An energy efficiency control method and system for shaft furnace combustion

Abstract

The present invention relates to the technical field of shaft furnace combustion control, and particularly to an energy efficiency control method and system for shaft furnace combustion. The method includes: extracting combustion acoustic fingerprint features by acquiring the combustion noise of the shaft furnace cavity and identifying the combustion state, setting a hierarchical early warning system according to the combustion state, obtaining the combustion early warning result, and dynamically adjusting the fuel flow value according to the combustion early warning result. By establishing an adaptive waste heat regulation center, the combustion early warning result is dynamically regulated, so as to conduct waste heat distribution. Through the present invention, the present invention effectively solves the problems that the traditional monitoring method is difficult to capture the subtle changes of the combustion state in time and is easy to miss the early signals of abnormal combustion. By collecting the combustion noise and extracting the combustion acoustic fingerprint features, the combustion state is dynamically monitored and accurately identified. Through the hierarchical early warning system and the dynamic adjustment of the fuel flow, the energy efficiency of the combustion process is significantly improved, and the energy waste and equipment failures caused by abnormal combustion are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of shaft furnace combustion control, and particularly to a method and system for controlling the energy efficiency of shaft furnace combustion. Background Art

[0002] With the continuous improvement of industrial automation and energy management requirements, as a common high-temperature industrial equipment, the combustion efficiency and energy utilization rate of the shaft furnace have an important impact on the overall production efficiency and energy consumption. The combustion efficiency of the shaft furnace directly affects the energy cost and production stability. Therefore, how to achieve precise and efficient energy management has become a key technical challenge in modern industrial production. Currently, the combustion process of the shaft furnace mainly relies on traditional physical quantities such as temperature and pressure for monitoring and adjustment. However, these monitoring methods often cannot reflect the combustion state in real time and comprehensively, and the adjustment of energy efficiency is relatively lagging.

[0003] Although the existing control technologies can perform a certain degree of combustion adjustment through the feedback of physical quantities, in actual applications, these methods often ignore the acoustic characteristics in the combustion process, resulting in the inability to accurately capture the subtle changes in the combustion state, making it difficult to detect abnormal combustion phenomena in a timely manner under complex working conditions, and thus affecting the combustion efficiency and energy utilization rate.

[0004] The information disclosed in this background art section is only intended to deepen the understanding of the overall background art of the present disclosure, and should not be regarded as an admission or any form of suggestion that this information constitutes the prior art known to those skilled in the art. Summary of the Invention

[0005] The present invention provides a method and system for controlling the energy efficiency of shaft furnace combustion, which can effectively solve the problems in the background art.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A method for controlling the energy efficiency of shaft furnace combustion, the method comprising:

[0008] Collecting the combustion noise of the shaft furnace cavity, extracting the combustion acoustic fingerprint features according to the combustion noise, and identifying the combustion state according to the combustion acoustic fingerprint features;

[0009] Setting up a hierarchical early warning system according to the combustion state, the hierarchical early warning system performing abnormal management according to the recognition result of the combustion state, obtaining a combustion early warning result, and dynamically adjusting the fuel flow value according to the combustion early warning result;

[0010] Establishing an adaptive waste heat regulation center, and dynamically adjusting the waste heat distribution strategy according to the combustion early warning result.

[0011] Further, extracting combustion acoustic fingerprint features according to the combustion noise includes:

[0012] Using a filter to denoise the combustion noise to obtain an audio preprocessing signal;

[0013] Performing time window segmentation on the audio preprocessing signal to obtain a number of time frames, and respectively performing amplitude-frequency conversion on the number of time frames to obtain a combustion noise spectrum;

[0014] Using a filter bank to weight the combustion noise spectrum to obtain combustion noise spectrum features;

[0015] Performing logarithmic transformation on the combustion noise spectrum features, and performing cepstral domain transformation on the transformed combustion noise spectrum features to obtain combustion acoustic fingerprint features.

[0016] Further, identifying the combustion state according to the combustion acoustic fingerprint features includes:

[0017] Collecting historical combustion noise signals to construct a training sample set, where the historical combustion noise signals include normal combustion signals and abnormal combustion signals;

[0018] Determining the kernel function type and penalty factor according to the signal features of the historical combustion noise signals in combination with a hyperparameter tuning algorithm to obtain a pre-identification model, where the kernel function is used to determine the boundary for distinguishing the combustion state, and the penalty factor is used to handle the deviation during model training;

[0019] Training the pre-identification model using the training sample set to obtain a combustion state identification model, and identifying the combustion state according to the combustion state for the combustion acoustic fingerprint features to obtain the combustion state.

