Boiler thermal efficiency prediction method, device, equipment, storage medium and program product

Abstract

The invention relates to a boiler thermal efficiency prediction method and device, equipment, a storage medium and a program product. The method comprises the steps of obtaining operation data of a target boiler, and constructing input sample data based on the operation data; the operation data is used for representing the heat efficiency of the target boiler; carrying out variational mode decomposition processing on the input sample data to obtain a plurality of mode components; for each modal component, performing multi-time scale feature extraction on the modal component to obtain time scale features respectively corresponding to the modal component under a plurality of time scales; time scale feature fusion results under all time scales are arranged in a directed mode, and time sequence features are obtained; and inputting the time sequence characteristics into a long short-term memory network, and performing time sequence prediction to obtain a thermal efficiency prediction result of the target boiler. By adopting the method, the accuracy of predicting the thermal efficiency of the boiler can be improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, equipment, storage medium, and program product for predicting boiler thermal efficiency. Background Technology

[0002] As a core thermal energy device in industrial production, the thermal efficiency of boilers directly affects energy consumption, production costs, and environmental emissions. Accurate prediction of boiler thermal efficiency is crucial for ensuring that boiler combustion meets emission requirements and for reducing production costs. However, current methods for predicting boiler thermal efficiency still have relatively low accuracy. Summary of the Invention

[0003] Therefore, it is necessary to provide a boiler thermal efficiency prediction method, apparatus, equipment, storage medium, and program product that can improve the accuracy of predicting boiler thermal efficiency in response to the above-mentioned technical problems.

[0004] In a first aspect, this application provides a method for predicting boiler thermal efficiency, including:

[0005] Obtain the operating data of the target boiler, and construct input sample data based on the operating data; the operating data is used to characterize the thermal efficiency of the target boiler.

[0006] The input sample data is subjected to variational mode decomposition to obtain multiple modal components;

[0007] For each modal component, multi-timescale feature extraction is performed on the modal component to obtain the timescale features corresponding to the modal component at multiple timescales.

[0008] The time-scale feature fusion results at each of the stated time scales are oriented to obtain time-series features;

[0009] The time series features are input into a long short-term memory network for time series prediction to obtain the predicted thermal efficiency of the target boiler.

[0010] In one embodiment, constructing input sample data based on the runtime data includes:

[0011] Extract the real-time running data from the running data, as well as the historical running data within a preset time step before the real-time running data;

[0012] Input sample data is constructed based on the real-time operation data and the historical operation data.

[0013] In one embodiment, the variational mode decomposition process on the input sample data to obtain multiple modal components includes:

[0014] Determine the number of target modes K; K is a natural number;

[0015] With the goal of minimizing the sum of modal bandwidths, modal components corresponding to K modes are extracted from the input sample data.

[0016] In one embodiment, the method further includes:

[0017] Obtain the decomposition information of each modal component;

[0018] If, in the case where each of the decomposed information represents a first modal component in which modal aliasing exists, the number of target modes is increased;

[0019] In cases where the decomposed information characterizes a second modal component with a redundant state in each modal component, the number of target modes is reduced.

[0020] Return to the step of extracting the modal components corresponding to K modes from the input sample data with the goal of minimizing the sum of modal bandwidths, until there are no modal components in the modal components that are in a modal aliasing state or a redundant state.

[0021] In one embodiment, the step of performing multi-timescale feature extraction on each of the modal components to obtain the timescale features corresponding to the modal components at multiple timescales includes:

[0022] The modal components are input into the feature extraction channel, and convolution processing is performed through multiple convolution kernels of different time scales in the feature extraction channel to extract the local temporal features of the modal components at each time scale.

[0023] Each of the local time features is pooled to obtain the time scale features corresponding to the modal components at multiple time scales.

[0024] In one embodiment, the step of inputting the time series features into a long short-term memory network for time series prediction to obtain the predicted thermal efficiency of the target boiler includes:

[0025] Long-term historical information in the time series features is filtered out by using the forgetting gate in the long short-term memory network.

[0026] Determine the injection of new information into the long-term historical information;

[0027] By fusing the long-term historical information and the injected new information, long-term and short-term fused information is obtained;

[0028] The combined long-term and short-term information is then converted into a predicted thermal efficiency result for the target boiler.

[0029] Secondly, this application also provides a boiler thermal efficiency prediction device, comprising:

[0030] The sample data construction module is used to acquire the operating data of the target boiler and construct input sample data based on the operating data; the operating data is used to characterize the thermal efficiency of the target boiler.

