Sintering waste heat power generation intelligent control method based on data driving

Abstract

The invention discloses a sintering waste heat power generation intelligent control method based on data driving. The method comprises the steps that data per second are collected from a controller; preprocessing the data, dividing the processed data into a training set and a verification set, and taking the sample data which does not finish speed regulation valve opening control as a prediction set; sorting the sample data of the training set from large to small according to the steam consumption rate, and taking the first 70% of the sample data for modeling; a multi-modal deep learning model algorithm is optimized based on an improved eagle optimization algorithm to perform model creation, training and optimization on the data; performing prediction according to the model to obtain a newest speed regulation valve opening degree appropriate value; and comparing the newest appropriate value of the opening of the speed regulation valve with the appropriate value of the last statistical period, if the absolute value of the variable quantity exceeds a threshold value 0.1 and lasts for 5 seconds, adjusting the opening of the speed regulation valve, and otherwise, not adjusting the opening of the speed regulation valve. The operation efficiency, stability and safety of the power generation system are improved, and a large amount of labor cost is saved.

Description

Technical Field

[0001] The present invention relates to the technical field of sintering waste heat power generation, and in particular to an intelligent control method for sintering waste heat power generation based on data driving. Background Art

[0002] Energy conservation and emission reduction in the iron and steel industry are crucial. A large amount of waste heat is released during the production process of sintering machines, and this waste heat can be used for power generation. The sintering waste heat power generation technology introduces high-temperature waste gas in the sintering production process into a waste heat boiler, generates superheated steam by heating feed water, and then drives a steam turbine or a generator to generate electricity. The waste heat power generation technology can effectively reduce the energy consumption of iron and steel plants, improve the energy utilization efficiency, and reduce environmental pollution.

[0003] The waste heat power generation system has strict requirements for the quality of main steam. However, the current sintering system's thermal system is very unstable, and the waste heat heat source has the characteristics of low overall quality, large fluctuations in waste gas temperature, flow rate, and pressure. The long-term and frequent fluctuations of steam parameters seriously threaten the safety, stability, and operating efficiency of the steam turbine. At present, most iron and steel plants adjust the opening degree of the speed control steam valve according to the experience of on-site operators. This method has uncertainty and non-quantifiability, cannot perform real-time and accurate calculations with the change of the working parameters of the waste heat recovery system, and has a long adjustment period, serious hysteresis, and high labor costs. In addition, due to different operating habits, control standards, and experiences of operators, the operating efficiency of the power generation systems controlled by each shift is different, resulting in poor system operation stability, low power generation efficiency, and high failure rates. During the actual production process, process parameters such as main steam pressure, main steam flow rate, and main steam temperature often change. The complexity and variability of working conditions cause operators to be unable to respond in real time and accurately, resulting in unstable system operation and affecting power generation efficiency. Summary of the Invention

[0004] In order to overcome the above problems existing in the prior art, the present invention proposes an intelligent control method for sintering waste heat power generation based on data driving.

[0005] The technical solution adopted by the present invention to solve its technical problems is: an intelligent control method for sintering waste heat power generation based on data driving, including the following steps: Step 1, historical data collection: Collect data per second from the controller and store it in the database; Step 2, data processing: Preprocess the data collected in Step 1, divide the processed data into a training set and a validation set, and use the sample data that has not completed the speed control steam valve opening control as the prediction set; Step 3, modeling sample selection: Sort the sample data in the training set from small to large according to the steam consumption rate, and take the first 70% of the sample data for modeling; Step 4, algorithm modeling: Based on the improved Tianying optimization algorithm to optimize the multi-modal deep learning model algorithm, create, train, and optimize the model for the sample data in Step 3; Step 5, predict the prediction set according to the model obtained in Step 4 to obtain the latest appropriate value of the governor valve opening, and store it in the database; Step 6, governor valve opening control: Compare the latest appropriate value of the governor valve opening obtained in Step 5 with the appropriate value in the previous statistical cycle. If the absolute value of the change amount exceeds the threshold of 0.1 and lasts for 5 seconds, then transmit the appropriate value of the governor valve opening to the controller for governor valve opening adjustment, otherwise no adjustment is made; In the above Step 6, set the upper and lower limit thresholds of the main steam pressure. If it is lower than the lower limit value, quickly reduce the governor valve opening. If the pressure cannot be restored to the normal range during the regulation process, maintain the turbine at 3000 rpm speed trip state; if the main steam pressure is too low, immediately automatically close the main steam valve for shutdown; if it is higher than the upper limit value, automatically adjust the governor valve opening to 100%.

