Intelligent combustion optimization method and system
By analyzing historical combustion data over time and optimizing the accurate analysis model, the problems of high energy consumption, low thermal efficiency, and excessive pollutants in traditional combustion optimization methods have been solved, achieving precise optimization of combustion parameters and improved system robustness.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-30
- Publication Date
- 2026-04-03
AI Technical Summary
Traditional combustion optimization methods suffer from problems such as multivariate coupling effects, insufficient utilization of historical data, and slow real-time control response, resulting in high energy consumption, low thermal efficiency, and excessive pollutants, especially lacking adaptive correction mechanisms under complex operating conditions.
By acquiring historical combustion data and performing time-series analysis, a historical task timeline and parameter curves are constructed. Curve alignment and random splitting are performed to build an optimized and accurate analysis model. Simulated combustion detection curves are used to evaluate optimization indicators and achieve intelligent combustion control.
It achieves precise optimization of combustion parameters, reduces energy consumption and emissions, improves system robustness, and ensures stability and safety under varying operating conditions.
Smart Images
Figure CN121782597A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of boiler combustion control optimization technology, and in particular to a smart combustion optimization method and system. Background Technology
[0002] In industrial production and energy utilization, the combustion process, as a core component, directly impacts the system's energy efficiency, emission control, and operational stability. Traditional combustion optimization methods primarily rely on engineering experience, manual parameter adjustment, or simple feedback control mechanisms. For example, combustion control parameters (such as fuel supply rate, air-fuel ratio, and ignition timing) are preset using empirical formulas and then adjusted retrospectively based on real-time monitored parameters (such as temperature, oxygen concentration, and pollutant emissions). However, this approach has significant limitations: First, the combustion process is affected by multiple coupled variables (such as ambient temperature, fuel quality fluctuations, and load changes), making it difficult to achieve global optimum in parameter optimization, often leading to excessive energy consumption or low thermal efficiency. Second, insufficient utilization of historical data prevents effective extraction of time-series patterns, resulting in weak generalization ability of the optimization model and low simulation and prediction accuracy. Third, slow real-time control response and accumulated deviations can easily lead to pollutant exceedances or safety hazards, especially under complex operating conditions, where the lack of adaptive correction mechanisms further amplifies optimization deviations. Summary of the Invention
[0003] The purpose of this invention is to provide a smart combustion optimization method and system that can optimize boiler combustion control.
[0004] This invention discloses a smart combustion optimization method, see reference. Figure 1 ,include: Step S100: Obtain historical combustion data, perform time series analysis on the historical combustion data, and configure the combustion control parameters and combustion detection parameters at different time points in the preset task timeline template to obtain the historical task timeline. Parallel to the historical task timeline, construct several historical combustion control parameter curves and historical combustion detection parameter curves. Step S200: Based on the alignment method of several historical task timelines, align the historical combustion control parameter curves and historical combustion detection parameter curves belonging to the same historical task timeline to obtain a group of historical combustion parameter curves. Then, randomly split each group of historical combustion parameter curves several times to obtain several groups of historical combustion parameter curves. Step S300: Based on grouping several historical combustion parameter curves, construct an optimized and accurate analysis model; Step S400: Construct several sets of simulated combustion parameter curves, evaluate the optimization index of the simulated combustion parameter curve sets using the simulated combustion detection curves, and analyze the accuracy index of the simulated combustion parameter curve sets using the optimization accuracy analysis model. Combine the optimization index and the accuracy index to determine the selected set of simulated combustion parameter curves. Step S500: Intelligent combustion control is performed based on the selected set of simulated combustion parameter curves.
[0005] In some embodiments disclosed in this invention, the method for constructing an optimized and accurate analysis model includes: Step S301: Perform variation adjustment on the historical combustion control parameter curve segment and the historical combustion detection parameter curve segment in each historical combustion parameter curve group to obtain several variation combustion parameter curve groups. Step S302: Compare each group of variable combustion parameter curves with different groups of historical combustion parameter curves to identify several similar groups of historical combustion parameter curves. The method for determining similarity includes determining the degree of similarity between them. If the degree of similarity is greater than or equal to a preset value, the corresponding group of historical combustion parameter curves is identified as similar. Step S303: Based on the similarity of several similar historical combustion parameter curves, determine the accurate index of the variable combustion parameter curve grouping. Step S304: Group the variable combustion parameter curves into training data, use the accuracy index as the training label, train the model, and obtain the optimized accuracy analysis model.