[0031] The variational mode decomposition module is used to perform variational mode decomposition processing on the input sample data to obtain multiple mode components;

[0032] The scale feature extraction module is used to extract multi-time-scale features for each modal component to obtain the time-scale features corresponding to the modal component at multiple time scales.

[0033] The scale feature fusion module is used to arrange the time scale feature fusion results under each time scale in a directed manner to obtain time series features.

[0034] The prediction result output module is used to input the time series features into a long short-term memory network for time series prediction to obtain the predicted thermal efficiency of the target boiler.

[0035] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0036] Obtain the operating data of the target boiler, and construct input sample data based on the operating data; the operating data is used to characterize the thermal efficiency of the target boiler.

[0037] The input sample data is subjected to variational mode decomposition to obtain multiple modal components;

[0038] For each modal component, multi-timescale feature extraction is performed on the modal component to obtain the timescale features corresponding to the modal component at multiple timescales.

[0039] The time-scale feature fusion results at each of the stated time scales are oriented to obtain time-series features;

[0040] The time series features are input into a long short-term memory network for time series prediction to obtain the predicted thermal efficiency of the target boiler.

[0041] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0042] Obtain the operating data of the target boiler, and construct input sample data based on the operating data; the operating data is used to characterize the thermal efficiency of the target boiler.

[0043] The input sample data is subjected to variational mode decomposition to obtain multiple modal components;

[0044] For each modal component, multi-timescale feature extraction is performed on the modal component to obtain the timescale features corresponding to the modal component at multiple timescales.

[0045] The time-scale feature fusion results at each of the stated time scales are oriented to obtain time-series features;

[0046] The time series features are input into a long short-term memory network for time series prediction to obtain the predicted thermal efficiency of the target boiler.

[0047] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0048] Obtain the operating data of the target boiler, and construct input sample data based on the operating data; the operating data is used to characterize the thermal efficiency of the target boiler.

[0049] The input sample data is subjected to variational mode decomposition to obtain multiple modal components;

[0050] For each modal component, multi-timescale feature extraction is performed on the modal component to obtain the timescale features corresponding to the modal component at multiple timescales.

[0051] The time-scale feature fusion results at each of the stated time scales are oriented to obtain time-series features;

[0052] The time series features are input into a long short-term memory network for time series prediction to obtain the predicted thermal efficiency of the target boiler.

[0053] The aforementioned boiler thermal efficiency prediction method, apparatus, equipment, storage medium, and program products acquire the operating data of the target boiler and construct input sample data based on the operating data. The operating data is used to characterize the thermal efficiency of the target boiler. Variational mode decomposition is performed on the input sample data to obtain multiple modal components. For each modal component, multi-timescale feature extraction is performed to obtain the timescale features corresponding to the modal component at multiple time scales. The timescale feature fusion results at each time scale are oriented to obtain time series features. The time series features are input into a long short-term memory network for time series prediction to obtain the thermal efficiency prediction result of the target boiler. Based on the operating data used to characterize the thermal efficiency of the target boiler, variational mode decomposition can separate the multimodal information in the input sample data and eliminate the interference caused by mode aliasing. Multi-timescale feature extraction can obtain the short-term and long-term time scale features of each modal data and fuse them into time series features. Then, the time series feature capture capability of the long short-term memory network is used to perform time series prediction to obtain the thermal efficiency prediction result. This method can adapt to changes in boiler operating conditions and characteristics, thus improving the accuracy of boiler thermal efficiency prediction. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a diagram illustrating the application environment of a boiler thermal efficiency prediction method in one embodiment.

[0056] Figure 2 This is a flowchart illustrating a boiler thermal efficiency prediction method in one embodiment;

[0057] Figure 3 This is a flowchart illustrating the process of performing variational mode decomposition on input sample data to obtain multiple modal components in one embodiment.

[0058] Figure 4 This is a diagram showing the results of variational mode decomposition in one embodiment;

[0059] Figure 5 This is a flowchart illustrating the process of extracting multi-timescale features from each modal component and fusing the extracted timescale features to obtain time series features in one embodiment.

[0060] Figure 6This is a flowchart illustrating the process of inputting time series features into a long short-term memory network to perform time series prediction and obtain the predicted thermal efficiency of the target boiler in one embodiment.

[0061] Figure 7 This is a diagram of the neuronal structure of a long short-term memory network in one embodiment;

[0062] Figure 8 This is a diagram illustrating the organizational structure of a Long Short-Term Memory (LSTM) network in one embodiment.

