Boiler combustion optimization system with self-learning function
By introducing self-learning functions into the boiler combustion optimization system and optimizing the combustion process using neural networks and genetic algorithms, the problems of low efficiency and large emissions of the existing system are solved, and combustion efficiency improvement and emission reduction are achieved, which meets environmental protection requirements.
Patent Information
- Application Number
- CN202411004708.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-25
- Publication Date
- 2024-11-12
AI Technical Summary
The existing boiler combustion optimization system lacks a self-learning and updating mechanism, resulting in low combustion efficiency and high nitrogen oxide emissions.
A boiler combustion optimization system with self-learning function was designed, including data acquisition, effective data extraction, simulation training set creation, simulation model training, optimization data set creation, optimization model training, historical data storage and model self-update The module optimizes the combustion process through neural networks and genetic algorithms, generates optimization models and automatically updates.
Through continuous update of the optimization model, the combustion efficiency of the boiler is improved, the emission of nitrogen oxides is reduced, the energy utilization efficiency is improved, the operation costs are reduced, and the strict environmental protection regulations are complied with strict environmental regulations.
Smart Images

Figure CN118940628A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the cross-technical field of environmental protection, energy saving and artificial intelligence, and in particular to a boiler combustion optimization system with a self-learning function. Background Art
[0003] At present, thermal power generation is still the main source of electricity in my country, and boiler combustion optimization is an important technical means for thermal power units to achieve energy conservation and emission reduction. With the development of information construction in the power industry, the thermal control systems of most combustion equipment in my country are basically equipped with distributed control systems (DCS) to realize monitoring and alarm, data collection, and automatic control of the unit during the operation of the equipment. At the same time, the existing DCS system generates and accumulates a large amount of operating data in real time as the unit operates. In recent years, the development of artificial intelligence technologies such as neural networks and intelligent optimization algorithms has also provided effective tools for deep mining and utilization of accumulated DCS system data, and the intelligence of boiler combustion optimization systems has also become an inevitable development trend. However, the existing boiler combustion optimization system lacks a self-learning and updating mechanism, which is not conducive to improving the combustion efficiency of boilers and reducing nitrogen oxide emissions in the long run.
[0004] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention
[0005] In view of this, the purpose of the embodiments of the present application is to provide a boiler combustion optimization system with a self-learning function to overcome the above-mentioned technical problems existing in the existing related technologies.
[0006] To this end, the specific technical solution adopted by the present invention is as follows:
[0007] A boiler combustion optimization system with a self-learning function comprises a data acquisition module, an effective data extraction module, a simulation training set creation module, a simulation model training module, an optimization data set creation module, an optimization model training module, a historical data storage module and a model self-update module.
[0008] Among them, the data acquisition module is used to extract the phased historical data generated during the operation process from the distributed control system of the unit using the communication protocol.
[0009] The effective data extraction module is used to merge the new operation data with the historical data, perform data cutting and eliminate erroneous data to form a qualified data set.
[0010] The simulation training set creation module is used to rebalance the data distribution of the qualified data set according to the pre-set data balance index to form the simulation training set data.
[0011] The simulation model training module is used to train the simulation model of core indicators based on the simulation training set data and generate the corresponding simulation model.
[0012] The optimization data set creation module is used to obtain the best standard, and based on the best standard, perform full-data genetic algorithm optimization on qualified data and generate an optimization training set.
[0013] The optimization model training module is used to train the optimization model according to the optimization training set and generate the corresponding optimization model.
[0014] The historical data storage module is used to back up the simulation training set data, optimization training set, optimization model and simulation model.
[0015] The model self-update module is used to provide an automatic reminder function that the model can be upgraded.
[0016] Furthermore, the phased unit distributed control system operation data is the operation data of several months;
[0017] Historical data refers to data several months before the earliest date of extraction of operating data.
