into a pre-trained target fly ash carbon content prediction model to obtain the target fly ash carbon content corresponding to the target moment.

[0114] In an optional manner, the fly ash carbon content prediction device 600 also includes a screening module 604, which is used to use the target fly ash carbon content prediction model to determine the factor weight corresponding to each of the key influencing factors; and based on the factor weight corresponding to each of the key influencing factors, determine the target key influencing factor from multiple key influencing factors; use the target fly ash carbon content prediction model to determine the target fly ash carbon content corresponding to the target moment based on the target operating data corresponding to the target key influencing factor.

[0115] In an optional manner, the acquisition module 603 is also used to construct an operation database, wherein the operation database includes multiple data sets, and one data set includes multiple operation data corresponding to multiple key influencing factors at a moment; a training set is constructed according to the operation database; and the preset fly ash carbon content prediction model is trained using the training set to obtain the target fly ash carbon content prediction model.

[0116] In an optional manner, the target fly ash carbon content prediction model includes a data screening module and a time series screening module, and the acquisition module 603 is also used to use the data screening module to determine the factor weight corresponding to each of the key influencing factors, and determine the target key influencing factor from the multiple key influencing factors based on the factor weights corresponding to the multiple key influencing factors; use the time series screening module to determine the moment weights corresponding to the multiple moments according to the moments corresponding to the multiple data sets, and determine the multiple target moments from the multiple moments according to the moment weights corresponding to the multiple moments; construct a training set based on the operating data corresponding to the multiple key influencing factors at the multiple target moments.

[0117] In an optional manner, the obtaining module 603 is further configured to determine the key influencing factor whose factor weight is greater than or equal to a preset factor weight as the target key influencing factor.

[0118] In an optional manner, the obtaining module 603 is further configured to determine the moment whose moment weight is greater than or equal to a preset moment weight as the target moment.

[0119] An embodiment of the present disclosure further provides an electronic device, comprising: at least one processor; a memory for storing instructions executable by at least one processor; wherein the at least one processor is configured to execute instructions to implement the steps of the above method disclosed in the embodiment of the present disclosure.

[0120] Figure 7 This is a schematic diagram of the structure of an electronic device provided by an embodiment of the present disclosure. Figure 7As shown, the electronic device 700 includes at least one processor 701 and a memory 702 coupled to the processor 701. The processor 701 can execute corresponding steps in the above method disclosed in the embodiment of the present disclosure.

[0121] The processor 701 can also be referred to as a central processing unit (CPU), which can be an integrated circuit chip with signal processing capabilities. Each step in the above method disclosed in the embodiment of the present disclosure can be completed by the hardware integrated logic circuit in the processor 701 or by instructions in the form of software. The processor 701 can be a general-purpose processor, a digital signal processor (DSP), an ASIC, a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiment of the present disclosure can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in the memory 702, such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, or other mature storage media in the art. The processor 701 reads the information in the memory 702 and completes the steps of the above method in combination with its hardware.

[0122] In addition, when various operations / processes according to the present disclosure are implemented by software and / or firmware, they can be transferred from a storage medium or a network to a computer system having a dedicated hardware structure, for example, Figure 8 The computer system 800 shown is installed with the programs constituting the software. When the various programs are installed, the computer system can perform various functions, including the functions described above. Figure 8 A schematic diagram of the structure of a computer system provided in one embodiment of the present disclosure.

[0123] Computer system 800 is intended to represent various forms of digital electronic computer equipment, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic equipment can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0124] like Figure 8 As shown, the computer system 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. Various programs and data required for the operation of the computer system 800 can also be stored in the RAM 803. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0125] Multiple components within computer system 800 are connected to I / O interface 805, including an input unit 806, an output unit 807, a storage unit 808, and a communication unit 809. Input unit 806 can be any type of device capable of inputting information into computer system 800. Input unit 806 can receive input numeric or character information and generate key input signals related to user settings and / or function control of an electronic device. Output unit 807 can be any type of device capable of presenting information and may include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. Storage unit 808 may include, but is not limited to, a magnetic disk or an optical disk. Communication unit 809 allows computer system 800 to exchange information / data with other devices over a network, such as the Internet, and may include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver and / or chipset, such as a Bluetooth™ device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0126] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units for running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 801 performs the various methods and processes described above. For example, in some embodiments, the above-mentioned method disclosed in the embodiments of the present disclosure may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on an electronic device via ROM 802 and / or communication unit 809. In some embodiments, the computing unit 801 may be configured to perform the above-mentioned method disclosed in the embodiments of the present disclosure by any other appropriate means (e.g., by means of firmware).