[0063] Figure 9 This is a flowchart illustrating the boiler thermal efficiency prediction method in another embodiment;

[0064] Figure 10 This is a structural block diagram of a boiler thermal efficiency prediction device in one embodiment;

[0065] Figure 11 This is an internal structural diagram of a computer device in one embodiment;

[0066] Figure 12 This is a diagram of the internal structure of a computer device in another embodiment. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0068] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0069] In related technologies, boiler thermal efficiency prediction methods mainly include mechanistic modeling and data-driven methods. Mechanistic modeling is based on the boiler heat balance equation to construct a physical model, which requires accurate acquisition of dozens of parameters such as fuel composition, flue gas composition, and heat loss. However, parameter measurements in industrial settings are easily affected by sensor accuracy and operating condition fluctuations, and mechanistic models are difficult to adapt to dynamic changes in boiler operating conditions such as scaling and wear, resulting in large prediction errors.

[0070] The boiler thermal efficiency prediction method provided in this application embodiment can be applied to, for example... Figure 1The application environment is shown. Terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on a cloud or other network server. Terminal 102 acquires the operating data of the target boiler and constructs input sample data based on the operating data. The operating data is used to characterize the thermal efficiency of the target boiler. Variational mode decomposition is performed on the input sample data to obtain multiple modal components. For each modal component, multi-timescale feature extraction is performed to obtain the timescale features corresponding to the modal component at multiple time scales. The timescale feature fusion results at each time scale are oriented to obtain time series features. The time series features are input into a long short-term memory network for time series prediction to obtain the predicted thermal efficiency of the target boiler. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, and tablets. Server 104 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0071] In one exemplary embodiment, such as Figure 2 As shown, a method for predicting boiler thermal efficiency is provided, which is then applied to... Figure 1 The following steps are used as an example of the terminal in the example, including steps 202 to 210.

[0072] Step 202: Obtain the operating data of the target boiler and construct input sample data based on the operating data.

[0073] Boilers in industry are thermal power equipment, mostly coal-fired. Their thermal efficiency refers to the ratio of effectively utilized heat to the heat input from the fuel. The target boiler is the boiler whose thermal efficiency is to be predicted. Operating data refers to data related to the boiler's operational status. Operating data may include feedwater flow, fuel quantity, flue gas temperature, total coal consumption, total air volume, and oxygen content in the flue gas. Input sample data refers to the data used as input to the overall model during the thermal efficiency prediction process. The target boiler's operating data can be obtained in real time from various sources. For subsequent multi-timescale feature extraction and time-series prediction processes, the input data can include operating data from multiple time steps.

[0074] Operational data is used to characterize thermal efficiency. It's understandable that operational data such as water supply and fuel quantity can be measured in real-time using instruments during actual production, but thermal efficiency is difficult to measure in real-time. Thermal efficiency is typically calculated using parameters such as fuel quantity, total coal quantity, total air volume, flue gas oxygen content, and exhaust gas temperature. For example, exhaust gas temperature characterizes the magnitude of exhaust heat loss; a higher exhaust gas temperature indicates lower thermal efficiency. Similarly, flue gas oxygen content indicates whether the excess air coefficient is at its optimal level; excessively high or low oxygen content indicates that thermal efficiency has not reached its peak.

[0075] For example, the target boiler's operating data is transmitted to a database in real time for storage. Operating data for a set time period is retrieved from the database, and input sample data is constructed based on this data. When boiler thermal efficiency prediction is required, the necessary operating data can be obtained from the database. The set time period can be determined based on the actual prediction situation and is related to the long-term information that the subsequent long short-term memory network can capture.

[0076] In one embodiment, constructing input sample data based on operational data includes: extracting real-time operational data from the operational data and historical operational data within a preset time step prior to the real-time operational data; and constructing input sample data based on the real-time operational data and the historical operational data.

[0077] Real-time operating data refers to the data generated by the target boiler during its operation. Historical operating data refers to the data generated by the target boiler within a preset time step preceding the real-time operating data. The length of the preset time step can be set according to actual prediction needs. The preset time step can contain multiple shorter time periods, each corresponding to a set of historical operating data. For example, if the real-time operating data is the data at time step t, then the historical operating data includes the operating data from the N time steps preceding time step t, and the total duration of the N time steps is the preset time step. During the construction of input sample data, the real-time operating data and historical operating data can be arranged in chronological order to obtain the input sample data.

[0078] Step 204: Perform variational mode decomposition on the input sample data to obtain multiple modal components.