[0006] In some embodiments disclosed in this invention, the method for determining the similarity between a group of variable combustion detection parameter curves and a group of historical combustion parameter curves includes: Step S3021: Determine the intersection range between each relative parameter curve segment, and as time progresses, determine the change in the intersection range area, including the unit intersection area per unit time corresponding to different time nodes, and the average unit intersection area. Step S3022: Determine the degree of similarity based on the change in the area of the intersection range between the relative parameter curve segments; The expression for calculating the degree of similarity is: ; Where S represents the degree of similarity. Let L be the average unit cross area, L be the normalization adjustment coefficient between the average unit cross area and the variance of the unit cross area, K be the adjustment coefficient for the influence of the variance of the unit cross area, b be the adjustment constant for the influence of the variance of the unit cross area, and T be the total number of time points involved in the alignment. Let be the unit cross area at time point t. To preset the maximum average unit cross area, This is the variance of the preset maximum unit cross area.
[0007] In some embodiments disclosed in this invention, the method for determining accurate indices for grouping variable combustion parameter curves includes: Step S3031: Analyze the determined similarity levels to determine the highest similarity level and the average similarity level, and determine the accurate index based on the highest similarity level and the average similarity level. The expression for calculating the accuracy index is as follows: ; Where Q is the accuracy indicator. For the highest degree of similarity, To achieve the average similarity, The weighted number of points representing the highest similarity. The weighted number of points representing the average similarity is denoted by , where and The similarity is determined based on the difference between the highest similarity and the average similarity. The determination method includes determining the preset similarity difference interval to which the difference belongs, and determining the corresponding weight allocation points based on the preset similarity difference interval.
[0008] In some embodiments disclosed in this invention, the method for training a model by grouping variable combustion parameter curves into training data and using accurate indicators as training labels includes: Step S3041: Feature extraction and preprocessing are performed on the variable combustion parameter curves, including quantizing and encoding the shape features, peak positions, slope change rate and periodicity of the curve segments, and pairing the encoded feature vectors with the corresponding accurate indicators to form an initial training dataset. Step S3042: The initial training dataset is randomly divided into a training subset, a validation subset, and a test subset, and data augmentation is performed on each subset, including adding Gaussian noise or slightly distorting the time axis. Step S3043: Select a supervised learning algorithm to construct an optimized and accurate analysis model, wherein the input layer receives feature vectors, the hidden layer captures the time series dependencies of the curves, and the output layer predicts the accurate index value. Step S3044: Iterative training is performed using a training subset, with mean squared error as the loss function. The model parameters are optimized through backpropagation until the loss on the verification subset converges or the preset number of iterations is reached. Step S3045: Evaluate the model performance on the test subset, calculate the root mean square error and coefficient of determination between the prediction accuracy index and the true label. If the coefficient of determination is greater than or equal to a preset threshold, the model is confirmed to be effective.
[0009] In some embodiments disclosed in this invention, the method for evaluating the optimization index of a set of simulated combustion parameter curves using the simulated combustion detection curves includes: Step S401: Based on the simulated combustion detection curve, extract key performance indicators in the simulated combustion process, including combustion temperature distribution, oxygen concentration change, pollutant emission concentration and thermal efficiency related parameters. Step S402: Standardize the extracted key performance indicators and calculate the deviation value of each indicator and the difference between the indicators under ideal combustion conditions. Step S403: Determine the preset index difference range to which the index difference amount belongs for each index, and then determine the corresponding optimization index.
[0010] In some embodiments disclosed in this invention, the method for determining a selected set of simulated combustion parameter curves by combining optimization indicators and accuracy indicators includes: Step S403: Determine whether the accuracy index of each simulated combustion parameter curve group is greater than or equal to the preset value. If it is greater than or equal to the preset value, retain the corresponding simulated combustion parameter curve group. Based on the order of the accuracy index, sort the retained simulated combustion parameter curve groups to obtain a simulated fuel consumption parameter curve group sequence. The simulated combustion parameter groups in the simulated combustion parameter curve group sequence are the selected simulated combustion parameter curve groups.
[0011] In some embodiments disclosed in this invention, the method for intelligent combustion control includes: Step S501: Map the selected set of simulated combustion parameter curves to the control parameter framework of the actual combustion system, including synchronizing and aligning the time nodes of the simulated parameter curve set with the real-time combustion task timeline, and adjusting the amplitude of the control parameters to adapt to the current environmental conditions. Step S502: During the combustion process, control commands are injected in real time, and the operation settings of the combustion equipment are dynamically adjusted according to the mapped control parameter curve. Step S503: Monitor the feedback data of the actual combustion detection parameters, compare them with the selected simulated combustion detection curve, calculate the real-time deviation value, and trigger the adaptive correction mechanism to minimize the deviation; Step S504: If the deviation value exceeds the preset threshold, a candidate curve is selected from the backup simulated combustion parameter curve group for switching and adjustment; otherwise, the instruction sequence of the current curve group continues to be executed.