[0020] Further, setting a hierarchical early warning system according to the combustion state includes:

[0021] Outputting a combustion state identification result according to the combustion state identification model, and calculating a confidence score;

[0022] Obtaining historical combustion data, and setting a combustion trend change standard according to the historical combustion data;

[0023] Assigning weights to the confidence score and the combustion trend change standard respectively, and obtaining a final early warning value;

[0024] Setting a hierarchical early warning according to the final early warning value.

[0025] Further, calculating the confidence score includes:

[0026] Obtaining combustion state discrimination conditions according to the historical combustion noise signals, and setting an initial decision boundary;

[0027] Set boundary constraint conditions according to the support vector machine algorithm, and optimize the initial decision boundary using mathematical methods according to the boundary constraint conditions to obtain an execution decision boundary;

[0028] Calculate the distance from each combustion noise signal point to the execution decision boundary, and the distance is expressed as the confidence score of each combustion noise signal point.

[0029] Furthermore, determine the kernel function type and penalty factor according to the signal characteristics of the historical combustion noise signal in combination with the hyperparameter tuning algorithm, including:

[0030] Extract combustion noise data characteristics from the historical combustion noise signal;

[0031] Set a candidate set of kernel functions according to the combustion noise data characteristics, and perform simulation training on several kernel functions in the candidate set of kernel functions using a training sample set, and select a kernel function based on the training results;

[0032] Set an initial penalty factor candidate value according to the historical combustion noise signal, perform cross-combination on the initial penalty factor candidate value using the grid search method, and perform simulation training respectively using the training sample set to determine the penalty factor.

[0033] Furthermore, dynamically adjust the fuel flow value according to the combustion warning result, including:

[0034] Record the abnormal combustion time according to the combustion warning result, obtain the corresponding confidence score, and calculate the thermal efficiency loss value according to the abnormal combustion time and the confidence score;

[0035] Feed back the thermal efficiency loss value to the fuel supply system, and the fuel supply system controls the fuel flow value based on the thermal efficiency loss value.

[0036] Furthermore, dynamically adjust the waste heat distribution strategy according to the combustion warning result, including:

[0037] Establish a joint control channel based on the hierarchical warning system, and perform real-time monitoring on the combustion state of the shaft furnace cavity to obtain real-time monitoring results;

[0038] When the combustion state is abnormal, extract the overflow heat according to the real-time monitoring result, and cooperate to control the waste heat distribution system to export the overflow heat;

[0039] When the combustion state is restored, schedule the waste heat based on a preset distribution strategy.

[0040] An energy efficiency control system for shaft furnace combustion, the system includes:

[0041] A combustion noise state recognition module that collects the combustion noise in the shaft furnace cavity, extracts combustion acoustic fingerprint features based on the combustion noise, and recognizes the combustion state based on the combustion acoustic fingerprint features;

[0042] A hierarchical warning and adjustment module that sets up a hierarchical warning system according to the combustion state. The hierarchical warning system performs abnormal management based on the recognition result of the combustion state, obtains a combustion warning result, and dynamically adjusts the fuel flow value according to the combustion warning result;

[0043] A waste heat adjustment and distribution module that establishes an adaptive waste heat adjustment center and dynamically adjusts the waste heat distribution strategy according to the combustion warning result.

[0044] Further, the combustion noise state recognition module includes:

[0045] A combustion noise denoising processing unit that uses a filter to denoise the combustion noise and obtains an audio preprocessing signal;

[0046] A combustion noise spectrum acquisition unit that performs time window segmentation on the audio preprocessing signal to obtain a number of time frames, and performs amplitude-frequency conversion on the number of time frames respectively to obtain a combustion noise spectrum;

[0047] A spectrum weighted feature extraction unit that weights the combustion noise spectrum using a filter bank to obtain combustion noise spectrum features;

[0048] A combustion acoustic fingerprint acquisition unit that performs logarithmic transformation on the combustion noise spectrum features and performs cepstral domain transformation on the transformed combustion noise spectrum features to obtain combustion acoustic fingerprint features.

[0049] Through the technical solution of the present invention, the following technical effects can be achieved:

[0050] The present invention effectively solves the problems that the traditional monitoring method is difficult to capture the subtle changes in the combustion state in time and is easy to miss the early signals of abnormal combustion. By collecting the combustion noise and extracting the combustion acoustic fingerprint features, the combustion state is dynamically monitored and accurately recognized. Through the hierarchical warning system and the dynamic adjustment of the fuel flow, the energy efficiency of the combustion process is significantly improved, and the energy waste and equipment failures caused by abnormal combustion are reduced.