[0018] Furthermore, the simulation training set creation module includes a grouping module, a number determination module and a balancing module;
[0019] Among them, the grouping module is used to group the data of the qualified data set based on the date, unit load and total coal supply as conditions;
[0020] The number determination module is used to determine the number of data records in each group;
[0021] The balancing module is used to determine whether the number of data records in any group exceeds a preset data balancing index. If so, some records in the corresponding group are randomly discarded.
[0022] Furthermore, the simulation model training module includes a simulation data partitioning module and a training judgment module;
[0023] The simulation data division module is used to divide the simulation training set data into a first training data set, a first verification set and a first test data set according to a certain ratio;
[0024] The training judgment module is used to train the simulation model based on the core indicators, and during training, if the degree of fit reaches the target, the corresponding simulation model is generated.
[0025] Furthermore, when training the simulation model based on the core indicators, the output of the neural network model is defined;
[0026] The neural network model is trained using the first training data set and the first validation set to obtain a simulation model;
[0027] The simulation model is verified using the first test data set. If the model fit reaches the target, the training is successful. Otherwise, the neural network model is retrained.
[0028] Furthermore, the core indicators include exhaust gas temperature, fly ash carbon content, flue gas CO content, flue gas NO x content, flue gas oxygen content, furnace pressure difference;
[0029] The output of the neural network model includes flue gas NO x content, exhaust temperature, fly ash carbon content, flue gas CO content, flue gas oxygen content and furnace pressure difference.
[0030] Furthermore, when the qualified data is optimized by genetic algorithm based on the best standard and the optimization training set is generated, the NO reduction x The objective function of emission and the objective function of improving combustion efficiency will reduce NO x The objective function of emission and the objective function of improving combustion efficiency are used as the objective function of the genetic algorithm, and the dimension of the objective function of the genetic algorithm is the number of adjustment variables;
[0031] Perform results analysis on the genetic algorithm to determine the optimal criteria for the number of iterations and related parameter values of the genetic algorithm;
[0032] The genetic algorithm is executed by adopting the best standard of relevant parameter values and taking the qualified data set as input. The execution results of the genetic algorithm are the optimization results and the optimization recommended values of the adjustment variables. At the same time, the execution results of the genetic algorithm are collected and organized into an optimization training set.
[0033] Furthermore, the optimization model training module includes an optimization data partitioning module, a combustion efficiency optimization module, and a NO reduction module. x Emission module;
[0034] The optimization data partitioning module is used to divide the data of the optimization training set into a second training data set, a second verification set and a second test data set;
[0035] Optimizing combustion efficiency module, used to train the optimization combustion efficiency model;
[0036] Reducing NO x Emissions module for training to reduce NO x Emission model.
[0037] Furthermore, when training the combustion efficiency optimization model, the output of the neural network model is defined;
[0038] The neural network model is trained using the second training data set and the second validation set to obtain an optimized combustion efficiency model, and during the training process, a simulation model is used to predict specific operating condition parameters, and the prediction results meet the optimization constraint conditions;
[0039] The optimized combustion efficiency model is verified using the second test data set. If the fit reaches the target, the training is successful. Otherwise, the neural network model is retrained.
[0040] Furthermore, training reduces NO x When emitting a model, define the output of the neural network model;
[0041] The neural network model is trained using the second training data set and the second validation set to obtain an optimized combustion efficiency model, and during the training process, a simulation model is used to predict specific operating condition parameters, and the prediction results meet the optimization constraint conditions;
[0042] The second test data set was used to reduce NO x The emission model is verified. If the fit reaches the target, the training is successful. Otherwise, the neural network model is retrained.
[0043] Embodiments of the present invention include the following benefits:
[0044] (1) The present invention provides a boiler combustion optimization system with a self-learning function, which can generate a corresponding optimization model based on the staged unit DCS system operation data and historical data, and after backing up the old version of the model, update and replace the optimization model, so that the unit can operate according to the new optimization model. The model is constantly self-learning and updated over and over again, and the performance of the boiler combustion optimization system is continuously improved, thereby ultimately improving the boiler's combustion efficiency and reducing nitrogen oxide emissions.