[0006] For the above data-driven intelligent control method for sintering waste heat power generation, the specific data preprocessing in Step 2 includes: performing mean filtering on the collected historical data with a time period of 20 seconds, performing normalization on the filtered data, delaying the time length of each index variable to the turbine, and aligning all index values in time.

[0007] For the above data-driven intelligent control method for sintering waste heat power generation, the specific calculation formula for the steam consumption rate in Step 3 is: steam consumption rate = total steam consumption / power generation.

[0008] For the above data-driven intelligent control method for sintering waste heat power generation, the specific improvement of the Tianying optimization algorithm to optimize the multi-modal deep learning model algorithm in Step 4 is: taking the main steam temperature, main steam flow, and main steam pressure as input variables and the governor valve opening as the output variable, inputting into the CNN layer for local feature extraction, and its output enters the BiGRU-Transformer layer, where the Transformer layer adopts the multi-head self-attention mechanism, and the model automatically focuses on the main influencing variables, and finally maps the fully connected layer to the prediction target.

[0009] For the above data-driven intelligent control method for sintering waste heat power generation, Step 4 specifically includes: Step 4.1, initialize and generate the initial solution: Use the Tent chaotic mapping to initialize the parameters of the Tianying optimization algorithm and randomly generate the initial parameters of the CNN-BiGRU-Transformer model; Step 4.2, Fitness calculation: Use the parameters of each individual to train the CNN-BiGRU-Transformer model, and evaluate the prediction performance of the model using the validation set. Step 4.3, Iterative optimization: Utilize the iterative strategy of the Tianying optimization algorithm, adjust the global search using adaptive weights, and update the parameters of each individual in the population until the termination condition is met. Step 4.4, Optimal parameter selection: Select the parameters corresponding to the individual with the minimum fitness value as the final model parameters to obtain the prediction model.

[0010] In the above intelligent control method for sintering waste heat power generation based on data-driven, the Tent chaos mapping in Step 4.1 is specifically as follows: Generate N chaotic sequences using the Tent chaos mapping, where N is the population size, map the chaotic sequences to the search space to generate the initial population, and the mathematical model of the Tent chaos mapping is: where, is , is the i-th data.

[0011] In the above intelligent control method for sintering waste heat power generation based on data-driven, the specific formula for the adaptive weight in Step 4.3 is: where t is the current iteration number, is the maximum iteration number; Adjust the global search using the adaptive weight, and the formula is: where, , is the current position, , is the adaptive position.

[0012] The beneficial effects of the present invention are as follows: The present invention provides an intelligent control method for sintering waste heat power generation. The server collects historical production big data, deeply mines the internal relationship between the throttle valve opening and its influencing factors through the improved IAO-CNN-BiGRU-Transformer algorithm, and summarizes and forms knowledge, so as to be able to intelligently process and solve the system fluctuation problems caused by the change of steam parameters under various working conditions. On the one hand, a large amount of labor costs are saved. On the other hand, the problems of lag in manual control, inconsistent manual operation standards, and the inability of manual to respond accurately and in real time to complex and changeable working conditions are solved, which not only saves a large amount of labor but also improves the long-term stability and power generation efficiency of the power generation system. Brief Description of the Drawings

[0013] Figure 1 It is a schematic flow diagram of the present invention. Detailed Embodiment

[0014] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be described in detail below in conjunction with the drawings and specific embodiments.

[0015] As Figure 1 shown, the present invention provides an intelligent control scheme for sintering waste heat power generation based on data-driven. This scheme collects historical production big data, and uses the CNN-BiGRU-Transformer intelligent algorithm to autonomously learn the inherent natural laws between the main steam temperature, main steam flow, main steam pressure, etc. and the governor valve opening under various working conditions, and uses the improved Tianying optimization algorithm (IAO) to optimize the parameters of the CNN-BiGRU-Transformer model, so that the created model can adapt to various working condition changes, improve the control accuracy and control response speed of the model for actual production, and solve the problem of system instability caused by complex and changeable working conditions. Finally, using the knowledge formed by the optimal model, the appropriate value of the governor valve opening is predicted accurately in real time, and feedforward control is realized, eliminating the lag effect.