[0051] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the specific embodiments of the present application are specifically described below. Brief Description of the Drawings

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0053] Figure 1 It is a schematic flowchart of an energy efficiency control method for shaft furnace combustion;

[0054] Figure 2 It is a schematic flowchart of combustion acoustic fingerprint feature extraction;

[0055] Figure 3 It is a schematic flowchart of the setting of a hierarchical early warning system;

[0056] Figure 4 It is a relationship diagram of the combustion noise state recognition module. Specific Embodiments

[0057] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. The terms used in the description of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0059] Embodiment 1;

[0060] As Figure 1 shown, the present application provides an energy efficiency control method for shaft furnace combustion, and the method includes:

[0061] S10: Collect the combustion noise of the shaft furnace cavity, extract the combustion acoustic fingerprint features according to the combustion noise, and identify the combustion state according to the combustion acoustic fingerprint features;

[0062] S20: Set up a hierarchical early warning system according to the combustion state. The hierarchical early warning system performs abnormal management according to the recognition result of the combustion state, obtains the combustion early warning result, and dynamically adjusts the fuel flow value according to the combustion early warning result;

[0063] S30: Establish an adaptive waste heat regulation center, and dynamically adjust the waste heat distribution strategy according to the combustion early warning result.

[0064] Specifically, first, install multi-channel noise acquisition equipment inside the shaft furnace cavity. In some embodiments, highly sensitive piezoelectric sensors or microphone arrays are used. Such equipment has good anti-interference performance and can work stably in high-temperature and harsh industrial environments. At the same time, the noise acquisition equipment has a wide frequency response range, which is 10 Hz to 10 kHz in this embodiment, and is used to capture the low-frequency and high-frequency noise characteristics during the shaft furnace combustion process. Then, perform noise reduction processing on the collected combustion noise signal. Based on the denoised signal, time-frequency analysis methods such as short-time Fourier transform and wavelet transform can be used to extract combustion acoustic fingerprint features to analyze the spectral characteristics of the signal. According to the extracted combustion acoustic fingerprint features, use pattern recognition algorithms (such as decision trees, neural networks, etc.) to identify the combustion state, judge whether it is in a normal combustion state or an abnormal combustion state, and then construct a hierarchical early warning system according to the recognition result of the combustion state. Obtain the early warning result according to the hierarchical early warning system, and dynamically adjust the fuel flow to ensure that the combustion process is in an optimal state. Then, based on the combustion early warning result, establish an adaptive waste heat regulation center, which can dynamically monitor data such as the temperature and pressure of the waste heat source according to the real-time combustion state and dynamically adjust the waste heat flow direction.

[0065] Through the technical solution of the present invention, the problems that the traditional monitoring method is difficult to capture the subtle changes in the combustion state in time and is easy to miss the early signals of abnormal combustion are effectively solved. By collecting combustion noise and extracting combustion acoustic fingerprint features, the combustion state is dynamically monitored and accurately identified. Through the hierarchical early warning system and the dynamic adjustment of the fuel flow, the energy efficiency of the combustion process is significantly improved, and the energy waste and equipment failures caused by abnormal combustion are reduced.

[0066] Furthermore, as Figure 2 shown, extracting combustion acoustic fingerprint features from combustion noise includes:

[0067] S11: Use a filter to perform noise reduction processing on the combustion noise to obtain an audio preprocessing signal;

[0068] S12: Perform time window segmentation on the audio preprocessing signal to obtain a number of time frames, and perform amplitude-frequency conversion on the number of time frames respectively to obtain the combustion noise spectrum;

[0069] S13: Use a filter bank to weight the combustion noise spectrum to obtain combustion noise spectrum features;

[0070] S14: Perform logarithmic transformation on the combustion noise spectrum features and perform cepstral domain transformation on the transformed combustion noise spectrum features to obtain combustion acoustic fingerprint features.