[0045] (2) The present invention significantly improves energy efficiency and reduces energy consumption by optimizing the combustion process. This not only reduces operating costs, but also helps to increase economic benefits and reduce emissions of harmful substances such as nitrogen oxides. By precisely controlling combustion conditions, environmental pollution is reduced, complying with increasingly stringent environmental regulations. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0047] Figure 1It is a framework diagram of a boiler combustion optimization system with a self-learning function according to an embodiment of the present invention.
[0048] In the figure:
[0049] 1. Communication module; 2. Data acquisition module; 3. Effective data extraction module; 4. Simulation training set creation module; 5. Simulation model training module; 6. Optimal data set creation module; 7. Optimal model training module; 8. Historical data storage module; 9. Model self-update module. DETAILED DESCRIPTION
[0050] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0051] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and 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 contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.
[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.
[0053] According to an embodiment of the present invention, a boiler combustion optimization system with a self-learning function is provided. The optimization system defines a set of fully automatic workflows. The dedicated host computer of the system can automatically retrieve the unit operation data and historical data from the boiler DCS system online, and apply the preset cleaning rules to process the data to form a training data set, and then use the neural network to train the simulation model of the core indicators (exhaust gas temperature, fly ash carbon content, flue gas CO content, flue gas NO x After the simulation model is generated, the GA genetic algorithm is used to perform full data optimization according to the preset optimization rules and iteration times to generate a training set for optimization. Then, the neural network is used for training to obtain the optimal combustion efficiency and NO reduction. xThe system is a comprehensive optimization model composed of two sub-models: emission and control. After the new version of the model is trained, the system automatically determines whether the model update conditions are met. If so, the model is updated and replaced, and the old model is backed up at the same time. The above process is repeated, and the model is automatically updated regularly based on the latest historical data, which can continuously improve the operation effect of the optimization system and adapt to the changes in the actual operating conditions of the unit without the need for manual intervention by the user. Effectively improve the combustion efficiency of the boiler and reduce nitrogen oxide emissions.
[0054] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to an embodiment of the present invention, a boiler combustion optimization system with a self-learning function includes a communication module 1, a data acquisition module 2, a valid data extraction module 3, a simulation training set creation module 4, a simulation model training module 5, an optimization data set creation module 6, an optimization model training module 7, a historical data storage module 8 and a model self-update module 9.
[0055] Wherein, the communication module 1 is used to communicate with a DCS system (distributed control system).
[0056] The data acquisition module 2 is used to extract the periodic historical data generated during the operation from the distributed control system of the unit using the relevant communication protocol.
[0057] In one embodiment, the phased unit distributed control system operation data is operation data of several months (generally 3-6 months);
[0058] Historical data refers to data from several months (6-24 months) before the earliest date of extraction of operating data.
[0059] The effective data extraction module 3 is used to merge the new operation data with the historical data, cut the data according to a certain range, and eliminate the erroneous data to form a qualified data set.
[0060] The simulation training set creation module 4 is used to rebalance the data distribution of the qualified data set according to the pre-set data balance index to form the simulation training set data.
[0061] In one embodiment, the simulation training set creation module 4 includes a grouping module, a number determination module and a balancing module;
[0062] Among them, the grouping module is used to group the data of the qualified data set based on the date, unit load and total coal supply as conditions;
[0063] The number determination module is used to determine the number of data records in each group;
[0064] The balancing module is used to determine whether the number of data records in any group exceeds a preset data balancing index. If so, some records in the corresponding group are randomly discarded to achieve a relative balance in the number of records in each group.
[0065] The simulation model training module 5 is used to perform simulation model training of core indicators based on the simulation training set data and generate a corresponding simulation model.