[0012] In some embodiments disclosed in this invention, a smart combustion optimization system is also disclosed, comprising: The first module is used to acquire historical combustion data, perform time-series analysis on the historical combustion data, and configure the combustion control parameters and combustion detection parameters at different time points in the preset task timeline template to obtain the historical task timeline. Parallel to the historical task timeline, several historical combustion control parameter curves and historical combustion detection parameter curves are constructed. The second module is used to align the historical combustion control parameter curves and historical combustion detection parameter curves belonging to the same historical task timeline according to the alignment method of several historical task timelines, to obtain a group of historical combustion parameter curves, and to randomly split each group of historical combustion parameter curves several times to obtain several groups of historical combustion parameter curves. The third module is used to construct an optimized and accurate analysis model based on grouping several historical combustion parameter curves; The fourth module is used to construct several sets of simulated combustion parameter curves, evaluate the optimization index of the simulated combustion parameter curve sets using the simulated combustion detection curves, and analyze the accuracy index of the simulated combustion parameter curve sets using the optimization accuracy analysis model. Combining the optimization index and the accuracy index, the selected simulated combustion parameter curve sets are determined. The fifth module is used for intelligent combustion control based on the selected set of simulated combustion parameter curves.
[0013] This invention discloses a smart combustion optimization method and system, relating to the field of boiler combustion control optimization technology. The method includes: acquiring historical combustion data, performing time-series analysis, configuring control parameters and detection parameters on a task timeline template, constructing a historical task timeline and parallel parameter curves; aligning curve groups on the same timeline and randomly splitting them into diverse historical parameter curve groups; based on these groups, constructing an accurate optimization analysis model through curve variation adjustment, similarity comparison, and supervised training; generating simulated parameter curve groups, using detection curves to evaluate optimization indicators, and using the model to predict accurate indicators, comprehensively ranking and selecting the optimal group; mapping the selected group to the actual system to achieve real-time control, feedback monitoring, and adaptive switching. Compared to traditional methods, this invention achieves precise parameter optimization, reduces energy consumption and emissions, and improves system robustness.
[0014] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0015] Figure 1 This is a flowchart illustrating the steps of a smart combustion optimization method disclosed in an embodiment of the present invention. Detailed Implementation
[0016] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.
[0017] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. It should be understood that the preferred embodiments described herein are only for illustration and explanation of the present invention and should not be construed as limiting the scope of protection of the present invention. Those skilled in the art can make some non-essential improvements and adjustments based on the following content of the present invention. In the present invention, unless otherwise expressly specified and limited, the technical terms used in the present invention should have the ordinary meaning understood by those skilled in the art.
[0018] Example: This invention discloses a smart combustion optimization method, see reference. Figure 1 ,include: Step S100: Acquire historical combustion data, perform time-series analysis on the historical combustion data, and configure the combustion control parameters and combustion detection parameters at different time points in the preset task timeline template to obtain the historical task timeline. Parallel to the historical task timeline, construct several historical combustion control parameter curves and historical combustion detection parameter curves.
[0019] The core of step S100 lies in constructing a structured task timeline and parameter curves through time-series analysis of historical combustion data, thereby laying a data foundation for subsequent optimization. Specifically, the acquired historical combustion data first includes multi-dimensional parameters such as fuel injection rate, airflow, temperature distribution, and emission concentration. After time-series analysis, this data can reveal the dynamic evolution of the combustion process, such as periodic fluctuations in peak temperature or the cumulative trend of pollutant emissions. Subsequently, combustion control parameters at different time points, such as ignition timing and valve opening, and combustion detection parameters, such as oxygen concentration and thermal efficiency, are configured in a preset task timeline template. This template is similar to a standardized framework, defining the stages of the combustion cycle, such as preheating, stable combustion, and decay, thereby generating historical task timelines. The timeline not only preserves the temporal integrity of the original data but also facilitates comparison across datasets. The historical combustion control parameter curves and historical combustion detection parameter curves constructed parallel to this timeline are transformed from discrete data into continuous curves through interpolation or smoothing algorithms such as spline interpolation. These curves capture the coupling relationships between parameters, such as how the adjustment of control parameters affects the response lag of detection parameters, thus providing a visual and quantitative basis for subsequent alignment and splitting. Ultimately, this achieves the transformation from massive historical data to operable model input, ensuring that the input data of the optimization process has high fidelity and temporal consistency.
[0020] Step S200: Based on the alignment method of several historical task timelines, align the historical combustion control parameter curves and historical combustion detection parameter curves belonging to the same historical task timeline to obtain a group of historical combustion parameter curves. Then, randomly split each group of historical combustion parameter curves several times to obtain several groups of historical combustion parameter curves.
