A method and system for optimizing the control of a furnace

Abstract

The application discloses a heating furnace control optimization method and system, and relates to the technical field of heating furnace control.The method comprises the following steps: obtaining working condition data of a heating furnace, and pre-processing and marking the working condition data to obtain energy consumption data corresponding to each working condition; calculating a weighted similarity measure under different working conditions based on the working condition data and the energy consumption data, and generating a preliminary control strategy corresponding to different working conditions based on the weighted similarity measure; constructing a heterogeneous network comprising the working condition data and the preliminary control strategy, and walking in the heterogeneous network based on a bidirectional random walk algorithm to generate a candidate control strategy; and determining a final control strategy corresponding to each working condition based on the candidate control strategy and a chaos optimization technique.The application mines the complex relationship between the working condition data and the control strategy, provides more information for optimizing the control strategy, generates a candidate control strategy with more globality based on the bidirectional random walk algorithm, and improves the comprehensiveness of strategy selection.

Description

Technical Field

[0001] This invention relates to the field of heating furnace control technology, and more specifically, to a heating furnace control optimization method and system. Background Technology

[0002] A heating furnace is a device used to heat materials or fluids, widely used in industrial production. Its main function is to transfer heat to the object being heated by burning fuels (such as natural gas, coal, or oil) or using electrical energy. Energy-saving furnace control refers to optimizing the furnace's operating parameters and control strategies to reduce energy consumption and improve thermal efficiency. The goal of energy-saving furnace control is to minimize energy consumption and environmental pollution while meeting production needs and quality requirements.

[0003] Collecting, storing, analyzing, and processing the large amounts of data generated during the operation of a heating furnace can optimize the furnace's control strategy and achieve more efficient energy-saving results. For example, Chinese Patent Application No. 202211115824.4 discloses a heating furnace combustion intelligent control method and device based on a big data cloud platform. This method integrates big data mining technology, intelligent algorithms, and traditional mechanism models to construct an intelligent control system for heating furnace combustion, thereby improving the automatic control level of furnace temperature, the uniformity of slab heating temperature, and the accuracy of furnace gas temperature control.

[0004] However, the above control method still has the following shortcomings: the method cannot effectively construct heterogeneous networks that include different working conditions, environmental conditions and control strategies, and it is difficult to analyze and optimize complex relationships, which makes it difficult to understand and analyze the optimal control strategy under different working conditions and environmental conditions in actual operation.

[0005] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0006] In view of the problems in the related technologies, the present invention proposes a heating furnace control optimization method and system to overcome the above-mentioned technical problems existing in the existing related technologies.

[0007] Therefore, the specific technical solution adopted by the present invention is as follows:

[0008] According to one aspect of the present invention, a method for optimizing the control of a heating furnace is provided, comprising:

[0009] S1. Obtain the operating data of the heating furnace, and preprocess and label the operating data to obtain the energy consumption data corresponding to each operating condition.

[0010] S2. Based on operating condition data and energy consumption data, calculate the weighted similarity metric under different operating conditions, and generate preliminary control strategies corresponding to different operating conditions based on the weighted similarity metric.

[0011] S3. Construct a heterogeneous network using current operating condition data, preliminary control strategy, historical operating condition data, and historical energy consumption data as nodes; define the number of walk steps and the number of walks, and perform a bidirectional random walk starting from a current operating condition data node and a preliminary control strategy node; obtain the walk path generated by the bidirectional random walk, and record the sequence of nodes visited during the walk; generate candidate control strategies based on the frequency and stability of the walk path.

[0012] S4. Based on candidate control strategies and chaotic optimization techniques, determine the final control strategy corresponding to each operating condition.

[0013] Furthermore, the operating condition data of the heating furnace is acquired, and the data is preprocessed and labeled to obtain the energy consumption data corresponding to each operating condition, including:

[0014] S11. Collect the operating data of the heating furnace, including the furnace's operating data and environmental data;

[0015] S12. Use data preprocessing techniques to clean and calibrate operational and environmental data.

[0016] S13. Label the collected operating data and record the corresponding energy consumption data.

[0017] Furthermore, based on operating condition data and energy consumption data, a weighted similarity metric is calculated for different operating conditions, and preliminary control strategies corresponding to different operating conditions are generated based on the weighted similarity metric, including:

[0018] S21. Extract the operational data features and environmental data features from the operating condition data, and calculate the statistics for each operational data feature and environmental data feature.