[0071] As a preference of the above embodiments, first, noise reduction processing is performed on the collected combustion noise, and the noise reduction processing solution is as follows: First, a high-pass filter (such as a band-pass filter) is used to set a suitable range for noise cancellation. In some embodiments, the lower limit of the filter is set to 20 Hz and the upper limit is set to 10 kHz to retain only the frequency components related to the combustion process. Then, a band-stop filter is used to remove the factory environmental noise (such as mechanical equipment vibration, fan noise, etc.) in the signal to obtain a purer combustion noise signal. Then, the denoised signal, which is the audio preprocessing signal, is saved as a digital signal. Subsequently, the audio preprocessing signal is segmented by time windows. In at least one embodiment, the window size is set to 50 ms to 100 ms, and the audio signal is segmented according to the window size. Each time window contains a time frame. When processing, the signal of each time frame will be subjected to separate spectral analysis, and then amplitude-frequency conversion is performed on each time frame. The amplitude-frequency conversion solution is as follows: The signal data within each time frame is prepared as an N-point discrete time series. Assuming there are N data points in a time frame, the fast Fourier transform is applied to the signal data of each time frame to convert it into a frequency-domain signal. The basic operation of the fast Fourier transform is to decompose the signal in the time domain into the sum of different frequency components, and the amplitude of each frequency can be obtained. After conversion by the fast Fourier transform, the amplitude corresponding to each frequency component is the intensity of that frequency component in the signal. For each frequency point, the amplitude is the modulus value of a complex number. Subsequently, based on the calculated spectral amplitude values, the intensity distribution of the signal at different frequencies is analyzed to output the spectrum. Then, a filter bank is used to perform weighted processing on the spectrum. For example, the high-frequency part can be weighted to emphasize the high-frequency noise characteristics of combustion, or the low-frequency part can be weighted to emphasize the low-frequency noise characteristics. After the weighting operation, the combustion noise spectrum characteristics are obtained. To enhance the characteristics of the low-frequency signal, logarithmic transformation is performed on the combustion noise spectrum characteristics, and the spectrum characteristic values are taken as logarithms. Finally, inverse cepstral domain conversion is performed on the logarithmically transformed combustion noise spectrum characteristics, and the discrete cosine transform (DCT) or discrete Fourier transform (DFT) can be used to complete this conversion to obtain the combustion voiceprint characteristics, which are used as the input data for subsequent combustion state recognition.

[0072] Furthermore, identifying the combustion state based on the combustion voiceprint characteristics includes:

[0073] Collecting historical combustion noise signals to construct a training sample set, where the historical combustion noise signals include normal combustion signals and abnormal combustion signals;

[0074] Determining the kernel function type and penalty factor according to the signal characteristics of the historical combustion noise signals in combination with the hyperparameter tuning algorithm to obtain a pre-identification model. The kernel function is used to determine the boundary for distinguishing the combustion state, and the penalty factor is used to handle the deviation during model training;

[0075] Train the pre-identification model using the training sample set to obtain a combustion state identification model, and identify the combustion acoustic fingerprint features according to the combustion state to obtain the combustion state.

[0076] In this embodiment, first, extract the historical combustion noise signals according to the shaft furnace operation records to ensure that there are two types of sample data, namely, sufficient normal combustion signals and abnormal combustion signals. Label the collected historical combustion noise signals according to their corresponding combustion states (normal or abnormal), and organize them into a training sample set. Each sample should include the corresponding combustion acoustic fingerprint features and their corresponding combustion state labels. Then, use a machine learning algorithm to construct a pre-identification model. For example, use support vector machines or random forests, and determine the optimal hyperparameters of the pre-identification model through hyperparameter tuning algorithms such as grid search and random search. The key hyperparameters include: the type of kernel function and the penalty factor. The main function of the kernel function is to map the data to a higher-dimensional space, making the data that is inseparable in the low-dimensional space separable in the high-dimensional space. Simply put, the kernel function makes the data separable by a simple decision boundary (such as a hyperplane) in the higher-dimensional space by adding additional features. The penalty factor is used to control the complexity of the model to avoid overfitting or underfitting. A smaller penalty factor allows more training errors, while a larger penalty factor tends to reduce errors but may lead to overfitting. Subsequently, use the extracted training sample set to train the pre-identification model. Once the pre-identification model is trained, it can be applied to real-time monitoring. In actual applications, the collected combustion noise signals are preprocessed and feature-extracted, and then input into the trained pre-identification model. The model identifies the current combustion state according to the input combustion acoustic fingerprint features.

[0077] Furthermore, as Figure 3 shown, set up a hierarchical early warning system according to the combustion state, including:

[0078] S21: Output the combustion state identification result according to the combustion state identification model and calculate the confidence score;

[0079] S22: Obtain the historical combustion data and set the combustion trend change standard according to the historical combustion data;

[0080] S23: Assign weights to the confidence score and the combustion trend change standard respectively and obtain the final early warning value;

[0081] S24: Set up a hierarchical early warning according to the final early warning value.