[0066] In one embodiment, the simulation model training module 5 includes a simulation data partitioning module and a training judgment module;
[0067] The simulation data partitioning module is used to divide the simulation training set data into a first training data set, a first validation set and a first test data set according to a certain ratio (for example, according to a ratio of 6:3:1);
[0068] The training judgment module is used to train the simulation model based on the core indicators. During the training, if the degree of fit reaches the target, for example, the goal of satisfying the degree of fit R2 ≥ 98% is met, a corresponding simulation model is generated.
[0069] In one embodiment, when training a simulation model based on core indicators, an output of a neural network model is defined;
[0070] The neural network model is trained using the first training data set and the first validation set to obtain a simulation model;
[0071] The simulation model is verified using the first test data set. If the model fit reaches the target, the training is successful. Otherwise, the neural network model is retrained.
[0072] In one embodiment, the core indicators include exhaust gas temperature, fly ash carbon content, flue gas CO content, flue gas NO x content, flue gas oxygen content, furnace pressure difference;
[0073] The output of the neural network model includes flue gas NO x content, exhaust temperature, fly ash carbon content, flue gas CO content, flue gas oxygen content and furnace pressure difference.
[0074] Flue gas NO x Training of content simulation model:
[0075] a. Define the output of the neural network model, namely, flue gas NO x content;
[0076] b. Use the first training data set and the first validation set obtained above to train the model;
[0077] c. Use the first test data set to verify the simulation model. When the model fit R2 ≥ 98%, the model training is successful; when the model fit R2 < 98%, retrain the model.
[0078] Training of exhaust temperature simulation model:
[0079] a. Define the model network output, i.e., exhaust temperature;
[0080] b. Use the first training data set and the first validation set obtained above to train the model;
[0081] c. Use the first test data set to verify the simulation model. When the model fit R2 ≥ 98%, the model training is successful; when the model fit R2 < 98%, retrain the model.
[0082] Training of the simulation model for carbon content in fly ash:
[0083] a. Define the model network output, i.e. the carbon content of fly ash;
[0084] b. Use the first training data set and the first validation set obtained above to train the model;
[0085] c. Use the first test data set to verify the simulation model. When the model fit R2 ≥ 98%, the model training is successful; when the model fit R2 < 98%, retrain the model.
[0086] Training of flue gas CO content simulation model:
[0087] a. Define the model network output, i.e., flue gas CO content;
[0088] b. Use the first training data set and the first validation set obtained above to train the model;
[0089] c. Use the first test data set to verify the simulation model. When the model fit R2 ≥ 98%, the model training is successful; when the model fit R2 < 98%, retrain the model.
[0090] Training of flue gas oxygen content simulation model:
[0091] a. Define the model network output, i.e., the oxygen content of flue gas;
[0092] b. Use the first training data set and the first validation set obtained above to train the model;
[0093] c. Use the first test data set to verify the simulation model. When the model fit R2 ≥ 98%, the model training is successful; when the model fit R2 < 98%, retrain the model.
[0094] Training of furnace pressure difference simulation model:
[0095] a. Define the model network output, i.e. furnace pressure difference;
[0096] b. Use the first training data set and the first validation set obtained above to train the model;
[0097] c. Use the first test data set to verify the simulation model. When the model fit R2 ≥ 98%, the model training is successful; when the model fit R2 < 98%, retrain the model.
[0098] The optimization data set creation module 6 is used to obtain the best standard (according to the pre-set optimization constraints, automatically determine the optimal standard for the number of iterations of the genetic algorithm and the value of the relevant parameters), and perform full-data genetic algorithm optimization on qualified data based on the best standard, and generate an optimization training set.
[0099] In one embodiment, when the qualified data is optimized by the full data genetic algorithm based on the best standard and the optimization training set is generated, the NO reduction x The objective function of emission and the objective function of improving combustion efficiency will reduce NO x The objective function of emission and the objective function of improving combustion efficiency are used as the objective function of the genetic algorithm, and the dimension of the objective function of the genetic algorithm is the number of adjustment variables;
[0100] Perform results analysis on the genetic algorithm to determine the optimal criteria for the number of iterations and related parameter values of the genetic algorithm;
[0101] The genetic algorithm is executed by adopting the best standard of relevant parameter values and taking the qualified data set as input. The execution results of the genetic algorithm are the optimization results and the optimization recommended values of the adjustment variables. At the same time, the execution results of the genetic algorithm are collected and organized into an optimization training set.