[0019] S22. Extract energy consumption features from energy consumption data and calculate the statistics of energy consumption data;

[0020] S23. Integrate the operating condition characteristics and corresponding energy consumption characteristics into overall characteristics, calculate the similarity between different overall characteristics, form a comprehensive similarity measure, assign weights to the operating condition characteristics and energy consumption characteristics, and calculate a weighted similarity measure.

[0021] S24. Based on weighted similarity measurement, historical operating condition data and historical energy consumption data, generate a preliminary control strategy for each operating condition.

[0022] Furthermore, the operating condition characteristics and corresponding energy consumption characteristics are integrated into overall characteristics. The similarity between different overall characteristics is calculated to form a comprehensive similarity measure. Weights are assigned to the operating condition characteristics and energy consumption characteristics. The weighted similarity measure is calculated as follows:

[0023] S231. Standardize the operational data characteristics, environmental data characteristics, and energy consumption characteristics using the calculated statistical quantities;

[0024] S232. Combine the operating data features and environmental data features into operating condition features, and integrate the operating condition features and corresponding energy consumption features into overall features, and use a similarity measurement algorithm to calculate the similarity between different overall features.

[0025] S233. Assign weights to the operating condition characteristics and energy consumption characteristics based on the statistical measures, and calculate the weighted similarity measure under different operating conditions;

[0026] The overall characteristics include the temperature, pressure, flow rate, external temperature, humidity, fuel consumption, and electrical energy consumption of the heating furnace;

[0027] When assigning weights, the mean and standard deviation of each feature in the overall feature set are calculated, and the standard deviation of the feature is divided by the mean of the feature to obtain the coefficient of variation of each feature; weights are assigned to each feature based on the coefficient of variation.

[0028] Furthermore, based on weighted similarity metrics, historical operating condition data, and historical energy consumption data, preliminary control strategies are generated for each operating condition, including:

[0029] S241. Based on weighted similarity measurement, select the historical operating condition data that is most similar to the current operating condition;

[0030] S242. Extract the control strategy corresponding to the most similar historical operating condition data, and use it as the preliminary control strategy for the current operating condition.

[0031] Furthermore, constructing a heterogeneous network using current operating data, preliminary control strategies, historical operating data, and historical energy consumption data as nodes includes:

[0032] S311. Obtain the node types in the heterogeneous network, including current operating condition data nodes, preliminary control strategy nodes, historical operating condition data nodes, and historical energy consumption data nodes.

[0033] S312. Based on the weighted similarity and correlation between nodes, determine the weights of the edges between the current operating condition data node and the historical operating condition data node, the edges between the preliminary control strategy node and the historical energy consumption data node, and the edges between the historical operating condition data node and the historical energy consumption data node.

[0034] Furthermore, the number of walk steps and the number of walks are defined. A bidirectional random walk starting from a current operating condition data node and a preliminary control strategy node includes:

[0035] Determine the number of steps and the number of walks, and select a current working condition data node and a preliminary control strategy node as the starting point for the walk;

[0036] Define the direction of the walk, the probability transfer mechanism, and the self-starting coefficient to achieve bidirectional random walk;

[0037] In the bidirectional random walk, the number of times each node is visited is recorded, nodes with a visit frequency higher than the threshold are identified, and the frequency and consistency of nodes appearing in multiple walks are analyzed. Based on frequent and stable walk paths, candidate control strategies are generated.

[0038] Furthermore, based on candidate control strategies and chaotic optimization techniques, the final control strategy for each operating condition is determined as follows:

[0039] S41. Determine the optimization objective and construct an evaluation function based on the optimization objective;

[0040] S42. Optimize the initial population using a chaotic optimization algorithm, and evaluate the performance of each individual based on the evaluation function;

[0041] S43. Update the position of individuals based on differences in fitness and attractiveness among them;

[0042] S44. Construct search parameters that vary with the number of iterations;

[0043] S45. After each iteration, sort the individuals according to their fitness.

[0044] S46. After multiple iterations, select the individual with the highest fitness as the final control strategy for each working condition.

[0045] The formula for the evaluation function is as follows:

[0046] ;

[0047] In the formula, E en This indicates the actual energy consumption value of the heating furnace; E eff This represents the reciprocal of the actual efficiency value of the heating furnace; E t This indicates the actual heating time of the heating furnace;

[0048] w 1, w 2, w 3 represents the weighting coefficients corresponding to the actual energy consumption value of the heating furnace, the reciprocal of the actual efficiency value of the heating furnace, and the actual heating time of the heating furnace, respectively.