[0082] Specifically, first, a combustion state recognition result is output according to the combustion recognition model. This result is usually divided into two categories: "normal combustion" and "abnormal combustion". After each recognition result, the confidence score of this result is calculated to measure the accuracy of the recognition result. In some embodiments, a support vector machine or a probabilistic regression method is used to calculate the confidence score. Then, based on historical combustion data, a standard for combustion trend change is set to judge the fluctuations during the combustion process. For example, the standard includes: the change range of combustion temperature (such as abnormal if the combustion temperature fluctuation exceeds the set threshold), the change trend of fuel flow (such as abnormal if the fuel flow suddenly increases or decreases by more than a certain proportion). Then, according to the confidence score of the combustion state recognition result, corresponding weights can be set. A recognition result with a higher confidence score can be given a higher weight, while a recognition result with a lower confidence score should have its influence reduced. The weight setting method can be as follows: confidence score > 0.8 indicates that the model is very confident in the judgment of the combustion state. At this time, the confidence score should have a greater impact on the final warning value. Therefore, a higher weight is given, such as 0.7. 0.5 < confidence score ≤ 0.8 means that the influence of the confidence score on the warning value is relatively moderate, and the weight can be set to 0.5 for reference. Confidence score ≤ 0.5, the weight is set to 0.3. According to the severity and urgency of the combustion trend change standard, corresponding weights are also set. The confidence score and the weights of the combustion trend change standard are combined to calculate the final warning value. According to the final warning value, a graded warning is set. The reference plan is as follows: warning value > 0.8, the system judges it as a serious abnormality, triggers an emergency warning, and immediate measures need to be taken, such as adjusting the fuel flow, suspending the operation, or conducting equipment inspections. 0.5 < warning value ≤ 0.8, the system judges it as a medium abnormality, triggers a medium-level warning, and requires enhanced monitoring and real-time adjustment, and may require manual intervention. Warning value ≤ 0.5, the system considers the state normal or slightly abnormal, triggers a low-level warning, continues to monitor, and conducts periodic inspections.

[0083] Furthermore, calculating the confidence score includes:

[0084] Obtaining the combustion state discrimination condition according to the historical combustion noise signal and setting the initial decision boundary;

[0085] Setting boundary constraint conditions according to the support vector machine algorithm, and optimizing the initial decision boundary by using mathematical methods according to the boundary constraint conditions to obtain the execution decision boundary;

[0086] Calculating the distance from each combustion noise signal point to the execution decision boundary, and the distance represents the confidence score of each combustion noise signal point.

[0087] As a preference of the above embodiments, combustion state discrimination conditions are defined according to historical combustion noise signals. For example, by analyzing the performance of noise signals in terms of frequency and amplitude under different combustion states, features capable of distinguishing normal combustion from abnormal combustion are extracted, and an initial decision boundary is set. The initial decision boundary is a dividing line that separates normal combustion and abnormal combustion data points in a multi-dimensional feature space. Then, the support vector machine algorithm is used to train the support vector machine model based on the features of historical combustion noise signals (such as spectral, time-domain features, etc.). The goal of the support vector machine is to find an optimal decision boundary to separate samples of normal combustion and abnormal combustion. Then, boundary constraint conditions are set according to the support vector machine: First, the support vector machine algorithm will find a hyperplane or a set of hyperplanes in the multi-dimensional feature space to separate normal and abnormal combustion data points. To ensure the accuracy of the boundary, the goal of the support vector machine is to maximize the margin between the boundary and the data points, and the robustness of the model is improved by introducing a soft margin (i.e., allowing a small number of misclassifications). After obtaining the optimal execution decision boundary according to the boundary constraint conditions, the operator needs to calculate the distance from each combustion noise signal point to the decision boundary. This distance reflects the degree of proximity of the signal point to the decision boundary. By calculating the distance from each signal point to the execution decision boundary and normalizing it, the confidence score of each signal point is obtained.

[0088] Furthermore, the kernel function type and penalty factor are determined according to the signal features of historical combustion noise signals in combination with a hyperparameter tuning algorithm, including:

[0089] Extract combustion noise data features according to historical combustion noise signals;

[0090] Set a candidate set of kernel functions according to the combustion noise data features, and perform simulation training on several kernel functions in the candidate set of kernel functions using a training sample set, and select a kernel function based on the training results;

[0091] Set an initial candidate value of the penalty factor according to historical combustion noise signals, perform cross-combination on the initial candidate value of the penalty factor using the grid search method, and perform simulation training respectively using the training sample set to determine the penalty factor.