[0016] The control process includes historical data collection, data processing, intelligent algorithm modeling, calculation of the appropriate value of the governor valve opening, and system control. The specific process steps are as follows: The First Step: Data Collection Use the KepServer configuration software to collect data from the PLC (programmable logic controller) every second and store it in the database. The collected data mainly includes main steam temperature, main steam flow, main steam pressure, governor valve opening, power generation, etc., providing data support for subsequent algorithm modeling and optimization.

[0017] The Second Step: Data Processing Perform mean filtering on the collected historical data with a time period of 20 seconds, and perform normalization on the filtered data; delay the time length of each index variable to the steam turbine to align all index values in time. Divide the processed data into data sets. The sample data that has completed the governor valve opening control is divided into a training set and a validation set according to a ratio of 7:3, and the sample data that has not completed the governor valve opening control is used as the prediction set.

[0018] The purpose of this step is to provide a high-quality input data set for the construction of the subsequent prediction model through data preprocessing and structured division, playing a key role in improving data quality and enhancing the generalization ability of the model in the overall scheme.

[0019] The mean filtering process and normalization process are performed on the data, effectively enhancing the effectiveness of the process data participating in control and reducing the negative impact of abnormal data fluctuations on the system control accuracy; by aligning the sample data in terms of time, the change relationships between process parameters can be accurately mined, further improving the quality of the sample data.

[0020] Through the division of sample data, the sample data that has completed the control of the governor valve opening is divided into a training set and a validation set according to a ratio of 7:3. This division ratio is designed based on the time series correlation characteristics of industrial process data, which not only retains sufficient historical patterns for model training and learning but also reserves enough data to verify the prediction accuracy of the model and its adaptability to working conditions, helping to optimize model parameters during subsequent modeling, ensuring model stability, improving the efficiency of model training, and ensuring the quality of model training. Thus, when predicting the appropriate value of the governor valve opening through the model subsequently, the accuracy is higher, ensuring the stability of the entire system. The training set is used to learn data features, and the validation set is used to adjust hyperparameters to avoid performance degradation caused by over-optimizing training data. At the same time, the sample data that has not completed the control of the governor valve opening is used as the prediction set, constructing a historical-real-time dual-channel data processing mechanism, enabling the model to not only learn historical operation rules but also adapt to new working conditions, improving the adaptability of the model.

[0021] Step 3: Selecting Optimal Modeling Samples The selection of high-quality cases is determined by the power generation efficiency. The specific formula for power generation efficiency needs to be determined according to the specific power generation system and waste heat recovery technology. For example, the steam consumption rate, which is a negative indicator. After sorting the training set data in ascending order of the steam consumption rate, the first 70% are taken as high-quality samples for modeling.

[0022] The steam consumption rate refers to the ratio of the cumulative value of the main steam flow during the statistical period of the steam turbine unit to the power generation of the unit. The calculation formula is: Steam consumption rate = Total steam consumption (kg) / Power generation (kW·h).

[0023] By selecting the first 70% of the samples with a small steam consumption rate, a high-quality input data set is provided for subsequent algorithm modeling, training, and optimization, enabling the model to focus on learning and capturing the internal laws and key features of high-quality operation modes; eliminating the latter 30% of the sample data is equivalent to performing a strong data cleaning based on the target variable, significantly reducing the risk of the model learning incorrect or irrelevant modes, and effectively improving the generalization ability of the model. The model established based on high-quality samples naturally biases its predictions and suggestions towards the high-efficiency operation area, and can more directly and effectively provide decision support for operation optimization, parameter setting, and operation guidance, guiding the model to learn the non-linear relationship between the efficiency of the power generation system and operation parameters in the high-efficiency area.

[0024] In summary, this step plays a key role in improving data quality and enhancing the model generalization ability in the overall solution, making the accuracy of the predicted appropriate value of the governor valve opening obtained by training higher, improving the control accuracy, further increasing the power generation efficiency, and ensuring the stability of the entire system.

[0025] Step 4: Intelligent algorithm modeling Based on the improved Aquila Optimizer (IAO), the multi-modal deep learning model CNN-BiGRU-Transformer algorithm is used to create, train, and optimize the model for the training set formed in the third step. By integrating the advantages of the convolutional neural network (CNN), bidirectional gated recurrent unit (BiGRU), and Transformer module, this hybrid model realizes the efficient extraction of data spatial features, the efficient modeling of bidirectional temporal features, and the long-range global dependencies. Taking the main steam temperature, main steam flow rate, and main steam pressure as input variables and the governor valve opening as the output variable, they are input into the CNN layer for local feature extraction. The output then enters the BiGRU-Transformer layer, where the Transformer layer adopts the multi-head self-attention mechanism. The model automatically focuses on the main influencing variables, and finally maps the fully connected layer to the prediction target. When training the model, the hyperparameters to be optimized mainly include the learning rate, the number of CNN convolutional kernels, the number of BiGRU neurons, the number of attention heads, etc.