[0049] Furthermore, optimizing the initial population using chaotic optimization algorithms includes:

[0050] S421. Set the control parameters of the piecewise linear chaotic mapping algorithm, use the piecewise linear chaotic mapping algorithm to optimize the initial population, and generate a chaotic sequence;

[0051] S422. Combine the chaotic sequence with the individuals in the initial population, and update the position of the individuals through chaotic mapping.

[0052] According to another aspect of the present invention, a heating furnace control optimization system is also provided, including a data acquisition module, a preliminary screening module, a global selection module, and an optimization module; wherein the data acquisition module, the preliminary screening module, the global selection module, and the optimization module are sequentially connected.

[0053] The data acquisition module is used to acquire the operating condition data of the heating furnace, and to preprocess and label the operating condition data to obtain the energy consumption data corresponding to each operating condition.

[0054] The preliminary screening module is used to calculate the weighted similarity metric under different operating conditions based on operating condition data and energy consumption data, and to generate preliminary control strategies corresponding to different operating conditions based on the weighted similarity metric.

[0055] The global selection module is used to construct a heterogeneous network using current operating condition data, preliminary control strategy, historical operating condition data, and historical energy consumption data as nodes; it defines the number of walk steps and the number of walks, and performs a bidirectional random walk starting from a certain current operating condition data node and preliminary control strategy node; it obtains the walk path generated by the bidirectional random walk, records the sequence of nodes visited during the walk, and generates candidate control strategies based on the frequency and stability of the walk path.

[0056] The optimization module is used to optimize and determine the final control strategy for each operating condition based on the candidate control strategies.

[0057] The beneficial effects of this invention are as follows:

[0058] (1) The present invention provides a heating furnace control optimization method and system, which identifies similar operating conditions by calculating the weighted similarity metric between different operating conditions, provides a basis for generating a preliminary control strategy, generates a preliminary control strategy based on similar operating conditions, improves the rationality of the initial strategy, and reduces blindness; by constructing a heterogeneous network, it explores the complex relationship between operating condition data and control strategy, provides more information for optimizing the control strategy, and generates a more global candidate control strategy based on a bidirectional random walk algorithm, improving the comprehensiveness of strategy selection.

[0059] (2) This invention not only considers current operating data and preliminary control strategies, but also introduces historical operating data and historical energy consumption data, constructing a heterogeneous network containing multiple node types. This more comprehensively captures the complex relationships between different operating conditions, control strategies, and energy consumption data, helping to discover potential correlations and patterns, and improving the rationality and effectiveness of control strategies. The bidirectional random walk algorithm can explore multiple times and in multiple directions in the heterogeneous network, generating diverse candidate control strategies and avoiding local optima. Combining similarity measurement, path frequency, and stability as multi-dimensional evaluation criteria, more reliable and comprehensive candidate control strategies are generated.

[0060] (3) The present invention uses an optimization algorithm to evaluate the generated candidate control strategies, calculates their scores under different evaluation indicators, and selects the control strategy with the highest fitness value as the final control strategy based on the optimization results. This ensures that the selected control strategy takes into account energy saving, combustion efficiency and heating time, and guarantees the economy of the final control strategy for each working condition. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is a flowchart of a furnace control optimization method according to an embodiment of the present invention;

[0063] Figure 2 This is a schematic diagram of a heating furnace control optimization system according to an embodiment of the present invention.

[0064] In the picture:

[0065] 1. Data acquisition module; 2. Preliminary screening module; 3. Global selection module; 4. Optimization module. Detailed Implementation

[0066] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0067] According to an embodiment of the present invention, a method and system for optimizing the control of a heating furnace are provided.

[0068] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, according to an embodiment of the present invention, a method for optimizing the control of a heating furnace is provided, comprising:

[0069] S1. Obtain the operating data of the heating furnace, and preprocess and label the operating data to obtain the energy consumption data corresponding to each operating condition. The operating data includes the operating data of the heating furnace and the environmental data.

[0070] In a further embodiment, the operating condition data of the heating furnace is acquired, and the operating condition data is preprocessed and labeled to obtain the energy consumption data corresponding to each operating condition, including:

[0071] S11. Collect operating and environmental data from the heating furnace. Operating data includes temperature, pressure, and flow rate, while environmental data includes external temperature and humidity. High-precision temperature, pressure, flow, and humidity sensors are used to collect the corresponding data, ensuring accuracy and real-time performance. The collected data is transmitted in real-time to a central database for storage and processing via wired or wireless networks. S12. Utilize data preprocessing techniques to clean and calibrate the operating and environmental data. Data cleaning involves processing the raw data to remove noise, outliers, and missing values, ensuring data quality. Data calibration involves correcting and adjusting the collected data to ensure consistency and accuracy. S13. Label the collected operating data and record the corresponding energy consumption data. Energy consumption data is collected in real-time by energy metering devices installed on the heating furnace, such as fuel flow meters and electricity meters.