[0102] In addition, reducing NO x The emission objective function, in the execution logic of this function, uses the target variable simulation model trained in the simulation model training module 5 to simulate and predict a certain operating condition data. While satisfying the optimization constraints, the return value of the objective function is NO x Predicted value of content;
[0103] The objective function for improving combustion efficiency uses the target variable simulation model trained in the simulation model training module 5 in the execution logic of this function to simulate and predict a certain operating condition data and calculate the corresponding combustion efficiency loss. While optimizing the constraints, the return value of the objective function is the loss value of the combustion efficiency.
[0104] The optimization model training module 7 is used to train the optimization model according to the optimization training set and generate a corresponding optimization model.
[0105] In one embodiment, the optimization model training module 7 includes an optimization data partitioning module, a combustion efficiency optimization module, and a NO reduction module. x Emission module;
[0106] The optimization data division module is used to divide the data of the optimization training set into a second training data set, a second verification set and a second test data set according to a ratio of 6:3:1;
[0107] Optimizing combustion efficiency module, used to train the optimization combustion efficiency model;
[0108] Reducing NO x Emissions module for training to reduce NO x Emission model.
[0109] At the same time, when the training process meets the goal of fitting degree R2 ≥ 98%, the corresponding optimization model is generated.
[0110] In one embodiment, when training the combustion efficiency optimization model, the output of the neural network model is defined as follows: 36 secondary air door opening instructions;
[0111] The neural network model is trained using the second training data set and the second validation set to obtain an optimized combustion efficiency model, and during the training process, the simulation model in the simulation model training module 5 is used to predict the specific operating condition parameters, and the prediction results meet the optimization constraint conditions;
[0112] The optimized combustion efficiency model is verified using the second test data set. If the fit reaches the target, the training is successful. Otherwise, the neural network model is retrained. For example, when the model fit R2 ≥ 98%, the model training is successful; when the model fit R2 < 98%, the model is retrained.
[0113] In one embodiment, training to reduce NO x When the emission model is used, the output of the neural network model is defined as follows: 36 secondary air door opening instructions;
[0114] The neural network model is trained using the second training data set and the second validation set to obtain an optimized combustion efficiency model, and during the training process, the simulation model in the simulation model training module 5 is used to predict the specific operating condition parameters, and the prediction results meet the optimization constraint conditions;
[0115] The second test data set was used to reduce NO x The emission model is verified. If the fit reaches the target, the training is successful. Otherwise, the neural network model is retrained. For example, when the model fit R2 ≥ 98%, the model training is successful; when the model fit R2 < 98%, the model is retrained.
[0116] The historical data storage module 8 is used to back up the simulation training set data, the optimization training set, the optimization model and the simulation model to prevent loss and prepare for subsequent inspection.
[0117] The model self-update module 9 is used to provide an automatic reminder function that the model can be upgraded, which means that the execution end (platform or DCS) provides an automatic reminder function that the system model can be upgraded. After the personnel confirms that the model upgrade needs to be executed, the new model replaces the existing optimization model.
[0118] A boiler combustion optimization system with a self-learning function is used for large thermal power generating units. Boiler combustion is a complex and nonlinear process. There are many parameters that need to be adjusted for combustion. Signal variables include load, exhaust gas temperature, ambient temperature, water supply, water supply temperature, total fuel amount entering the furnace, coal feeder opening, total air volume, total primary air volume, total secondary air volume, burner swing angle, primary air damper opening, secondary air damper opening, burnout air damper opening, total primary air pressure, coal quality parameters (calorific value, moisture, ash content, volatile matter, fly ash carbon content, slag carbon content, C, H, O, N, S), economizer outlet oxygen content, furnace pressure difference, etc.