[0079] Variational mode decomposition (VMD) refers to the process of decomposing a signal into modal components with finite bandwidth using a variational framework. A modal component is a component within a certain frequency range separated from the input sample data. The input sample data can be considered a non-stationary time-series signal containing information from multiple frequency components. Through VMD processing, the information from different frequency components can be separated, noise interference can be filtered out, and more detailed analysis can be performed based on each modal component.

[0080] For example, a preset number of decomposed modes, bandwidth constraints, and noise tolerance are obtained. Based on the number of decomposed modes, bandwidth constraints, and noise tolerance, variational mode decomposition is performed on the input sample data to obtain multiple modal components. Among them, the number of decomposed modes limits the number of modal components obtained by decomposition, the bandwidth constraint limits the frequency range of the modal components, and the noise tolerance limits the tolerance to noise in the modal components.

[0081] Step 206: For each modal component, perform multi-timescale feature extraction on the modal component to obtain the timescale features corresponding to the modal component at multiple timescales.

[0082] Multi-timescale feature extraction refers to the operation of extracting features at multiple different time scales, and the extracted features are called time-scale features. Through multi-timescale feature extraction, the local correlation patterns of each modal component in the short, medium and long time ranges can be captured, and transient features of different durations in the input sample data can be obtained.

[0083] For example, a set time scale information is obtained, and multi-time scale feature extraction is performed on each modal component based on the time scale information to obtain the time scale features at different time scales represented by the time scale information.

[0084] Step 208: Directly arrange the time scale feature fusion results at each time scale to obtain time series features.

[0085] Time-scale feature fusion refers to the result of fusing time-scale features at the same time scale. By fusing time-scale features, relatively broad trend fluctuations in modal components can be identified, achieving a combination of multimodal and multi-scale features to obtain time-series features with high information density. Time-scale feature fusion results can be obtained by concatenating time-scale features at the same time scale, or by weighted summation of time-scale features at the same time scale. Furthermore, time-scale feature fusion results at different time scales can be arranged in a directed manner according to chronological order to obtain time-series features.

[0086] Step 210: Input the time series features into the Long Short-Term Memory network to perform time series prediction and obtain the predicted thermal efficiency of the target boiler.

[0087] Long Short-Term Memory (LSTM) networks are a type of recurrent neural network that addresses the long-term dependency problem inherent in recurrent neural networks. Thermal efficiency prediction results are forecasts of the target boiler thermal efficiency based on input operating data. By inputting time-series features into an LSM network, the network learns the temporal relationships within those features and uses the acquired contextual information to perform time-series predictions, thus obtaining the thermal efficiency prediction result. The thermal efficiency prediction result can include the predicted thermal efficiency value and the prediction confidence level.

[0088] For example, time series features are input into a long short-term memory network in chronological order, and the time series features are dynamically encoded through a gating mechanism to obtain global time series dynamic information. The global time series dynamic information is then passed to the fully connected layer and the output layer to obtain the thermal efficiency prediction result of the target boiler.

[0089] In the above-mentioned boiler thermal efficiency prediction method, the operating data of the target boiler is acquired, and input sample data is constructed based on the operating data. Variational mode decomposition is performed on the input sample data to obtain multiple modal components. Multi-timescale feature extraction is performed on each modal component, and the extracted timescale features are fused to obtain time series features. The time series features are input into a long short-term memory network for time series prediction to obtain the thermal efficiency prediction result of the target boiler. Variational mode decomposition can separate the multi-modal information in the input sample data and eliminate the interference caused by mode aliasing. Multi-timescale feature extraction can obtain the short-term and long-term timescale features of each modal data. Then, the time series feature capture capability of the long short-term memory network is used to perform time series prediction to obtain the thermal efficiency prediction result. This method can adapt to the changes in boiler operating conditions and characteristics, and improve the accuracy of boiler thermal efficiency prediction.

[0090] In one exemplary embodiment, such as Figure 3 As shown, variational mode decomposition is performed on the input sample data to obtain multiple modal components, including steps 302 to 304.

[0091] Step 302: Determine the number of target modes K.

[0092] The target number of modes refers to the number of modal components to be obtained after variational mode decomposition. During model training, rules for signal decomposition using variational mode decomposition can be configured to determine the number of modal components that yield the best results from a set range of modes, serving as the initial target number of modes in the variational mode decomposition process. K is a natural number.

[0093] Step 304: With the goal of minimizing the sum of modal bandwidths, extract the modal components corresponding to K modes from the input sample data.

[0094] The sum of modal bandwidths refers to the sum of the bandwidths of all modal components. For input sample data in time series form... Modal decomposition yields multiple modal components. This can be achieved by iteratively finding each modal component. (k=1,2,...,K, where K is the number of target modes) and their corresponding center frequencies Minimize the constrained optimization problem shown in Equations 1 and 2 below.