[0072] S2. Based on operating condition data and energy consumption data, calculate the weighted similarity metric under different operating conditions, and generate preliminary control strategies corresponding to different operating conditions based on the weighted similarity metric.

[0073] In a further embodiment, based on operating condition data and energy consumption data, a weighted similarity metric is calculated for different operating conditions, and a preliminary control strategy corresponding to different operating conditions is generated based on the weighted similarity metric, including:

[0074] S21. Extract operational and environmental data features from the operating condition data, and calculate the mean, variance, maximum, and minimum values ​​for each feature. S22. Extract energy consumption features from the energy consumption data, such as energy consumption per unit time and energy consumption per unit output, and calculate energy consumption statistics, such as mean and variance. S23. Integrate the operating condition features and corresponding energy consumption features into a holistic feature set, calculate the similarity between different holistic features to form a comprehensive similarity measure, assign weights to the operating condition features and energy consumption features, and calculate a weighted similarity measure to obtain a more accurate similarity result. This integration method comprehensively reflects the relationship between operating conditions and energy consumption. Simultaneously, similarity measurement algorithms, such as Euclidean distance and cosine similarity, are used to calculate the similarity between different holistic features. S24. Based on the weighted similarity measure, historical operating condition data, and historical energy consumption data, generate a preliminary control strategy corresponding to each operating condition. This ensures that the selected historical operating condition data are similar not only in operating condition features but also in energy consumption features.

[0075] In a further embodiment, the operating condition characteristics and corresponding energy consumption characteristics are integrated into a holistic feature, the similarity between different holistic features is calculated to form a comprehensive similarity measure, and weights are assigned to the operating condition characteristics and energy consumption characteristics. The calculation of the weighted similarity measure includes:

[0076] S231. Standardize the operational data features, environmental data features, and energy consumption features using the calculated statistics. Standardization methods include Z-score standardization, which converts the data into a distribution with a mean of 0 and a standard deviation of 1, and Min-Max standardization, which scales the data to the [0, 1] interval. S232. Combine the operational data features and environmental data features into operating condition features, and integrate the operating condition features and corresponding energy consumption features into overall features. Calculate the similarity between different overall features using a similarity measurement algorithm. S233. Assign weights to the operating condition features and energy consumption features based on the statistics, and calculate the weighted similarity measure under different operating conditions. For example, calculate the coefficient of variation (COP) of the operating condition features and energy consumption features. The COP is the ratio of the variance to the mean of the statistics. Use the COP as the basis for weight allocation; the larger the COP, the higher the weight.

[0077] In a further embodiment, based on weighted similarity measurement, historical operating condition data, and historical energy consumption data, the preliminary control strategy corresponding to each operating condition is generated as follows:

[0078] S241. Based on weighted similarity measurement, select the historical operating condition data most similar to the current operating condition, ensuring that the selected historical operating condition data are similar not only in operating condition characteristics but also in energy consumption characteristics. S242. Extract the control strategy corresponding to the most similar historical operating condition data as the preliminary control strategy for the current operating condition, thus ensuring that the preliminary control strategy has high reference value and practical feasibility.

[0079] In this invention, operating condition characteristics and energy consumption characteristics are integrated into a unified feature, including temperature, pressure, flow rate, external temperature, humidity, fuel consumption, and electrical energy consumption. The coefficient of variation is used to assign weights to these features. A larger coefficient of variation indicates greater volatility and variability in the feature, resulting in more changes under different operating conditions; therefore, it is assigned a higher weight. The coefficient of variation can be used directly as a weight, or it can be normalized before being used as a weight.

[0080] For example, the following weights are assigned to each feature: temperature, 0.014; pressure, 0.02; flow rate, 0.05; external temperature, 0.04; humidity, 0.083; fuel consumption, 0.04; and electrical energy consumption, 0.05. Based on a weighted similarity metric, the historical operating condition data most similar to the current operating condition is selected, and the control strategy corresponding to the most similar historical operating condition data is extracted. The corresponding control strategy is to adjust the fuel flow rate to reduce fuel consumption to the target value. The temperature control parameters of the heating furnace are adjusted to optimize energy consumption. The above is an illustrative example. The control strategy includes temperature control, fuel control, pressure control, flow rate control, external temperature control, and external humidity control of the heating furnace. Specifically, temperature control includes setting the target temperature value, the furnace inlet set temperature, and the outlet temperature.