[0021] Step S200 ensures the diversity and representativeness of historical data through curve alignment and random splitting mechanisms, providing a robust sample set for model training. Specifically, the process first aligns historical combustion control parameter curves (e.g., fuel curves) and historical combustion detection parameter curves (e.g., temperature curves) on the same timeline using a dynamic time warping algorithm, eliminating offsets caused by differences in sampling frequency or operating conditions. This forms a set of historical combustion parameter curves, which represents the complete dynamic spectrum of the combustion process. Subsequently, each historical combustion parameter curve set is randomly split several times, for example, by randomly cutting the complete curve into sub-segments such as the first half of the cycle or the peak segment, generating multiple historical combustion parameter curve groups. This randomness introduces data variability, simulating noise and incompleteness in actual operating conditions, such as sensor failures or sudden load changes. The principle of the entire mechanism is to utilize the diversity of groupings to simulate potential optimization scenarios. Combined with subsequent variation adjustments as described in claim 2, this provides rich comparative benchmarks for similarity comparison and accuracy index calculation, ultimately improving the model's generalization ability to unknown operating conditions.
[0022] Step S300: Based on the grouping of several historical combustion parameter curves, construct an optimized and accurate analysis model.
[0023] Step S300 constructs an optimized and accurate analysis model based on the grouping of historical combustion parameter curves. The principle is to achieve a quantitative assessment of the accuracy of the simulated parameter curves through an iterative framework of variation comparison training, thereby bridging historical experience and future prediction. Specifically, firstly, the curve segments in each group are mutated and adjusted, such as by adding Gaussian noise, time warping, or amplitude scaling, as detailed in S301 of claim 2, to generate mutated combustion parameter curve groups to simulate the uncertainty in the optimization process. Then, these mutated groups are compared with different historical groups, and the similarity is calculated by the change in the cross-area. This method quantifies the spatiotemporal consistency of the curve shape by the average and variance of the cross-area per unit time. If the similarity is greater than or equal to a preset threshold, it is considered similar. The accuracy index is determined based on the highest and average values of the similarity, where the accuracy index is calculated by a weighted combination of the highest similarity and the average similarity, and the weighting points are dynamically adjusted according to the difference between the two. Finally, the mutated groups are used as training data, and the accuracy index is used as labels to train a supervised learning model, including feature extraction such as shape encoding, data augmentation, recurrent neural networks or long short-term memory network hidden layers to capture temporal dependencies, and backpropagation of mean squared error loss, until the coefficient of determination is greater than or equal to the threshold. This model essentially learns the accuracy boundary of historical patterns, providing a prediction engine for simulation evaluation and ensuring that the optimization does not deviate from the reliable historical trajectory.
[0024] Step S400: Construct several sets of simulated combustion parameter curves, evaluate the optimization index of the simulated combustion parameter curve sets using the simulated combustion detection curves, and analyze the accuracy index of the simulated combustion parameter curve sets using the optimization accuracy analysis model. Combine the optimization index and the accuracy index to determine the selected set of simulated combustion parameter curves.
[0025] The principle of step S400 lies in achieving efficient parameter curve optimization and selection through simulation generation, dual-index evaluation, and comprehensive screening, integrating performance optimization and historical accuracy to avoid blind simulation. Specifically, the process first constructs several sets of simulated combustion parameter curves, for example, using genetic algorithms or particle swarm optimization to generate candidate sets from the initial historical curves, such as random perturbations of control parameters. Then, simulated combustion detection curves, such as temperature, oxygen, and emission curves derived through physical models like computational fluid dynamics simulations, are used to evaluate optimization indicators, extracting key performance aspects such as combustion temperature distribution uniformity, peak pollutant concentration, and thermal efficiency deviation. Standardized interval mapping is then used, such as low-deviation intervals corresponding to high-optimization fractional combustion efficiency and environmental friendliness. Simultaneously, the constructed optimized accuracy analysis model is input into the simulation set, outputting accurate indicators, and evaluating its matching degree with historical patterns, such as prediction deviation being less than a threshold. Finally, combining the two, if the accurate indicators are greater than or equal to preset values, they are retained and sorted sequentially to form a sequence. The ingenuity of this mechanism lies in the fact that optimization indicators drive performance improvement, while accurate indicators anchor realistic feasibility, avoiding high-performance but low-reliability phantom schemes, thereby generating a set of parameter curves balancing efficiency, emissions, and stability, providing high-confidence guidance for actual regulation.
[0026] Step S500: Intelligent combustion control is performed based on the selected set of simulated combustion parameter curves.