[0095] Formula 1: .

[0096] Formula 2: .

[0097] in, For time derivative, For the Dirac function, This is a convolution operation.

[0098] Then update using the alternating direction multiplier method. , The process continues with Lagrange multipliers until the convergence condition is met (iteration error less than 10). -7 The final output consists of K modal components. Each modal component can correspond to a frequency mode, such as high-frequency noise mode, mid-frequency operating condition mode, low-frequency trend mode, etc., to achieve multi-modal information separation.

[0099] In one embodiment, the method further includes: acquiring the decomposition information of each modal component; increasing the number of target modes when the decomposition information indicates that a first modal component in each modal component has a modal aliasing state; decreasing the number of target modes when the decomposition information indicates that a second modal component in each modal component has a redundant state; and returning to the step of extracting the modal components corresponding to K modes from the input sample data with the goal of minimizing the sum of modal bandwidths, until there are no modal components in each modal component with a modal aliasing state or a redundant state.

[0100] Decomposition information is used to characterize the state-related information of modal components after decomposition, and may include bandwidth, center frequency, noise, etc. The first modal component refers to the modal component exhibiting mode aliasing. The second modal component refers to the component with redundant states. During the deployment phase before model training, the K value can be selected and adjusted within an empirical range (e.g., 3-8) based on different data characteristics to achieve the best decomposition effect. By evaluating the decomposition effect under different K values, the K value that best separates the frequency components with different physical meanings can be selected. In the actual thermal efficiency prediction process, the initial target number of modes can be adaptively corrected and the K value adjusted according to the modal decomposition results.

[0101] Modal aliasing refers to a state where a single modal component simultaneously contains information of similar but different frequencies; for example, it may contain both high-frequency noise and mid-frequency fluctuations. The presence of modal aliasing indicates that the set number of target modes is too small, failing to adequately separate frequency components of different characteristics. Increasing the number of target modes allows for finer decomposition. The number of modes can be gradually increased within a set range. When the operating data is highly complex and rich in frequency components, it may be necessary to increase the number of modes.

[0102] Redundant modes refer to mode decomposition states where the center frequencies of multiple modal components are too close, or where modal components exhibit extremely small amplitudes that are indistinguishable from random noise. The existence of redundant modes indicates that the number of target modes is set too high, resulting in over-decomposition. The number of target modes can be reduced to extract more representative main modes, while avoiding the introduction of unnecessary noise and computational burden.

[0103] Figure 4 This is a result diagram of variational mode decomposition in one embodiment, where imf1, imf2, imf3, imf4, and imf5 represent different modal components. Figure 4 As can be seen, different modal components exhibit distinctly different characteristics. IMF1 is relatively stable overall, with a relatively slow trend of change, showing the relatively low-frequency part of the signal; while IMF2, IMF3, IMF4, and IMF5 show more frequent fluctuations. These components contain the relatively high-frequency components of the signal, reflecting information about the signal in terms of detail and rapid changes.

[0104] In this embodiment, the separation of physical meaning in the signal through variational mode decomposition can provide a more interpretable learning foundation than the original signal for subsequent multi-timescale feature extraction and long short-term memory networks. The model can extract features from different modal components more specifically.

[0105] In one exemplary embodiment, such as Figure 5 As shown, multi-timescale feature extraction is performed on the modal components to obtain the timescale features corresponding to the modal components at multiple timescales, including steps 502 to 504.

[0106] Step 502: Input the modal components into the corresponding feature extraction channels respectively, and perform convolution processing through multiple convolution kernels of different time scales in the feature extraction channels to extract the local temporal features of the modal components at each time scale.

[0107] Multi-scale convolutional neural networks can be used for multi-temporal-scale feature extraction. Feature extraction channels are the channels used for feature extraction within the convolutional neural network. Each channel can include multiple convolutional kernels of different sizes. Convolutional processing with kernels of different sizes can yield local temporal features at different time scales. In the convolutional layer, the input feature maps correspond to each modal component. With convolution kernel Features can be extracted through local weighted summation to capture local correlation patterns, as shown in Formula 3 below.

[0108] Formula 3: .

[0109] in, The kernel size is the convolution kernel size. For the number of channels, For bias.

[0110] Step 504: Perform pooling processing on each local time feature to obtain the time scale features corresponding to the modal components at multiple time scales.