[0026] The Aquila Optimizer (AO) is a new population-based metaheuristic algorithm that explores the optimal solution by mimicking the global search and local search behaviors of eagles when hunting for prey. The improved Aquila Optimizer (IAO) optimizes the algorithm parameters and search strategies on this basis. On the one hand, it reduces the computational complexity and the requirements for hardware configuration; on the other hand, it improves the global search ability and convergence speed, and the optimized parameters are more accurate and timely, enabling the model to adapt to different working conditions and changes, and improving the control accuracy.

[0027] The optimization process of the traditional AO algorithm is divided into 4 parts: 1) Extended search stage: Identify the prey area through high flight and vertical bending, and select the best hunting area; 2) Shrinking exploration stage: Search within the divergent search space in the way of short gliding, and prepare for landing attack; 3) Expanding exploitation stage: When the prey area is accurately specified, conduct a low-altitude and slow-descent attack; 4) Shrinking exploitation stage: Raid in the way of walking and predation.

[0028] Although the traditional AO algorithm balances exploration and exploitation capabilities and has excellent search capabilities, there is still room for improvement in the extended exploration and reduced exploration phases: Initialize the population using the Tent chaotic map The traditional AO algorithm uses a method of randomly initializing the population, which easily leads to uneven population distribution and affects the algorithm performance. This solution uses the Tent chaotic map to initialize the population, which can effectively increase the diversity of the population, thereby improving the global search ability of the algorithm, making the searched parameters more generalizable and adaptable to various working conditions. Specifically, N chaotic sequences are generated using the Tent chaotic map, where N is the population size, and the chaotic sequences are mapped to the search space to generate the initial population.

[0029] The mathematical model of the Tent chaotic map is: Adaptive weight To accelerate the algorithm search rate, a larger weight is required in the initial stage of iteration to strengthen the global search ability, and a smaller weight is required in the later stage of iteration to strengthen the local exploration ability. It can perform fine search near the optimal solution, improving the convergence speed while enhancing the search accuracy, making the searched parameters more accurate, reducing the computational amount and improving the control accuracy.

[0030] The adaptive weight formula is: where t is the current iteration number, is the maximum number of iterations.

[0031] The global search is adjusted using the adaptive weight, and the formula is: where, 、 is the current position, 、 is the adaptive position.

[0032] The specific steps of the improved IAO-CNN-BiGRU-Transformer modeling are as follows: 1) Initialization to generate the initial solution: The parameters of the Eagle optimization algorithm, such as population size, number of iterations, etc., are initialized using Tent chaotic mapping to randomly generate the initial parameters of the CNN-BiGRU-Transformer model. The parameters of the Eagle optimization algorithm are initialized by Tent chaotic mapping, which has ergodic uniformity, initial value sensitivity, and non-periodicity, and can efficiently cover the entire solution space. Compared with ordinary random initialization, it significantly avoids population aggregation and falling into local optimality. The high diversity of the initial population forces the IAO algorithm to explore a wider parameter combination space in subsequent iterations. This makes the CNN-BiGRU-Transformer parameters finally optimized not only fit the current training data, but also adapt to unseen and complex working condition fluctuations, which directly responds to the problem that "it is difficult for humans to accurately respond to complex and changing working conditions in real time", and lays a strong foundation for the robustness of the model.

[0033] 2) Fitness calculation: The parameters of each individual are used to train the CNN-BiGRU-Transformer model, and the prediction performance of the model is evaluated using the validation set. For example, the root mean square error RMSE is used as the fitness function, and its calculation formula is: in, represents the predicted value, Indicates the actual value, and n indicates the number of samples. In this step, the above-mentioned adaptive weight algorithm is used. In the early stage (exploration): Increase the global search weight to encourage the algorithm to explore the parameter space extensively and avoid premature convergence. In the later stage (development): Increase the local search weight, focus on potential areas for detailed search, and quickly approach the optimal solution. Dynamic balance: Automatically adjust the weight according to the iteration progress or population dispersion to achieve an intelligent balance between exploration and development.