[0027] Step S500 performs intelligent combustion control based on a selected set of simulated combustion parameter curves. Its principle is to achieve a seamless transition and dynamic optimization from simulation to reality through mapping synchronization, real-time injection, and adaptive feedback closed-loop. Specifically, firstly, the selected curve set is mapped to the actual system control framework. Time-node synchronization, such as Kalman filtering to align the task timeline, and amplitude adjustments, such as scaling factors based on the current environment (e.g., humidity compensation), ensure the parameter curves are integrated with real-time operating conditions. Subsequently, during combustion, real-time control commands are injected, such as a proportional-integral-derivative controller dynamically setting the fuel rate and air-to-air ratio, allowing the equipment to follow the curve trajectory. Simultaneously, the deviation between actual detection parameter feedback, such as sensor data streams and simulated detection curves (e.g., Euclidean distance or mean absolute error), is monitored, triggering an adaptive correction deviation minimization mechanism. The curve is fine-tuned through model re-prediction. If the deviation exceeds a threshold, a backup group is switched from the standby sequence. The essence of the entire closed-loop is to utilize historically driven simulations as priors, combined with real-time feedback to form a model predictive control framework. This not only improves combustion stability and efficiency but also reduces emission risks and ensures the system's robustness under varying operating conditions.
[0028] In some embodiments disclosed in this invention, the method for constructing an optimized and accurate analysis model includes: Step S301: For each historical combustion parameter curve group, the historical combustion control parameter curve segment and the historical combustion detection parameter curve segment are variably adjusted to obtain several variable combustion parameter curve groups.
[0029] Step S302: Compare each group of variable combustion parameter curves with different groups of historical combustion parameter curves to identify several similar groups of historical combustion parameter curves. The method for determining similarity includes determining the degree of similarity between them. If the degree of similarity is greater than or equal to a preset value, the corresponding group of historical combustion parameter curves is identified as similar.
[0030] Step S303: Based on the degree of similarity corresponding to several similar historical combustion parameter curve groups, determine the accurate index of the variable combustion parameter curve grouping.
[0031] Step S304: Group the variable combustion parameter curves into training data, use the accuracy index as the training label, train the model, and obtain the optimized accuracy analysis model.
[0032] In some embodiments disclosed in this invention, the method for determining the similarity between a group of variable combustion detection parameter curves and a group of historical combustion parameter curves includes: Step S3021: Determine the intersection range between each relative parameter curve segment, and as time progresses, determine the change in the intersection range area, including the unit intersection area per unit time corresponding to different time nodes, and the average unit intersection area. Step S3022: Determine the degree of similarity based on the change in the area of the intersection range between the relative parameter curve segments; The expression for calculating the degree of similarity is: ; Where S represents the degree of similarity. Let L be the average unit cross area, L be the normalization adjustment coefficient between the average unit cross area and the variance of the unit cross area, K be the adjustment coefficient for the influence of the variance of the unit cross area, b be the adjustment constant for the influence of the variance of the unit cross area, and T be the total number of time points involved in the alignment. Let be the unit cross area at time point t. To preset the maximum average unit cross area, This is the variance of the preset maximum unit cross area.
[0033] S represents the similarity level, used to indicate the overall similarity between two relative parameter curve segments (variation curve and historical curve) in the time series dimension.
[0034] The maximum average cross area per unit is a preset constant, which serves as the maximum average cross area as an ideal or historical benchmark.
[0035] L is the normalization adjustment coefficient between the average unit cross area and the variance of the unit cross area, a dimensionless adjustment parameter used to balance the relative weights of the linear deviation term and the exponential growth term.
[0036] K is the adjustment factor for the effect of variance per unit cross area, a positive amplification factor used to enhance the sensitivity of the variance penalty.
[0037] To predetermine the variance of the maximum unit cross area, a preset constant is used, representing the upper limit of the maximum allowable variance.
[0038] b is an adjustment constant for the effect of variance per unit cross area, used to adjust the rise of variance in the exponential function.
[0039] In some embodiments disclosed in this invention, the method for determining accurate indices for grouping variable combustion parameter curves includes: Step S3031: Analyze the determined similarity levels to determine the highest similarity level and the average similarity level, and determine the accurate index based on the highest similarity level and the average similarity level. The expression for calculating the accuracy index is as follows: ; Where Q is the accuracy indicator. For the highest degree of similarity, To achieve the average similarity, The weighted number of points representing the highest similarity. The weighted number of points representing the average similarity is denoted by , where and The similarity is determined based on the difference between the highest similarity and the average similarity. The determination method includes determining the preset similarity difference interval to which the difference belongs, and determining the corresponding weight allocation points based on the preset similarity difference interval.
[0040] Q is the accuracy index, representing the overall accuracy score of the variable combustion parameter curve group relative to the historical group.