[0111] Before pooling, non-linearity can be introduced through an activation function, enabling the convolutional neural network to learn complex relationships. The activation function used can be the ReLU activation function, expressed as follows: Pooling can be performed using max pooling, which downsamples the feature map, reduces the number of parameters, and enhances translation invariance. The expression for max pooling can be found in Equation 4 below.

[0112] Formula 4: .

[0113] in, This is the pooling step size.

[0114] Through convolution and pooling operations, convolutional neural networks can transform the original modal components into highly abstract, low-dimensional but more information-dense temporal features, reducing the sequence length that subsequent long short-term memory networks need to process, thereby alleviating the computational burden.

[0115] In this embodiment, multi-timescale feature extraction can achieve the combination of multimodal and multi-scale features, providing accurate data support for subsequent long short-term memory network processing.

[0116] In one exemplary embodiment, such as Figure 6 As shown, the time series features are input into the Long Short-Term Memory network for time series prediction to obtain the thermal efficiency prediction result of the target boiler, including steps 602 to 608.

[0117] Step 602: Use the forgetting gate in the Long Short-Term Memory network to filter out long-term historical information from the time series features.

[0118] Long Short-Term Memory (LSTM) networks selectively store, forget, and output information by introducing gating mechanisms and cell states, thereby effectively preserving key long-term information in long sequences while dynamically updating short-term information. Figure 7 Here is a diagram of the neuronal structure of a long short-term memory network in one embodiment. Figure 8 Here is a diagram of the organizational structure of a Long Short-Term Memory (LSTM) network in one embodiment, with reference to... Figure 7 and Figure 8 The forgetting gate filters out long-term historical information. It determines the cell state based on the current input and the short-term memory from the previous moment. To determine which long-term information should be retained and which should be forgotten, refer to Formula 5 below.

[0119] Formula 5: .

[0120] Where σ is the activation value of the forget gate; The hidden layer output of the previous time step The connection weight to the forget gate; This indicates that the previous state will be hidden. With the current input Concatenate along the feature dimension to form a dimension of Vectors; weight matrix With bias Through training, the final output forget gate vector In the diagram, the value of each element corresponds to the cell state. The retention ratio of the corresponding dimension information, with a value close to 1 indicating that the long-term information of that dimension is fully retained, and a value close to 0 indicating that it is completely forgotten.

[0121] Step 604: Determine the new information to be injected into the long-term historical information.

[0122] Injecting new information refers to injecting information about the cell state. After filtering historical information, the input gate and candidate cell states can jointly inject and filter new information. During this process, the input gate can determine which new information should be included in the cell state based on the current input and the short-term memory of the previous moment, referring to the following formula 6.

[0123] Formula 6: .

[0124] in, Similar to the forget gate, each element represents the proportion of new information included in the corresponding dimension, enabling selective filtering of new information.

[0125] Then, candidate cell state To generate the new information to be injected at the current moment, refer to the following formula 7.

[0126] Formula 7: .

[0127] The tanh activation function can be used to compress the values ​​of candidate information to [-1, 1], which avoids numerical overflow and also reflects the positive or negative contribution of information through positive or negative values. This represents all new information that can be injected into the cell at the current moment.

[0128] Step 606: Integrate long-term historical information and inject new information to obtain integrated long-term and short-term information.

[0129] Long-term and short-term fused information can be obtained by updating cell state and fusing long-term historical information with the injection of new information. Cell state The update process can be referenced in Formula 8 below.

[0130] Formula 8: .

[0131] Among them, the symbol " "" indicates element-wise multiplication. This formula achieves dimension-wise information fusion through the Hadamard product, the first term The second item represents the valid information preserved in the historical cell state. The new information represents the filtered data, and the sum of the two forms the cell state at time t. This cumulative update method allows cell states to continuously transmit key information over long sequences, avoiding long-term information loss due to gradient vanishing.

[0132] Step 608: Transform the long-term and short-term integrated information into the thermal efficiency prediction results of the target boiler.

[0133] Short-term memory can be generated using the output gate and hidden state to output the predicted thermal efficiency. (Output gate and hidden state) They can work together to convert the information in the current cell state into short-term memory, which can be used to output the thermal efficiency prediction result of the model at the current moment, or to continue to the next moment.

[0134] During the current short-term memory generation process, the output gate can determine the cell state based on the current input and the short-term memory from the previous moment. To determine which information needs to be output, refer to Formula 9 below.

[0135] Formula 9: .

[0136] Among them, the output Each element represents the output proportion of the cell state corresponding to the dimension information, enabling selective control of the output information.