[0081] S3. Construct a heterogeneous network containing operating condition data and preliminary control strategies, and generate candidate control strategies by walking through the heterogeneous network based on a bidirectional random walk algorithm.

[0082] In a further embodiment, a heterogeneous network containing operating condition data and a preliminary control strategy is constructed, and a bidirectional random walk algorithm is used to walk through the heterogeneous network to generate candidate control strategies, including:

[0083] S31. Construct a heterogeneous network using current operating condition data, preliminary control strategy, historical operating condition data, and historical energy consumption data as nodes. The heterogeneous network is composed of nodes and edges of different types. S32. Define the number of walk steps and the number of walks, and perform a bidirectional random walk starting from a current operating condition data node and a preliminary control strategy node. The bidirectional random walk is a random walk algorithm performed in a graph or network structure, starting simultaneously from two different initial nodes. This algorithm can increase the comprehensiveness of the exploration and improve the efficiency of finding target nodes or paths. S33. Obtain the walk paths generated by the bidirectional random walk and record the sequence of nodes visited during the walk, especially the preliminary control strategy node. Based on the frequency and stability of the walk paths, generate candidate control strategies. Multiple candidate control strategies can be selected, or only one can be chosen, for subsequent optimization and selection processes.

[0084] The node access frequency analysis includes:

[0085] Analyze each traversal path, counting the number of times each node, especially the node associated with the initial control strategy, is visited. Identify nodes with a visit frequency exceeding a certain threshold; these nodes may represent effective control strategies. Analyze the stability of the traversal paths, i.e., the consistency and repeatability observed across multiple traversals. Select traversal paths that are both frequent and stable; these paths may lead to effective control strategies.

[0086] In a further embodiment, constructing a heterogeneous network using current operating condition data, preliminary control strategy, historical operating condition data, and historical energy consumption data as nodes includes:

[0087] Obtain the node types in the heterogeneous network, including current operating condition data nodes, preliminary control strategy nodes, historical operating condition data nodes, and historical energy consumption data nodes; S312, based on the weighted similarity and correlation between nodes, determine the weights of the edges between current operating condition data nodes and historical operating condition data nodes, between preliminary control strategy nodes and historical energy consumption data nodes, and between historical operating condition data nodes and historical energy consumption data nodes.

[0088] For example, the edge weights between the current operating condition data node and the historical operating condition data node are based on the weighted similarity calculated in S2; the edge weights between the preliminary control strategy node and the historical energy consumption data node, as well as the edge weights between the historical operating condition data node and the historical energy consumption data node, are determined by similarity measurement, i.e., cosine similarity, or correlation measurement, i.e., Pearson correlation coefficient. At the same time, the data of the nodes are processed by standardization and normalization to eliminate the difference in magnitude between different feature values.

[0089] In a further embodiment, the number of walk steps and the number of walks are defined. A bidirectional random walk starting from a current operating condition data node and a preliminary control strategy node includes:

[0090] Determine the number of walk steps, i.e., the number of steps to move from one node to another during the walk, and the number of walks, i.e., the number of times the walk process is repeated. Select a current working condition data node and a preliminary control strategy node as the starting point for the walk. S322. Determine the direction of the walk, the probability transfer mechanism, and the self-starting coefficient to achieve bidirectional random walk.

[0091] Walk direction: Determines whether the walk is limited to nodes of the same type, or allows cross-type nodes, such as from a condition data node to a control strategy node. Probability transfer mechanism: Defines the probability of a walker moving from one node to another, based on edge weights. Self-starting coefficient: Sets a coefficient to adjust the initial distribution of the walk, allowing the walker to initially favor certain nodes or edges.

[0092] For example, the number of steps and the number of walks are 100. Node A, the current condition data node, and node B, the initial control strategy node, are selected as the starting points for the walk. A bidirectional walk is performed, starting simultaneously from nodes A and B. The transition probability is determined based on the edge weights. Higher weights result in a greater probability of transitioning to that node. An auto-start coefficient is set to increase the comprehensiveness of the exploration. The number of times each node is visited is counted. Frequently visited nodes are identified, and nodes with a visit frequency exceeding a certain threshold are determined. The stability of the walk paths is analyzed. Walk paths that are both frequent and stable are selected, as these paths may lead to effective control strategies. Candidate control strategies are generated based on these frequent and stable walk paths.

[0093] S4. Based on candidate control strategies, optimize and determine the final control strategy corresponding to each operating condition. This involves optimizing the candidate control strategies using optimization algorithms, such as particle swarm optimization or firefly optimization.