[0041] The weighted allocation of points for the highest similarity level is a dynamic weighting coefficient that represents the proportion of points allocated to the highest similarity level.
[0042] In some embodiments disclosed in this invention, the method for training a model by grouping variable combustion parameter curves into training data and using accurate indicators as training labels includes: Step S3041 involves feature extraction and preprocessing of the variable combustion parameter curves, including quantizing and encoding the shape features, peak positions, slope change rates, and periodic indicators of the curve segments, and pairing the encoded feature vectors with the corresponding accurate indicators to form an initial training dataset.
[0043] Step S3042: The initial training dataset is randomly divided into a training subset, a validation subset, and a test subset, and data augmentation is performed on each subset, including adding Gaussian noise or slightly distorting the time axis.
[0044] Step S3043: Select a supervised learning algorithm to construct an optimized and accurate analysis model, wherein the input layer receives feature vectors, the hidden layer captures the time series dependencies of the curves, and the output layer predicts the accurate index value.
[0045] Step S3044: Iterative training is performed using a training subset, with mean squared error as the loss function. The model parameters are optimized using the backpropagation algorithm until the loss on the verification subset converges or the preset number of iterations is reached.
[0046] Step S3045: Evaluate the model performance on the test subset, calculate the root mean square error and coefficient of determination between the prediction accuracy index and the true label. If the coefficient of determination is greater than or equal to a preset threshold, the model is confirmed to be effective.
[0047] In some embodiments disclosed in this invention, the method for evaluating the optimization index of a set of simulated combustion parameter curves using the simulated combustion detection curves includes: Step S401: Based on the simulated combustion detection curve, extract key performance indicators in the simulated combustion process, including combustion temperature distribution, oxygen concentration change, pollutant emission concentration and thermal efficiency related parameters.
[0048] Step S402: Standardize the extracted key performance indicators and calculate the deviation value of each indicator and the difference between the indicators under ideal combustion conditions.
[0049] Step S403: Determine the preset index difference range to which the index difference amount belongs for each index, and then determine the corresponding optimization index.
[0050] In some embodiments disclosed in this invention, the method for determining a selected set of simulated combustion parameter curves by combining optimization indicators and accuracy indicators includes: Step S403: Determine whether the accuracy index of each simulated combustion parameter curve group is greater than or equal to the preset value. If it is greater than or equal to the preset value, retain the corresponding simulated combustion parameter curve group. Based on the order of the accuracy index, sort the retained simulated combustion parameter curve groups to obtain a simulated fuel consumption parameter curve group sequence. The simulated combustion parameter groups in the simulated combustion parameter curve group sequence are the selected simulated combustion parameter curve groups.
[0051] In some embodiments disclosed in this invention, the method for intelligent combustion control includes: Step S501: Map the selected set of simulated combustion parameter curves to the control parameter framework of the actual combustion system, including synchronizing and aligning the time nodes of the simulated parameter curve set with the real-time combustion task timeline, and adjusting the amplitude of the control parameters to adapt to the current environmental conditions.
[0052] Step S502: During the combustion process, control commands are injected in real time, and the operation settings of the combustion equipment are dynamically adjusted according to the mapped control parameter curve.
[0053] Step S503: Monitor the feedback data of the actual combustion detection parameters, compare it with the selected simulated combustion detection curve, calculate the real-time deviation value, and trigger the adaptive correction mechanism to minimize the deviation.
[0054] Step S504: If the deviation value exceeds the preset threshold, a candidate curve is selected from the backup simulated combustion parameter curve group for switching and adjustment; otherwise, the instruction sequence of the current curve group continues to be executed.
[0055] In some embodiments disclosed in this invention, a smart combustion optimization system is also disclosed, comprising: The first module is used to acquire historical combustion data, perform time-series analysis on the historical combustion data, and configure the combustion control parameters and combustion detection parameters at different time points in the preset task timeline template to obtain the historical task timeline. Parallel to the historical task timeline, several historical combustion control parameter curves and historical combustion detection parameter curves are constructed.
[0056] The second module is used to align the historical combustion control parameter curves and historical combustion detection parameter curves belonging to the same historical task timeline according to the alignment method of several historical task timelines, to obtain a group of historical combustion parameter curves, and to randomly split each group of historical combustion parameter curves several times to obtain several groups of historical combustion parameter curves.
[0057] The third module is used to construct an optimized and accurate analysis model based on grouping several historical combustion parameter curves; The fourth module is used to construct several sets of simulated combustion parameter curves, evaluate the optimization index of the simulated combustion parameter curve sets using the simulated combustion detection curves, and analyze the accuracy index of the simulated combustion parameter curve sets using the optimization accuracy analysis model. Combining the optimization index and the accuracy index, the selected simulated combustion parameter curve sets are determined.