[0137] Next, the cell state is determined using the tanh function. The value of is compressed to [-1, 1], and then combined with the output gate. Perform the Hadamard product, referring to the following formula 10.

[0138] Formula 10: .

[0139] The final generated It can be used both as the output at time t, as a prediction result of thermal efficiency, and as... Transmitted to the next moment, along with the input of the next moment. Jointly participate in subsequent information processing.

[0140] In this embodiment, by coordinating the control of different components in the Long Short-Term Memory network, the long-term variation pattern of time series characteristics is understood, and accurate thermal efficiency prediction results are output.

[0141] In one exemplary embodiment, such as Figure 9 As shown, the boiler thermal efficiency prediction method includes the following steps S1 to S8.

[0142] S1. Collect historical operating data of the target boiler and preprocess the data. This provides basic data for subsequent modeling. Preprocessing methods may include outlier removal, missing value completion, and normalization.

[0143] Outlier removal can be achieved using the 3σ criterion to remove outliers from dataset X. Outlier data (μ is the mean, σ is the standard deviation). Missing values ​​are filled in using linear interpolation. Normalization maps the preprocessed data to the [0,1] interval using Min-Max normalization, as shown in Formula 11 below.

[0144] Formula 11: .

[0145] in, , These are the minimum and maximum values ​​of the parameter sequence, respectively.

[0146] S2. Configure the parameters for Variational Mode Decomposition (VMD). This includes the number of decomposition modes, bandwidth constraints, noise tolerance, etc., to determine the rules for signal decomposition.

[0147] S3. Set the parameters of the Convolutional Neural Network (CNN). This includes the number of kernels, kernel size, stride, and pooling kernel size, which are used for subsequent spatial feature extraction.

[0148] S4. Configure the parameters of the Long Short-Term Memory (LSTM) network. This includes the learning rate, number of iterations, and regularization parameters, to adapt to the time-dependent modeling requirements of sequence data.

[0149] S5. Based on the VMD-CNN-LSTM model with pre-set parameters, perform iterative prediction using historical running data, and output the model's prediction results for model training or validation.

[0150] S6. Determine whether a real-time forecasting task is required; if no real-time forecasting is required, the process ends.

[0151] S7. If real-time prediction is required, input the real-time operating data of the target boiler and combine it with the historical data of the previous few time steps to construct a sample that meets the model input requirements and input it into the VMD-CNN-LSTM model.

[0152] S8. The model processes the input samples, obtains real-time prediction values, and outputs them. This completes the real-time prediction process.

[0153] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0154] Based on the same inventive concept, this application also provides a boiler thermal efficiency prediction device for implementing the boiler thermal efficiency prediction method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the boiler thermal efficiency prediction device provided below can be found in the limitations of the boiler thermal efficiency prediction method described above, and will not be repeated here.

[0155] In one exemplary embodiment, such as Figure 10 As shown, a boiler thermal efficiency prediction device 1000 is provided, including: a sample data construction module 1001, a variational mode decomposition module 1002, a scale feature extraction module 1003, a scale feature fusion module 1004, and a prediction result output module 1005.

[0156] The sample data construction module 1001 is used to acquire the operating data of the target boiler and construct input sample data based on the operating data; the operating data is used to characterize the thermal efficiency of the target boiler.

[0157] Variational mode decomposition module 1002 is used to perform variational mode decomposition on input sample data to obtain multiple mode components.

[0158] The scale feature extraction module 1003 is used to extract multi-time-scale features for each modal component, so as to obtain the time-scale features corresponding to the modal component at multiple time scales.

[0159] The scale feature fusion module 1004 is used to arrange the time scale feature fusion results at each time scale in a directed manner to obtain time series features.

[0160] The prediction result output module 1005 is used to input time series features into a long short-term memory network for time series prediction to obtain the predicted thermal efficiency of the target boiler.

[0161] In an exemplary embodiment, the sample data construction module 1001 is further configured to: extract real-time running data from the running data, and historical running data within a preset time step before the real-time running data; and construct input sample data based on the real-time running data and the historical running data.

[0162] In an exemplary embodiment, the variational mode decomposition module 1002 is further configured to: determine the target number of modes K; K is a natural number; and extract the modal components corresponding to the K modes from the input sample data with the goal of minimizing the sum of modal bandwidths.

[0163] In an exemplary embodiment, the variational mode decomposition module 1002 is further configured to: acquire the decomposition information of each modal component; increase the target number of modes when the decomposition information indicates that a first modal component in each modal component has a modal aliasing state; decrease the target number of modes when the decomposition information indicates that a second modal component in each modal component has a redundant state; return to the step of extracting the modal components corresponding to K modes from the input sample data with the goal of minimizing the sum of modal bandwidths, until there are no modal components in each modal component with a modal aliasing state or a redundant state.