[0034] The advantages of RMSE as fitness are: It directly quantifies the deviation between the predicted value and the true value, and its goal is to minimize the prediction error, which is highly consistent with the core goal of "improving prediction accuracy".

[0035] Through the above steps, both accuracy and efficiency can be improved. (a) Accuracy improvement: More accurate parameter search directly improves the model's ability to predict the key control variable, the speed regulating valve opening. Accurate prediction is the prerequisite for optimizing control, thereby ensuring power generation efficiency.

[0036] (b) Improved efficiency: Adaptive weights reduce invalid searches and significantly shorten the number of iterations or computing time required to find high-quality parameters.

[0037] Key to industrial responsiveness: Improved computing efficiency enables the model to complete training / tuning faster when facing rapid fluctuations in core process parameters such as main steam pressure, flow, and temperature, and quickly generate a new optimal model after adapting to changes in operating conditions. Support online / near-line applications: Lay the computing foundation for possible real-time optimization control in the future. Ensure system stability: Fast response means that the control system can be adjusted more promptly to avoid control imbalance and system oscillation caused by model lag.

[0038] 3) Iterative optimization: Using the iterative strategy of the IAO algorithm, update the parameters of each individual in the population until the termination condition is met, such as the maximum number of iterations or accuracy requirements.

[0039] The four intelligent hunting behaviors of the Sky Eagle (high-altitude dive, low-altitude gliding, round-up, and raid) are used to update parameters, simulating efficient global exploration and local development capabilities. The improvements of IAO make it converge faster when dealing with high-dimensional and complex CNN-BiGRU-Transformer parameter optimization problems; the probability of finding the global optimal or approximate global optimal solution is higher; and it is relatively more robust to initialization and parameter settings.

[0040] Optimal parameter selection: Select the parameters corresponding to the individual with the smallest fitness value as the final model parameters.

[0041] The CNN-BiGRU-Transformer model is trained with the optimal parameters. The model at this time is the optimal model, which is saved for the next prediction.

[0042] Step 5: Prediction of the appropriate value of the speed regulating valve opening The prediction set formed in the second step is predicted according to the optimal model generated in the fourth step to obtain the latest appropriate value of the speed regulating valve opening and store it in the database.

[0043] Step 6: Speed regulating valve opening control Compare the latest appropriate value of the governing valve opening with the appropriate value in the previous statistical period. If the absolute value of the change amount exceeds the threshold of 0.1 and lasts for 5 seconds, transmit the appropriate value of the governing valve opening to the PLC for adjusting the governing valve opening; otherwise, no adjustment is made. To ensure the stable and safe operation of the waste heat power generation system, it is necessary to set the upper and lower limits of the main steam pressure, such as the upper limit of 1.65 MPa and the lower limit of 1.25 MPa. If it is lower than the lower limit value, quickly reduce the governing valve opening to reduce the steam turbine load. If the pressure cannot be restored to the normal range during the regulation process, maintain the steam turbine at a speed of 3000 rpm in the tripped state to avoid water hammer accidents on the steam turbine rotor. If the main steam pressure is too low, such as lower than 1.0 MPa, immediately automatically close the main steam valve for shutdown. If it is higher than the upper limit value, automatically adjust the governing valve opening to 100% to prevent the steam specific volume from decreasing and the main steam flow from increasing due to excessive pressure for a long time, which may cause overloading of the last-stage blades, accelerate the bearing vibration of the steam turbine unit, and lead to serious production safety accidents such as fracture of the large shaft of the unit.

[0044] In this step, by setting an opening threshold of 0.1, it is possible to avoid frequent adjustments caused by small fluctuations in the model or measurement noise, reducing mechanical wear; the 5-second duration is conducive to confirming the effectiveness of the trend and preventing false triggering by instantaneous interference, achieving a balance between the response speed and equipment safety.

[0045] Through the above process, the AI technology is used to replace manual operation to achieve the feedforward regulation of the governing valve, solving the problem of manual regulation lag and saving labor costs; the AI intelligent algorithm is used to deeply explore the internal relationship between the main steam temperature, main steam flow, main steam pressure and the governing valve opening under various working conditions, realizing the timely and accurate control of the power generation system, improving the long-term stability and power generation efficiency of the power generation system, and effectively reducing the risks of fluctuations and abnormal working conditions.