[0094] In a further embodiment, the final control strategy for each operating condition is determined based on candidate control strategies and chaotic optimization techniques, including:

[0095] S41. Determine the optimization objective, such as minimizing energy consumption, improving heating efficiency, or reducing heating time, and construct an evaluation function based on the objective. Assign a fitness value to each individual based on the evaluation function results. S42. Optimize the initial population using a chaotic optimization algorithm to ensure population diversity, improve the algorithm's global search capability, and evaluate each individual based on the evaluation function, representing the performance of a candidate control strategy. S43. Update the individual's position based on the fitness differences and attractiveness among individuals. In optimization problems, the quality of each solution is measured by its fitness or objective function value. Quality difference refers to the difference in fitness between different solutions. Solutions with higher fitness are considered closer to the optimal solution of the optimization problem. Attractiveness describes the influence of solutions with higher fitness on other solutions, especially those with lower fitness. Attractiveness can encourage the search process to move towards regions containing high-quality solutions. S44. Construct search parameters that vary with the number of iterations, such as the learning rate or step size, to balance the ability to explore globally and exploit locally during the search process. S45. After each iteration, sort the individuals based on their fitness, eliminate poorer solutions, and generate new solutions. S46. After multiple iterations, select the individual with the highest fitness as the final control strategy for each working condition.

[0096] The formula for the evaluation function is as follows:

[0097] ;

[0098] In the formula, E en This indicates the actual energy consumption of the heating furnace; the lower the value, the less energy is consumed. E eff It represents the reciprocal of the actual efficiency value of the heating furnace, or a quantity that is inversely proportional to the efficiency, because the higher the efficiency, the smaller its reciprocal should be; E t This indicates the actual heating time of the furnace; the lower the value, the shorter the heating time. w 1, w 2, w 3 represents the weighting coefficients for the actual energy consumption value, the reciprocal of the actual efficiency value, and the actual heating time, respectively. All weighting coefficients must add up to one, and their selection depends on the relative importance of each performance indicator. For example, if energy consumption is the most important consideration, then... w 1 will be given the highest weight.

[0099] For example, candidate control strategy one has a fuel flow rate of 48 kg / h and a target temperature of 700°C. Candidate control strategy two has a fuel flow rate of 49 kg / h and a target temperature of 705°C. Candidate control strategy three has a fuel flow rate of 47 kg / h and a target temperature of 695°C. After multiple rounds of iteration and evaluation, the final control strategy with the highest fitness was obtained: a fuel flow rate of 48.5 kg / h and a target temperature of 702°C.

[0100] In a further embodiment, optimizing the initial population using a chaotic optimization algorithm includes:

[0101] S421. Set the control parameters for the piecewise linear chaotic mapping algorithm, such as the mapping segmentation points and the parameter values ​​controlling chaos. Use the piecewise linear chaotic mapping algorithm to optimize the initial population and generate chaotic sequences. These sequences will be used to adjust the positions of individuals in the initial population to increase population diversity and avoid the limitations of the initial solution. Chaotic optimization algorithms are methods that utilize the randomness and unpredictability of chaos theory to improve optimization algorithms. Chaotic optimization algorithms can increase population diversity, help the algorithm escape local optima, and improve global search capabilities. Piecewise linear chaotic mapping generates chaotic sequences in a piecewise linear manner. In the chaotic mapping algorithm, the control parameters determine the characteristics of chaotic behavior, such as the mapping segmentation points and the parameter values ​​controlling chaos. These parameters have a significant impact on the distribution and characteristics of the chaotic sequences. S422. Combine the chaotic sequences with individuals in the initial population, and update the positions of individuals through chaotic mapping, thereby increasing population diversity. A series of values ​​generated by the chaotic mapping algorithm, these values ​​are random and unpredictable. In optimization algorithms, chaotic sequences are used to perturb the initial population, increasing its diversity. By introducing chaos, patterns or clusters that may exist in the initial population are broken, thereby promoting population diversity.

[0102] like Figure 2 As shown, according to another embodiment of the present invention, a heating furnace control optimization system is also provided, including a data acquisition module 1, a preliminary screening module 2, a global selection module 3, and an optimization module 4; wherein the data acquisition module 1, the preliminary screening module 2, the global selection module 3, and the optimization module 4 are connected in sequence.

[0103] Data acquisition module 1 is used to acquire the operating condition data of the heating furnace, and to preprocess and label the operating condition data to obtain the energy consumption data corresponding to each operating condition. The operating condition data includes the operating data of the heating furnace and the environmental data.