[0127] An embodiment of the present disclosure further provides a computer-readable storage medium, wherein when instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the above method disclosed in the embodiment of the present disclosure.

[0128] The computer-readable storage medium in the embodiments of the present disclosure can be a tangible medium that can contain or store a program for use by an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. The above-mentioned computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specifically, the above-mentioned computer-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0129] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0130] Figure 9 Schematic diagram of a computer program product provided by an embodiment of the present disclosure. Figure 9 As shown, the computer program product 900 includes a computer program 901 , wherein the computer program 901 implements the above method disclosed in the embodiment of the present disclosure when executed by a processor.

[0131] In embodiments of the present disclosure, computer program code for performing the operations of the present disclosure may be written in one or more programming languages ​​or combinations thereof, including but not limited to object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer.

[0132] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0133] The modules, components, or units described in the embodiments of the present disclosure may be implemented in software or hardware. The names of the modules, components, or units do not necessarily limit the modules, components, or units themselves.

[0134] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, and without limitation, exemplary hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0135] The above descriptions are merely some embodiments of the present disclosure and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but also encompasses other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned disclosed concepts. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in the present disclosure.

[0136] Although some specific embodiments of the present disclosure have been described in detail by way of examples, those skilled in the art will appreciate that the above examples are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Those skilled in the art will appreciate that modifications may be made to the above embodiments without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A method for predicting carbon content in fly ash, characterized in that: include: Identify the key factors that influence the carbon content of fly ash from coal-fired boilers; At a target time, obtaining a plurality of target operating data of the coal-fired boiler corresponding to a plurality of the key influencing factors; The plurality of target operation data are input into a pre-trained target fly ash carbon content prediction model to obtain the target fly ash carbon content corresponding to the target time.

2. The method according to claim 1, characterized in that The method further comprises: inputting the plurality of target operation data into a pre-trained target fly ash carbon content prediction model to obtain the target fly ash carbon content corresponding to the target time. Determining the factor weight corresponding to each of the key influencing factors using the target fly ash carbon content prediction model; and determining a target key influencing factor from the plurality of key influencing factors according to the factor weight corresponding to each key influencing factor; The target fly ash carbon content prediction model is used to determine the target fly ash carbon content corresponding to the target time according to the target operation data corresponding to the target key influencing factors.

3. The method according to claim 1, characterized in that Inputting the plurality of target operation data into a pre-trained target fly ash carbon content prediction model to obtain the target fly ash carbon content corresponding to the target time comprises: Constructing an operation database, wherein the operation database includes multiple data sets, and one data set includes multiple operation data corresponding to multiple key influencing factors at a time; constructing a training set based on the running database; The training set is used to train a preset fly ash carbon content prediction model to obtain the target fly ash carbon content prediction model.

4. The method according to claim 3, characterized in that The target fly ash carbon content prediction model includes a data screening module and a time series screening module. The training set is constructed according to the operation database, including: Determining the factor weight corresponding to each of the key influencing factors using the data screening module, and determining a target key influencing factor from the multiple key influencing factors based on the factor weights corresponding to the multiple key influencing factors; Utilizing the time series screening module, determining time weights corresponding to a plurality of time moments according to the time moments corresponding to the plurality of data sets, and determining a plurality of target time moments from the plurality of time moments according to the time weights corresponding to the plurality of time moments; A training set is constructed according to the operating data corresponding to the multiple key influencing factors at the multiple target moments.

5. The method according to claim 4, characterized in that Determining a target key influencing factor from the plurality of key influencing factors according to the factor weights corresponding to the plurality of key influencing factors includes: The key influencing factor whose factor weight is greater than or equal to the preset factor weight is determined as the target key influencing factor.

6. The method according to claim 4, characterized in that The step of determining a plurality of target moments from the plurality of moments according to the moment weights corresponding to the plurality of moments includes: The moment whose moment weight is greater than or equal to a preset moment weight is determined as the target moment.

7. A device for predicting carbon content in fly ash, characterized in that: include: A determination module is used to determine multiple key influencing factors affecting the carbon content of fly ash from coal-fired boilers; An acquisition module, configured to acquire, at a target time, a plurality of target operating data of the coal-fired boiler corresponding to the plurality of key influencing factors; The acquisition module is used to input the plurality of target operation data into a pre-trained target fly ash carbon content prediction model to obtain the target fly ash carbon content corresponding to the target moment.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory, wherein: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images