[0164] In an exemplary embodiment, the scale feature extraction module 1003 is further configured to: input the modal components into the feature extraction channel respectively, perform convolution processing through multiple convolution kernels of different time scales in the feature extraction channel, and extract the local temporal features of the modal components at each time scale; perform pooling processing on each local temporal feature respectively to obtain the temporal scale features corresponding to the modal components at multiple time scales.

[0165] In an exemplary embodiment, the prediction result output module 1005 is further configured to: filter out long-term historical information in the time series features through the forgetting gate in the long short-term memory network; determine the injected new information in the long-term historical information; fuse the long-term historical information and the injected new information to obtain long-short-term fused information; and convert the long-short-term fused information into the thermal efficiency prediction result of the target boiler.

[0166] Each module in the aforementioned boiler thermal efficiency prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0167] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 11 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores the operating data of the target boiler. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a boiler thermal efficiency prediction method.

[0168] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 12As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a boiler thermal efficiency prediction method. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0169] Those skilled in the art will understand that Figure 11 and Figure 12 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0170] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0171] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0172] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0173] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0174] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0175] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for predicting boiler thermal efficiency, characterized in that, The method includes: Obtain the operating data of the target boiler, and construct input sample data based on the operating data; the operating data is used to characterize the thermal efficiency of the target boiler. The input sample data is subjected to variational mode decomposition to obtain multiple modal components; For each modal component, multi-timescale feature extraction is performed on the modal component to obtain the timescale features corresponding to the modal component at multiple timescales. The time-scale feature fusion results at each of the stated time scales are oriented to obtain time-series features; The time series features are input into a long short-term memory network for time series prediction to obtain the predicted thermal efficiency of the target boiler.

2. The method according to claim 1, characterized in that, The construction of input sample data based on the operational data includes: Extract the real-time running data from the running data, as well as the historical running data within a preset time step before the real-time running data; Input sample data is constructed based on the real-time operation data and the historical operation data.

3. The method according to claim 1, characterized in that, The variational mode decomposition process is performed on the input sample data to obtain multiple modal components, including: Determine the number of target modes K; K is a natural number; With the goal of minimizing the sum of modal bandwidths, modal components corresponding to K modes are extracted from the input sample data.

4. The method according to claim 3, characterized in that, The method further includes: Obtain the decomposition information of each modal component; If, in the case where each of the decomposed information represents a first modal component in which modal aliasing exists, the number of target modes is increased; In cases where the decomposed information characterizes a second modal component with a redundant state in each modal component, the number of target modes is reduced. Return to the step of extracting the modal components corresponding to K modes from the input sample data with the goal of minimizing the sum of modal bandwidths, until there are no modal components in the modal components that are in a modal aliasing state or a redundant state.

5. The method according to claim 1, characterized in that, The step of extracting multi-timescale features from the modal components to obtain the timescale features corresponding to the modal components at multiple timescales includes: The modal components are input into the feature extraction channel, and convolution processing is performed through multiple convolution kernels of different time scales in the feature extraction channel to extract the local temporal features of the modal components at each time scale. Each of the local time features is pooled to obtain the time scale features corresponding to the modal components at multiple time scales.

6. The method according to claim 1, characterized in that, The step of inputting the time series features into a long short-term memory network for time series prediction to obtain the predicted thermal efficiency of the target boiler includes: Long-term historical information in the time series features is filtered out by using the forgetting gate in the long short-term memory network. Determine the injection of new information into the long-term historical information; By fusing the long-term historical information and the injected new information, long-term and short-term fused information is obtained; The combined long-term and short-term information is then converted into a predicted thermal efficiency result for the target boiler.

7. A boiler thermal efficiency prediction device, characterized in that, The device includes: The sample data construction module is used to acquire the operating data of the target boiler and construct input sample data based on the operating data; the operating data is used to characterize the thermal efficiency of the target boiler. The variational mode decomposition module is used to perform variational mode decomposition processing on the input sample data to obtain multiple mode components; The scale feature extraction module is used to extract multi-time-scale features for each modal component to obtain the time-scale features corresponding to the modal component at multiple time scales. The scale feature fusion module is used to arrange the time scale feature fusion results under each time scale in a directed manner to obtain time series features. The prediction result output module is used to input the time series features into a long short-term memory network for time series prediction to obtain the predicted thermal efficiency of the target boiler.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

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

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

Images