[0104] The preliminary screening module 2 is used to calculate the weighted similarity measure under different operating conditions based on operating condition data and energy consumption data, and to generate preliminary control strategies corresponding to different operating conditions based on the weighted similarity measure.

[0105] The global selection module 3 is used to construct a heterogeneous network using current operating condition data, preliminary control strategy, historical operating condition data, and historical energy consumption data as nodes; it defines the number of walk steps and the number of walks, and performs a bidirectional random walk starting from a certain current operating condition data node and a preliminary control strategy node; it obtains the walk path generated by the bidirectional random walk, records the sequence of nodes visited during the walk, and generates candidate control strategies based on the frequency and stability of the walk path.

[0106] Optimization module 4 is used to optimize and determine the final control strategy for each operating condition based on candidate control strategies.

[0107] In summary, the heating furnace control optimization method and system provided by this invention identifies similar operating conditions by calculating a weighted similarity metric between different operating conditions, providing a basis for generating preliminary control strategies. Generating preliminary control strategies based on similar operating conditions improves the rationality of the initial strategy and reduces blind spots. By constructing a heterogeneous network, the complex relationship between operating condition data and control strategies is explored, providing more information for optimizing the control strategy. Based on a bidirectional random walk algorithm, more global candidate control strategies are generated, improving the comprehensiveness of strategy selection. This invention not only considers current operating condition data and preliminary control strategies but also introduces historical operating condition data and historical energy consumption data, constructing a heterogeneous network containing multiple node types. This more comprehensively captures the complex relationships between different operating conditions, control strategies, and energy consumption data, helping to discover potential correlations and patterns, and improving the rationality and effectiveness of the control strategy. The bidirectional random walk algorithm can explore multiple times and in multiple directions within the heterogeneous network, generating diverse candidate control strategies and avoiding local optima. Combining multi-dimensional evaluation criteria such as similarity metrics, path frequency, and stability generates more reliable and comprehensive candidate control strategies. This invention uses an optimization algorithm to evaluate the generated candidate control strategies, calculates their scores under different evaluation indicators, and selects the control strategy with the highest fitness value as the final control strategy based on the optimization results. This ensures that the selected control strategy takes into account energy saving, combustion efficiency, and heating time, thus guaranteeing the economic efficiency of the final control strategy for each operating condition.

[0108] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing the control of a heating furnace, characterized in that, include: S1. Obtain the operating data of the heating furnace, and preprocess and label the operating data to obtain the energy consumption data corresponding to each operating condition; S2. Based on operating condition data and energy consumption data, calculate the weighted similarity metric under different operating conditions, and generate the preliminary control strategy corresponding to different operating conditions based on the weighted similarity metric. Specifically, it includes: S21. Extract operational data features and environmental data features from the operating condition data, and calculate the statistics for each operational data feature and environmental data feature; S22. Extract energy consumption features from the energy consumption data, and calculate the statistics for the energy consumption data; S23. Integrate the operating condition features and corresponding energy consumption features into overall features, calculate the similarity between different overall features, form a comprehensive similarity measure, assign weights to the operating condition features and energy consumption features, and calculate a weighted similarity measure; S24. Based on the weighted similarity measure, historical operating condition data, and historical energy consumption data, generate a preliminary control strategy corresponding to each operating condition; S23 includes: standardizing the operational data features, environmental data features, and energy consumption features using calculated statistics; combining the operational data features and environmental data features into operating condition features, and integrating the operating condition features and corresponding energy consumption features into overall features, and calculating the similarity between different overall features using a similarity measurement algorithm; assigning weights to the operating condition features and energy consumption features based on statistics, and calculating weighted similarity measurements under different operating conditions; wherein, the overall features include the furnace's temperature, pressure, flow rate, external temperature, humidity, fuel consumption, and electrical energy consumption; when assigning weights, calculating the mean and standard deviation of each feature in the overall features, and dividing the standard deviation of the feature by the mean of the feature to obtain the coefficient of variation of each feature; assigning weights to each feature based on the coefficient of variation; S24 includes: selecting the historical operating condition data most similar to the current operating condition based on a weighted similarity metric; extracting the control strategy corresponding to the most similar historical operating condition data as the preliminary control strategy corresponding to the current operating condition; S3. Construct a heterogeneous network using current operating condition data, preliminary control strategy, historical operating condition data, and historical energy consumption data as nodes; define the number of walk steps and the number of walks, and perform a bidirectional random walk starting from a current operating condition data node and a preliminary control strategy node; obtain the walk path generated by the bidirectional random walk, and record the sequence of nodes visited during the walk; generate candidate control strategies based on the frequency and stability of the walk path. S4. Based on candidate control strategies and chaotic optimization techniques, determine the final control strategy corresponding to each working condition; The construction of a heterogeneous network using current operating data, preliminary control strategies, historical operating data, and historical energy consumption data as nodes includes: Obtain the node types in the heterogeneous network, including current operating condition data nodes, preliminary control strategy nodes, historical operating condition data nodes, and historical energy consumption data nodes; based on the weighted similarity and correlation between nodes, determine the weights of the edges between current operating condition data nodes and historical operating condition data nodes, between preliminary control strategy nodes and historical energy consumption data nodes, and between historical operating condition data nodes and historical energy consumption data nodes.