[0092] In this embodiment, first, a candidate set of kernel functions is set according to the extracted combustion noise data features, such as spectral features, time-domain features, etc., including linear kernel functions, radial basis kernel functions, and polynomial kernel functions. Then, a kernel function is selected according to the quantity and complexity of the combustion noise data features: using the training sample set, each kernel function in the candidate set of kernel functions is simulated and trained to evaluate its performance on the training set. The training process can use the support vector machine algorithm. Under each kernel function, the support vector machine will learn how to separate normal combustion and abnormal combustion data through the optimal hyperplane, and then evaluate the performance of each kernel function through cross-validation. The cross-validation method is as follows: one subset is used to test the model, and the remaining K - 1 subsets are used to train the model. Cross-validation is usually performed K times iteratively, each time using a different subset as the validation set, and finally calculating the average evaluation result of all iterations; finally, a kernel function is selected according to the evaluation result; then, in some embodiments, the method for setting the penalty factor is as follows: first, several initial candidate values of the penalty factor are selected, usually with a value range of several different values from 0.01 to 100. For example, the candidate set of the penalty factor can be set as {0.1, 1, 10, 100}, and then the grid search method can be used to traverse multiple candidate values and combine them in sequence. Combining cross-validation, the optimal penalty factor is finally obtained. This penalty factor can effectively balance the bias and variance of the model and improve the generalization ability of the support vector machine model.

[0093] Furthermore, the fuel flow value is dynamically adjusted according to the combustion warning result, including:

[0094] Record the abnormal combustion time according to the combustion warning result, and obtain the corresponding confidence score. Calculate the thermal efficiency loss value according to the abnormal combustion time and the confidence score;

[0095] Feed back the thermal efficiency loss value to the fuel supply system, and the fuel supply system controls the fuel flow value based on the thermal efficiency loss value.

[0096] Specifically, first, when the system detects abnormal combustion, record the duration of the abnormal combustion. The abnormal combustion time refers to the time period when the combustion state is determined to be abnormal. The specific operation is as follows: when abnormal combustion is detected, the system automatically starts timing until the combustion state returns to normal. The system records this time period as the abnormal combustion time and stores the time value for subsequent calculation. Then, obtain the confidence score according to the combustion state recognition model and calculate the thermal efficiency loss value. The thermal efficiency loss refers to the energy loss caused by abnormal combustion during the combustion process. The abnormal combustion state usually leads to incomplete combustion or energy waste, thus reducing the overall thermal efficiency. The calculation formula is as follows: L heat = T ab × S conf × η, T ab represents the duration of abnormal combustion, Sconf The confidence score indicating abnormal combustion is denoted as η. Subsequently, the heat efficiency loss value is fed back to the fuel supply system, which dynamically adjusts the fuel flow based on the feedback heat efficiency loss value. If the heat efficiency loss value is high, indicating incomplete combustion or excessive fuel, the system will reduce the fuel flow and adjust it to an appropriate value to optimize the combustion process.

[0097] Furthermore, the waste heat distribution strategy is dynamically adjusted according to the combustion warning result, including:

[0098] A joint control channel is established based on the hierarchical warning system, and the combustion state of the shaft furnace cavity is monitored in real time to obtain the real-time monitoring result;

[0099] When the combustion state is abnormal, the overflow heat is extracted according to the real-time monitoring result, and the waste heat distribution system is coordinated to export the overflow heat;

[0100] When the combustion state is restored, the waste heat is scheduled based on the preset distribution strategy.

[0101] As a preference of the above embodiment, first, a joint control channel is established for the shaft furnace control system and the waste heat distribution system according to the hierarchical warning system. This channel is used to transmit the combustion state information in real time so as to dynamically adjust the waste heat distribution strategy according to the combustion warning result. Then, various sensors (such as temperature sensors, pressure sensors, gas concentration sensors, etc.) are deployed inside the shaft furnace cavity to monitor the combustion state in real time. The monitoring data will be transmitted to the combustion state recognition system in real time. The system analyzes the real-time data to judge whether the combustion state is normal and outputs the combustion state warning result according to the preset threshold. If combustion abnormality occurs, the system will automatically trigger the waste heat adjustment process. When the system identifies that the combustion state is abnormal (such as excessive temperature fluctuation, abnormal oxygen concentration, etc.), it is necessary to trigger the waste heat adjustment mechanism according to the degree of combustion abnormality: when incomplete combustion or abnormality generates excess heat, which is called overflow heat, the waste heat distribution system will automatically adjust the waste heat flow direction according to the size and type of the overflow heat. For example, the overflow heat can be directed to a waste heat boiler, an air preheater or other energy recovery devices to achieve heat energy recovery. When the combustion state returns to the normal level, the operator should ensure that the system promptly returns to the normal waste heat distribution strategy. At this time, the waste heat distribution system will reschedule the heat energy flow direction according to the preset distribution strategy to ensure the efficient operation of the system after the combustion state is restored.