[0119] A combustion optimization system with self-learning function is based on operating data and historical data, and is data-driven to form an optimized combustion efficiency model and reduce NO x Emission model. The system is configured with corresponding workstations, configuration stations and data storage stations to form the operating platform of the combustion optimization software. By adding a dedicated host computer for the system to establish a communication connection with the unit DCS system, the operating data and historical data related to the combustion optimization system are collected to analyze and optimize the combustion data, and the optimization results are sent to the DCS to realize combustion closed-loop optimization control, and the combustion optimization related data are sent to the data center.
[0120] Implementation process of a boiler combustion optimization system with self-learning function:
[0121] (1) Collect and analyze the power plant DCS data to obtain the power plant's phased operation data and historical data that can be used for analysis;
[0122] (2) Design the power plant boiler optimization system, define system functions, data screening conditions, AI model input and output conditions, and system deployment methods;
[0123] (3) Clean and organize the analyzed DCS data and extract model training data;
[0124] (4) Based on the data, the simulation models of various target indicators are trained to establish a power plant boiler combustion simulator;
[0125] (5) Obtaining a boiler combustion optimization model that meets optimization constraints based on data training;
[0126] (6) According to the actual conditions of the power plant, the function of the boiler combustion optimization system is optimized and a combustion optimization system that can be used in a power plant is established;
[0127] (7) On-site deployment: the combustion optimization system obtains specific data from the power plant DCS system and sends the optimization instructions calculated by the combustion optimization system to the power plant DCS system;
[0128] (8) Complete the software and hardware testing of the development object to ensure that the power plant combustion optimization system meets the usage standards;
[0129] (9) Realize the online deployment of the power plant combustion optimization system and complete the system operation.
[0130] In summary, the boiler combustion optimization system with self-learning function provided by the present invention can generate a corresponding optimization model according to the staged unit DCS system operation data and historical data, and after backing up the old version of the model, update and replace the optimization model, so that the unit can operate according to the new optimization model, over and over again, and the model will be continuously self-learned and updated to continuously improve the performance of the boiler combustion optimization system, thereby ultimately improving the combustion efficiency of the boiler and reducing the emission of nitrogen oxides. The present invention significantly improves energy utilization efficiency and reduces energy consumption by optimizing the combustion process. This not only reduces operating costs, but also helps to increase economic benefits and reduce emissions of harmful substances such as nitrogen oxides. By precisely controlling combustion conditions, environmental pollution is reduced, complying with increasingly stringent environmental protection regulations.
[0131] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional charging modules / units in the systems and devices may be implemented as software, firmware, hardware, or appropriate combinations thereof.
[0132] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0133] 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 principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A boiler combustion optimization system with self-learning function, characterized in that: The boiler combustion optimization system with self-learning function includes a data acquisition module, an effective data extraction module, a simulation training set creation module, a simulation model training module, an optimization data set creation module, an optimization model training module, a historical data storage module and a model self-update module; Wherein, the data acquisition module is used to extract the periodic historical data generated during the operation process from the distributed control system of the unit using the communication protocol; The effective data extraction module is used to merge the new operation data with the historical data, perform data segmentation and remove erroneous data to form a qualified data set; The simulation training set creation module is used to rebalance the data distribution of the qualified data set according to the pre-set data balance index to form the simulation training set data; The simulation model training module is used to perform simulation model training of core indicators based on simulation training set data and generate a corresponding simulation model; The optimization data set creation module is used to obtain the best standard, and perform full-data genetic algorithm optimization on qualified data based on the best standard, and generate an optimization training set; The optimization model training module is used to train the optimization model according to the optimization training set and generate a corresponding optimization model; The historical data storage module is used to back up the simulation training set data, the optimization training set, the optimization model and the simulation model accordingly; The model self-update module is used to provide an automatic prompt function that the model can be upgraded.