[0058] The fifth module is used for intelligent combustion control based on the selected set of simulated combustion parameter curves.
[0059] This invention discloses a smart combustion optimization method and system, relating to the field of boiler combustion control optimization technology. The method includes: acquiring historical combustion data, performing time-series analysis, configuring control parameters and detection parameters on a task timeline template, constructing a historical task timeline and parallel parameter curves; aligning curve groups on the same timeline and randomly splitting them into diverse historical parameter curve groups; based on these groups, constructing an accurate optimization analysis model through curve variation adjustment, similarity comparison, and supervised training; generating simulated parameter curve groups, using detection curves to evaluate optimization indicators, and using the model to predict accurate indicators, comprehensively ranking and selecting the optimal group; mapping the selected group to the actual system to achieve real-time control, feedback monitoring, and adaptive switching. Compared to traditional methods, this invention achieves precise parameter optimization, reduces energy consumption and emissions, and improves system robustness.
[0060] Through the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented in hardware or by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.
[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A smart combustion optimization method, characterized in that, include: Step S100: Obtain historical combustion data, perform time series analysis on the historical combustion data, and configure the combustion control parameters and combustion detection parameters at different time points in the preset task timeline template to obtain the historical task timeline. Parallel to the historical task timeline, construct several historical combustion control parameter curves and historical combustion detection parameter curves. Step S200: Based on the alignment method of several historical task timelines, align the historical combustion control parameter curves and historical combustion detection parameter curves belonging to the same historical task timeline to obtain a group of historical combustion parameter curves. Then, randomly split each group of historical combustion parameter curves several times to obtain several groups of historical combustion parameter curves. Step S300: Based on grouping several historical combustion parameter curves, construct an optimized and accurate analysis model; Step S400: Construct several sets of simulated combustion parameter curves, evaluate the optimization index of the simulated combustion parameter curve sets using the simulated combustion detection curves, and analyze the accuracy index of the simulated combustion parameter curve sets using the optimization accuracy analysis model. Combine the optimization index and the accuracy index to determine the selected set of simulated combustion parameter curves. Step S500: Intelligent combustion control is performed based on the selected set of simulated combustion parameter curves.
2. The intelligent combustion optimization method according to claim 1, characterized in that, Methods for constructing optimized and accurate analytical models include: Step S301: Perform variation adjustment on the historical combustion control parameter curve segment and the historical combustion detection parameter curve segment in each historical combustion parameter curve group to obtain several variation combustion parameter curve groups. Step S302: Compare each group of variable combustion parameter curves with different groups of historical combustion parameter curves to identify several similar groups of historical combustion parameter curves. The method for determining similarity includes determining the degree of similarity between them. If the degree of similarity is greater than or equal to a preset value, the corresponding group of historical combustion parameter curves is identified as similar. Step S303: Based on the similarity of several similar historical combustion parameter curves, determine the accurate index of the variable combustion parameter curve grouping. Step S304: Group the variable combustion parameter curves into training data, use the accuracy index as the training label, train the model, and obtain the optimized accuracy analysis model.
3. The intelligent combustion optimization method according to claim 2, characterized in that, Methods for determining the similarity between groups of variable combustion detection parameter curves and groups of historical combustion parameter curves include: Step S3021: Determine the intersection range between each relative parameter curve segment, and as time progresses, determine the change in the intersection range area, including the unit intersection area per unit time corresponding to different time nodes, and the average unit intersection area. Step S3022: Determine the degree of similarity based on the change in the area of the intersection range between the relative parameter curve segments; The expression for calculating the degree of similarity is: ； Where S represents the degree of similarity. Let L be the average unit cross area, L be the normalization adjustment coefficient between the average unit cross area and the variance of the unit cross area, K be the adjustment coefficient for the influence of the variance of the unit cross area, b be the adjustment constant for the influence of the variance of the unit cross area, and T be the total number of time points involved in the alignment. Let be the unit cross area at time point t. To preset the maximum average unit cross area, This is the variance of the preset maximum unit cross area.
4. The intelligent combustion optimization method according to claim 2, characterized in that, Methods for determining accurate indices for grouping variable combustion parameter curves include: Step S3031: Analyze the determined similarity levels to determine the highest similarity level and the average similarity level, and determine the accurate index based on the highest similarity level and the average similarity level. The expression for calculating the accuracy index is as follows: ； Where Q is the accuracy indicator. For the highest degree of similarity, To achieve the average similarity, The weighted number of points representing the highest similarity. The weighted number of points representing the average similarity is denoted by , where and The similarity is determined based on the difference between the highest similarity and the average similarity. The determination method includes determining the preset similarity difference interval to which the difference belongs, and determining the corresponding weight allocation points based on the preset similarity difference interval.