[0102] Embodiment 2;

[0103] Based on the same inventive concept as a method for controlling the energy efficiency of shaft furnace combustion in the foregoing embodiment, the present invention also provides a control system for the energy efficiency of shaft furnace combustion, and the system includes:

[0104] The combustion noise state recognition module collects the combustion noise in the shaft furnace cavity, extracts the combustion acoustic texture features based on the combustion noise, and recognizes the combustion state according to the combustion acoustic texture features;

[0105] The hierarchical early warning and adjustment module sets up a hierarchical early warning system according to the combustion state. The hierarchical early warning system conducts abnormal management based on the recognition result of the combustion state, obtains the combustion early warning result, and dynamically adjusts the fuel flow value according to the combustion early warning result;

[0106] The waste heat adjustment and distribution module establishes an adaptive waste heat adjustment center and dynamically adjusts the waste heat distribution strategy according to the combustion early warning result.

[0107] The above adjustment system in the present invention can effectively implement the energy efficiency control method for shaft furnace combustion, and the technical effects that can be achieved are as described in the above embodiments, which will not be elaborated here.

[0108] Furthermore, as Figure 4 shown, the combustion noise state recognition module includes:

[0109] The combustion noise noise reduction processing unit uses a filter to denoise the combustion noise and obtains an audio preprocessing signal;

[0110] The combustion noise spectrum acquisition unit performs time window segmentation on the audio preprocessing signal, obtains a plurality of time frames, and performs amplitude-frequency conversion on the plurality of time frames respectively to obtain the combustion noise spectrum;

[0111] The spectrum weighted feature extraction unit weights the combustion noise spectrum using a filter bank to obtain the combustion noise spectrum features;

[0112] The combustion acoustic texture acquisition unit performs logarithmic transformation on the combustion noise spectrum features and performs cepstral domain transformation on the transformed combustion noise spectrum features to obtain the combustion acoustic texture features.

[0113] Similarly, for the above optimization solutions of the system, the corresponding optimization effects of the method in the first embodiment can also be respectively achieved, which will not be elaborated here either.

[0114] Although the present application has been described in combination with specific features and their embodiments, it is obvious that various modifications and combinations can be made without departing from the spirit and scope of the present application. Accordingly, this specification and the drawings are merely exemplary descriptions of the present application defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A method for controlling the energy efficiency of shaft furnace combustion, characterized in that, The method includes: Collecting the combustion noise of the shaft furnace cavity, extracting combustion acoustic fingerprint features based on the combustion noise, and identifying the combustion state based on the combustion acoustic fingerprint features; Setting a hierarchical early warning system according to the combustion state, where the hierarchical early warning system performs abnormal management based on the recognition result of the combustion state, obtains a combustion early warning result, and dynamically adjusts the fuel flow value according to the combustion early warning result; Establishing an adaptive waste heat regulation center and dynamically adjusting the waste heat distribution strategy according to the combustion early warning result; Extracting combustion acoustic fingerprint features based on the combustion noise, including: Performing denoising processing on the combustion noise using a filter to obtain an audio preprocessing signal; Performing time window segmentation on the audio preprocessing signal to obtain a number of time frames, respectively performing amplitude-frequency transformation on the number of time frames to obtain a combustion noise spectrum; Weighting the combustion noise spectrum using a filter bank to obtain combustion noise spectrum features; Performing logarithmic transformation on the combustion noise spectrum features and performing cepstral domain transformation on the transformed combustion noise spectrum features to obtain combustion acoustic fingerprint features; Identifying the combustion state based on the combustion acoustic fingerprint features, including: Collecting historical combustion noise signals and constructing a training sample set, where the historical combustion noise signals include normal combustion signals and abnormal combustion signals; Determining the kernel function type and penalty factor according to the signal features of the historical combustion noise signals in combination with a hyperparameter tuning algorithm to obtain a pre-identification model, where the kernel function is used to determine the boundary for distinguishing the combustion state, and the penalty factor is used to handle the deviation during model training; Training the pre-identification model using the training sample set to obtain a combustion state identification model, identifying the combustion acoustic fingerprint features according to the combustion state, and obtaining the combustion state.