2. A boiler combustion optimization system with self-learning function according to claim 1, characterized in that: The phased unit distributed control system operation data is the operation data of several months; Historical data refers to data several months before the earliest date of extraction of operating data.
3. A boiler combustion optimization system with self-learning function according to claim 2, characterized in that: The simulation training set creation module includes a grouping module, a number determination module and a balancing module; The grouping module is used to group the data of the qualified data set based on the date, unit load and total coal supply as conditions; The number determination module is used to determine the number of data records in each group; The balancing module is used to determine whether the number of data records in any group exceeds a preset data balancing index. If so, some records in the corresponding group are randomly discarded.
4. A boiler combustion optimization system with self-learning function according to claim 3, characterized in that: The simulation model training module includes a simulation data division module and a training judgment module; The simulation data division module is used to divide the simulation training set data into a first training data set, a first verification set and a first test data set according to a certain ratio; The training judgment module is used to train the simulation model based on the core indicators, and during training, if the degree of fit reaches the target, a corresponding simulation model is generated.
5. A boiler combustion optimization system with self-learning function according to claim 4, characterized in that: When the simulation model is trained based on the core indicators, the output of the neural network model is defined; The neural network model is trained using the first training data set and the first validation set to obtain a simulation model; The simulation model is verified using the first test data set. If the model fit reaches the target, the training is successful. Otherwise, the neural network model is retrained.
6. A boiler combustion optimization system with self-learning function according to claim 5, characterized in that: The core indicators include exhaust gas temperature, fly ash carbon content, flue gas CO content, flue gas NO x content, flue gas oxygen content, furnace pressure difference; The output of the neural network model includes flue gas NO x content, exhaust temperature, fly ash carbon content, flue gas CO content, flue gas oxygen content and furnace pressure difference.
7. A boiler combustion optimization system with self-learning function according to claim 6, characterized in that: When the qualified data is optimized by genetic algorithm based on the best standard and the optimization training set is generated, the NO reduction x The objective function of emission and the objective function of improving combustion efficiency will reduce NO x The objective function of emission and the objective function of improving combustion efficiency are used as the objective function of the genetic algorithm, and the dimension of the objective function of the genetic algorithm is the number of adjustment variables; Perform results analysis on the genetic algorithm to determine the optimal criteria for the number of iterations and related parameter values of the genetic algorithm; The genetic algorithm is executed by adopting the best standard of relevant parameter values and taking the qualified data set as input. The execution results of the genetic algorithm are the optimization results and the optimization recommended values of the adjustment variables. At the same time, the execution results of the genetic algorithm are collected and organized into an optimization training set.
8. A boiler combustion optimization system with self-learning function according to claim 7, characterized in that: The optimization model training module includes an optimization data partitioning module, a combustion efficiency optimization module and a NO reduction module. x Emission module; The optimization data partitioning module is used to divide the data of the optimization training set into a second training data set, a second verification set and a second test data set; The optimized combustion efficiency module is used to train the optimized combustion efficiency model; The reduction of NO x Emissions module for training to reduce NO x Emission model.
9. A boiler combustion optimization system with self-learning function according to claim 8, characterized in that: When the training optimization combustion efficiency model is used, the output of the neural network model is defined; The neural network model is trained using the second training data set and the second validation set to obtain an optimized combustion efficiency model, and during the training process, a simulation model is used to predict specific operating condition parameters, and the prediction results meet the optimization constraint conditions; The optimized combustion efficiency model is verified using the second test data set. If the fit reaches the target, the training is successful. Otherwise, the neural network model is retrained.
10. A boiler combustion optimization system with self-learning function according to claim 9, characterized in that: The training reduces NO x When emitting a model, define the output of the neural network model; The neural network model is trained using the second training data set and the second validation set to obtain an optimized combustion efficiency model, and during the training process, a simulation model is used to predict specific operating condition parameters, and the prediction results meet the optimization constraint conditions; The second test data set was used to reduce NO x The emission model is verified. If the fit reaches the target, the training is successful. Otherwise, the neural network model is retrained.