5. The intelligent combustion optimization method according to claim 2, characterized in that, Methods for training models by grouping variable combustion parameter curves into training data and using accurate metrics as training labels include: Step S3041: Feature extraction and preprocessing are performed on the variable combustion parameter curves, including quantizing and encoding the shape features, peak positions, slope change rate and periodicity of the curve segments, and pairing the encoded feature vectors with the corresponding accurate indicators to form an initial training dataset. Step S3042: The initial training dataset is randomly divided into a training subset, a validation subset, and a test subset, and data augmentation is performed on each subset, including adding Gaussian noise or slightly distorting the time axis. Step S3043: Select a supervised learning algorithm to construct an optimized and accurate analysis model, wherein the input layer receives feature vectors, the hidden layer captures the time series dependencies of the curves, and the output layer predicts the accurate index value. Step S3044: Iterative training is performed using a training subset, with mean squared error as the loss function. The model parameters are optimized through backpropagation until the loss on the verification subset converges or the preset number of iterations is reached. Step S3045: Evaluate the model performance on the test subset, calculate the root mean square error and coefficient of determination between the prediction accuracy index and the true label. If the coefficient of determination is greater than or equal to a preset threshold, the model is confirmed to be effective.
6. The intelligent combustion optimization method according to claim 1, characterized in that, Methods for evaluating the optimization indices of simulated combustion parameter curve sets using simulated combustion detection curves include: Step S401: Based on the simulated combustion detection curve, extract key performance indicators in the simulated combustion process, including combustion temperature distribution, oxygen concentration change, pollutant emission concentration and thermal efficiency related parameters. Step S402: Standardize the extracted key performance indicators and calculate the deviation value of each indicator and the difference between the indicators under ideal combustion conditions. Step S403: Determine the preset index difference range to which the index difference amount belongs for each index, and then determine the corresponding optimization index.
7. The intelligent combustion optimization method according to claim 1, characterized in that, The methods for determining the selected set of simulated combustion parameter curves by combining optimization and accuracy indicators include: Step S403: Determine whether the accuracy index of each simulated combustion parameter curve group is greater than or equal to the preset value. If it is greater than or equal to the preset value, retain the corresponding simulated combustion parameter curve group. Based on the order of the accuracy index, sort the retained simulated combustion parameter curve groups to obtain a simulated fuel consumption parameter curve group sequence. The simulated combustion parameter groups in the simulated combustion parameter curve group sequence are the selected simulated combustion parameter curve groups.
8. The intelligent combustion optimization method according to claim 1, characterized in that, Methods for intelligent combustion control include: Step S501: Map the selected set of simulated combustion parameter curves to the control parameter framework of the actual combustion system, including synchronizing and aligning the time nodes of the simulated parameter curve set with the real-time combustion task timeline, and adjusting the amplitude of the control parameters to adapt to the current environmental conditions. Step S502: During the combustion process, control commands are injected in real time, and the operation settings of the combustion equipment are dynamically adjusted according to the mapped control parameter curve. Step S503: Monitor the feedback data of the actual combustion detection parameters, compare them with the selected simulated combustion detection curve, calculate the real-time deviation value, and trigger the adaptive correction mechanism to minimize the deviation; Step S504: If the deviation value exceeds the preset threshold, a candidate curve is selected from the backup simulated combustion parameter curve group for switching and adjustment; otherwise, the instruction sequence of the current curve group continues to be executed.
9. A smart combustion optimization system, characterized in that, The method for performing the intelligent combustion optimization method according to any one of claims 1-8 includes: The first module is used to acquire historical combustion data, perform time-series analysis on the historical combustion data, and configure the combustion control parameters and combustion detection parameters at different time points in the preset task timeline template to obtain the historical task timeline. Parallel to the historical task timeline, several historical combustion control parameter curves and historical combustion detection parameter curves are constructed. The second module is used to align the historical combustion control parameter curves and historical combustion detection parameter curves belonging to the same historical task timeline according to the alignment method of several historical task timelines, to obtain a group of historical combustion parameter curves, and to randomly split each group of historical combustion parameter curves several times to obtain several groups of historical combustion parameter curves. The third module is used to construct an optimized and accurate analysis model based on grouping several historical combustion parameter curves; The fourth module is used to construct several sets of simulated combustion parameter curves, evaluate the optimization index of the simulated combustion parameter curve sets using the simulated combustion detection curves, and analyze the accuracy index of the simulated combustion parameter curve sets using the optimization accuracy analysis model. Combining the optimization index and the accuracy index, the selected simulated combustion parameter curve sets are determined. The fifth module is used for intelligent combustion control based on the selected set of simulated combustion parameter curves.