2. The energy efficiency control method for shaft furnace combustion according to claim 1, characterized in that, Setting a hierarchical early warning system according to the combustion state, including: Outputting a combustion state recognition result according to the combustion state recognition model and calculating a confidence score; Obtaining historical combustion data and setting a combustion trend change standard according to the historical combustion data; Assigning weights to the confidence score and the combustion trend change standard respectively and obtaining a final early warning value; Setting a hierarchical early warning according to the final early warning value.

3. The energy efficiency control method for shaft furnace combustion according to claim 2, wherein Calculating the confidence score, including: Obtaining combustion state discrimination conditions according to the historical combustion noise signals and setting an initial decision boundary; Setting boundary constraint conditions according to the support vector machine algorithm, and optimizing the initial decision boundary using a mathematical method according to the boundary constraint conditions to obtain an execution decision boundary; Calculating the distance from each combustion noise signal point to the execution decision boundary, and the distance is expressed as the confidence score of each combustion noise signal point.

4. The energy efficiency control method for shaft furnace combustion according to claim 1, characterized in that, Determining the kernel function type and penalty factor according to the signal features of the historical combustion noise signals in combination with a hyperparameter tuning algorithm, including: Extracting combustion noise data features from the historical combustion noise signals; Setting a kernel function candidate set according to the combustion noise data features, respectively performing simulation training on a number of kernel functions in the kernel function candidate set using the training sample set, and selecting a kernel function based on the training results; Set the initial penalty factor candidate value according to the historical combustion noise signal, cross - combine the initial penalty factor candidate values using the grid search method, and perform simulation training respectively using the training sample set to determine the penalty factor.

5. The energy efficiency control method for shaft furnace combustion according to claim 1, wherein, Dynamically adjust the fuel flow value according to the combustion warning result, including: Record the abnormal combustion time according to the combustion warning result, obtain the corresponding confidence score, and calculate the thermal efficiency loss value according to the abnormal combustion time and the confidence score; Feed back the thermal efficiency loss value to the fuel supply system, and the fuel supply system controls the fuel flow value based on the thermal efficiency loss value.

6. The energy efficiency control method for shaft furnace combustion according to claim 1, characterized in that, Dynamically adjust the waste heat distribution strategy according to the combustion warning result, including: Establish a joint control channel based on the hierarchical warning system, and monitor the combustion state of the shaft furnace cavity in real - time to obtain the real - time monitoring result; When the combustion state is abnormal, extract the overflow heat according to the real - time monitoring result, and jointly control the waste heat distribution system to export the overflow heat; When the combustion state recovers, schedule the waste heat based on the preset distribution strategy.

7. An energy efficiency control system for shaft furnace combustion, characterized in that, Adopt the energy efficiency control method for shaft furnace combustion as described in Claim 1, and the system includes: A combustion noise state recognition module that collects the combustion noise of the shaft furnace cavity, extracts the combustion soundprint features according to the combustion noise, and identifies the combustion state according to the combustion soundprint features; A hierarchical warning adjustment module that sets up a hierarchical warning system according to the combustion state, the hierarchical warning system performs abnormal management according to the recognition result of the combustion state, obtains the combustion warning result, and dynamically adjusts the fuel flow value according to the combustion warning result; A waste heat adjustment and distribution module that establishes an adaptive waste heat adjustment center and dynamically adjusts the waste heat distribution strategy according to the combustion warning result; The combustion noise state recognition module includes: A combustion noise denoising processing unit that uses a filter to denoise the combustion noise to obtain an audio pre - processing signal; A combustion noise spectrum acquisition unit that performs time - window segmentation on the audio pre - processing signal to obtain a number of time frames, and performs amplitude - frequency conversion on the number of time frames respectively to obtain the combustion noise spectrum; A spectrum weighted feature extraction unit that weights the combustion noise spectrum using a filter bank to obtain the combustion noise spectrum features; A combustion soundprint acquisition unit that performs logarithmic transformation on the combustion noise spectrum features and performs cepstral domain transformation on the transformed combustion noise spectrum features to obtain the combustion soundprint features.

Images