[0046] The above embodiments are only exemplary embodiments of the present invention and are not used to limit the present invention. Those skilled in the art can make various modifications or equivalent replacements to the present invention within the essence and protection scope of the present invention, and such modifications or equivalent replacements should also be regarded as falling within the protection scope of the present invention.

Claims

1. An intelligent control method for sintering waste heat power generation based on data driving, characterized in that It includes the following steps: Step 1, historical data collection: Collect data per second from the controller and store it in the database; Step 2, data processing: Preprocess the data collected in Step 1, divide the processed data into a training set and a validation set, and use the sample data that has not completed the governing valve opening control as the prediction set; Step 3, selection of modeling samples: Sort the sample data in the training set in ascending order of steam consumption rate, and take the first 70% of the sample data for modeling; Step 4, algorithm modeling: Based on the improved Tianying optimization algorithm to optimize the multi-modal deep learning model algorithm, create, train, and optimize the model for the sample data in Step 3; Step 5, predict the prediction set according to the model obtained in Step 4 to get the latest appropriate value of the governing valve opening, and store it in the database; Step 6, governing valve opening control: Compare the latest appropriate value of the governing valve opening obtained in Step 5 with the appropriate value in the previous statistical cycle. If the absolute value of the change amount exceeds the threshold of 0.1 and lasts for 5 seconds, then transmit the appropriate value of the governing valve opening to the controller for governing valve opening adjustment, otherwise no adjustment is made; In Step 6, set the upper and lower limit thresholds of the main steam pressure. If it is lower than the lower limit value, quickly reduce the governing valve opening. If the pressure cannot be restored to the normal range during the regulation process, maintain the turbine at a speed of 3000 rpm and trip the unit; If the main steam pressure is too low, immediately automatically close the main steam valve for shutdown; If it is higher than the upper limit value, automatically adjust the governing valve opening to 100%.

2. The intelligent control method for sintering waste heat power generation based on data driving according to claim 1, wherein, The specific data preprocessing in Step 2 includes: Perform mean filtering on the collected historical data with a time period of 20 seconds, perform normalization on the filtered data, delay the time length of each index variable to the turbine, and align all index values in time.

3. A data-driven intelligent control method for sintering waste heat power generation according to claim 1, characterized in that The specific formula for calculating the steam consumption rate in Step 3 is: Steam consumption rate = total steam consumption / power generation.

4. A data-driven intelligent control method for sintering waste heat power generation according to claim 1, characterized in that The improved Tianying optimization algorithm to optimize the multi-modal deep learning model algorithm in Step 4 is specifically: Use the main steam temperature, main steam flow rate, and main steam pressure as input variables, and the governing valve opening as the output variable. Input it into the CNN layer for local feature extraction, and its output enters the BiGRU-Transformer layer. Among them, the Transformer layer uses the multi-head self-attention mechanism, and the model automatically focuses on the main influencing variables, and finally maps the fully connected layer to the prediction target.

5. A data-driven intelligent control method for sintering waste heat power generation according to claim 1, characterized in that Step 4 specifically includes: Step 4.1, initialize and generate the initial solution: Use the Tent chaotic mapping to initialize the parameters of the Tianying optimization algorithm, and randomly generate the initial parameters of the CNN-BiGRU-Transformer model; Step 4.2, fitness calculation: Use the parameters of each individual to train the CNN-BiGRU-Transformer model, and use the validation set to evaluate the prediction performance of the model; Step 4.3, iterative optimization: Use the iterative strategy of the Tianying optimization algorithm, adjust the global search with an adaptive weight, and update the parameters of each individual in the population until the termination condition is met; Step 4.4, Optimal parameter selection: Select the parameters corresponding to the individual with the smallest fitness value as the final model parameters to obtain the prediction model.

6. The intelligent control method for sintering waste heat power generation based on data driving according to claim 5, characterized in that, The Tent chaotic mapping in the step 4.1 is specifically as follows: Use the Tent chaotic mapping to generate N chaotic sequences, where N is the population size, map the chaotic sequences to the search space to generate the initial population, and the mathematical model of the Tent chaotic mapping is: Among them, is , the i-th data.

7. A data-driven intelligent control method for sintering waste heat power generation according to claim 5, characterized in that, The specific formula for the adaptive weight in the step 4.3 is: where t is the current iteration number, and is the maximum number of iterations; Adopt the adaptive weight to adjust the global search, and the formula is: Among them, , is the current position, , is the adaptive position.

Images