2. The furnace control optimization method according to claim 1, characterized in that, The process of acquiring the operating condition data of the heating furnace, and preprocessing and labeling the operating condition data to obtain the energy consumption data corresponding to each operating condition includes: S11. Collect the operating data of the heating furnace, including the furnace's operating data and environmental data; S12. Use data preprocessing techniques to clean and calibrate operational and environmental data. S13. Label the collected operating data and record the corresponding energy consumption data.

3. The method for optimizing the control of a heating furnace according to claim 1, characterized in that, The definition of the number of walk steps and the number of walks, and the bidirectional random walk starting from a certain current working condition data node and the preliminary control strategy node, includes: Determine the number of steps and the number of walks, and select a current working condition data node and a preliminary control strategy node as the starting point for the walk; Define the direction of the walk, the probability transfer mechanism, and the self-starting coefficient to achieve bidirectional random walk; During the bidirectional random walk, the number of times each node is visited is recorded, nodes with a visit frequency higher than the threshold are identified, and the frequency and consistency of nodes are analyzed during multiple walks. Based on frequent and stable walk paths, candidate control strategies are generated.

4. The furnace control optimization method according to claim 3, characterized in that, The determination of the final control strategy for each operating condition based on candidate control strategies and chaotic optimization techniques includes: S41. Determine the optimization objective and construct an evaluation function based on the optimization objective; S42. Optimize the initial population using a chaotic optimization algorithm, and evaluate the performance of each individual based on the evaluation function; S43. Update the position of individuals based on differences in fitness and attractiveness among them; S44. Construct search parameters that vary with the number of iterations; S45. After each iteration, sort the individuals according to their fitness. S46. After multiple iterations, select the individual with the highest fitness as the final control strategy for each working condition. The formula for the evaluation function is as follows: ； In the formula, E en This indicates the actual energy consumption value of the heating furnace; E eff This represents the reciprocal of the actual efficiency value of the heating furnace; E t This indicates the actual heating time of the heating furnace; w 1, w 2, w 3 represents the weighting coefficients corresponding to the actual energy consumption value of the heating furnace, the reciprocal of the actual efficiency value of the heating furnace, and the actual heating time of the heating furnace, respectively.

5. The furnace control optimization method according to claim 4, characterized in that, The optimization of the initial population using the chaotic optimization algorithm includes: S421. Set the control parameters of the piecewise linear chaotic mapping algorithm, use the piecewise linear chaotic mapping algorithm to optimize the initial population, and generate a chaotic sequence; S422. Combine the chaotic sequence with the individuals in the initial population, and update the position of the individuals through chaotic mapping.

6. A heating furnace control optimization system, used to implement the heating furnace control optimization method according to any one of claims 1-5, characterized in that, include: Data acquisition module, preliminary screening module, global selection module, and optimization module; The data acquisition module, the preliminary screening module, the global selection module, and the optimization module are sequentially connected. The data acquisition module is used to acquire the operating condition data of the heating furnace, and to preprocess and label the operating condition data to obtain the energy consumption data corresponding to each operating condition. The preliminary screening module is used to calculate the weighted similarity metric under different operating conditions based on operating condition data and energy consumption data, and to generate preliminary control strategies corresponding to different operating conditions based on the weighted similarity metric. The global selection module is used to construct a heterogeneous network using current operating condition data, preliminary control strategy, historical operating condition data, and historical energy consumption data as nodes; define the number of walk steps and the number of walks; perform a bidirectional random walk starting from a current operating condition data node and a preliminary control strategy node; obtain the walk path generated by the bidirectional random walk; record the sequence of nodes visited during the walk; and generate candidate control strategies based on the frequency and stability of the walk path. The optimization module is used to optimize and determine the final control strategy corresponding to each operating condition based on the candidate control strategies.

Images