CN-121981174-A - Pulverized coal distribution uniformity control method and device, electronic equipment and storage medium

CN121981174ACN 121981174 ACN121981174 ACN 121981174ACN-121981174-A

Abstract

The application discloses a coal powder distribution uniformity control method, a device, electronic equipment and a storage medium, which relate to the technical field of thermal power generation and comprise the steps of obtaining a historical operation data set and a current monitoring data set, performing characteristic alignment treatment, taking the historical operation data set as a source domain and the current monitoring data set as a target domain, training a control model by adopting an integrated migration learning framework, dynamically adjusting contribution of a sample weight balance domain and the model training, improving model robustness by adopting an integrated learning strategy, predicting key state parameters of a target process by utilizing a control model which is completed by training to generate a prediction result, generating a control instruction according to the prediction result, adjusting action of an actuating mechanism, and feeding back an adjusted process state to the control model to realize online optimization. The method has the technical effects of improving the control precision of the uniformity of pulverized coal distribution, enhancing the adaptability of the model to coal fluctuation and working condition change, and guaranteeing the long-term stable operation of the pulverizing system.

Inventors

CAO WEI
ZHAO WEIDONG
GAO MINGYANG
YAN HAIBO
LIU GUANGYUAN
JU RONG
LI ZHENDONG
YANG PEIJUN
YUAN PANFENG
WANG YAN
LIU DINGPO

Assignees

国能怀安热电有限公司
西安西热锅炉环保工程有限公司

Dates

Publication Date: 20260505
Application Date: 20251211

Claims (10)

1. A method for controlling uniformity of pulverized coal distribution, comprising: Acquiring a historical operation data set and a current monitoring data set, and performing feature alignment processing on the historical operation data set and the current monitoring data set; Taking the historical operation data set as a source domain, taking the current monitoring data set as a target domain, adopting an integrated transfer learning framework to train a control model, wherein the contribution of the source domain and the target domain to model training is balanced by dynamically adjusting sample weight, and adopting an integrated learning strategy to improve the robustness of the model; predicting key state parameters of the target process by using the trained control model to generate a prediction result; generating a control instruction according to the prediction result, adjusting the action of the executing mechanism according to the control instruction, and feeding back the adjusted process state to the control model to realize online optimization.
2. The method of claim 1, wherein training a control model using an integrated transfer learning framework comprises: constructing a mixed model integrating data driving and knowledge driving as a basic model; in the iterative training process, the weight of the source domain sample is adaptively updated according to the classification performance of the sample, and the method comprises the steps of carrying out weight enhancement or inhibition on the source domain sample according to the comparison result of the accumulated error classification times and a preset threshold value, and carrying out weight adjustment on the target domain sample according to the prediction error of the target domain sample so as to focus on the sample which is difficult to predict correctly.
3. The method of claim 2, wherein the weight enhancement or suppression is performed on the source domain samples according to the comparison result of the accumulated error classification times and the preset threshold, and the weight adjustment is performed on the target domain samples according to the prediction errors, and the method comprises the following steps: for source domain samples whose cumulative number of misclassifications does not exceed the threshold, increasing their weight to strengthen learning; for source domain samples whose cumulative number of error classifications exceeds the threshold, reducing their weight to suppress their interference with model training; and for the target domain samples, dynamically adjusting weights according to the prediction errors of the target domain samples, and giving higher weights to samples with larger prediction errors so as to improve the attention of the model.
4. The method of claim 1, wherein predicting the key state parameters of the target process using the trained control model to generate the prediction result comprises: Acquiring key physical quantity data reflecting the process state in real time through a multi-source sensing network deployed on an industrial site; Inputting the key physical quantity data into the control model to obtain an optimized control quantity for enabling the process state to trend to a target value.
5. The method of claim 1, wherein the generating a control command according to the prediction result, adjusting the action of the actuator according to the control command, and feeding back the adjusted process state to the control model to realize online optimization, comprises: According to the optimization control instruction, the action of the executing mechanism is adjusted to change the process parameter; And feeding back the adjusted process state to the control model so as to optimize model parameters on line and correct subsequent control instructions.
6. The pulverized coal distribution uniformity control method according to claim 1, characterized by further comprising: and establishing a self-adaptive compensation mechanism of the sensor data, and starting a calibration process to correct the data deviation when the deviation between the monitoring data and the expected value established based on the historical data exceeds a preset range, so as to ensure the reliability of the data input into the control model.
7. A pulverized coal distribution uniformity control apparatus, comprising: the alignment module is configured to acquire a historical operation data set and a current monitoring data set, and perform characteristic alignment processing on the historical operation data set and the current monitoring data set; The training module is configured to take the historical operation data set as a source domain, the current monitoring data set as a target domain, train a control model by adopting an integrated transfer learning framework, balance the contribution of the source domain and the target domain to model training by dynamically adjusting sample weights, and improve the robustness of the model by adopting an integrated learning strategy; The prediction module is configured to predict key state parameters of the target process by using the trained control model so as to generate a prediction result; the execution module is configured to generate a control instruction according to the prediction result, adjust the action of the execution mechanism according to the control instruction, and feed back the adjusted process state to the control model to realize online optimization.
8. An electronic device, comprising: At least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the pulverized coal distribution uniformity control method of any one of claims 1-6.
9. A non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the pulverized coal distribution uniformity control method according to any one of claims 1 to 6.
10. A computer program product comprising a computer program which, when executed by a processor, implements the pulverized coal distribution uniformity control method according to any one of claims 1-6.

Description

Pulverized coal distribution uniformity control method and device, electronic equipment and storage medium Technical Field The application relates to the technical field of thermal power generation, in particular to a pulverized coal distribution uniformity control method, a pulverized coal distribution uniformity control device, electronic equipment and a storage medium. Background The uniformity of pulverized coal distribution of a pulverizing system of a thermal power generating unit is a core factor for determining the combustion efficiency of a boiler, the emission level of pollutants and the running stability of the unit, and particularly when complex working conditions such as deep peak shaving, coal blending and the like of the thermal power generating unit are increasingly frequent, higher requirements are put forward on the accurate control and dynamic adaptation capability of pulverized coal distribution. The good pulverized coal distribution effect can avoid potential safety hazards such as local heat load deflection, overheat heating surface and the like in the furnace, and can optimize the combustion process and reduce the energy consumption and the environmental protection treatment cost, so that the pulverized coal distribution uniformity control technology is always a key research direction in the field of optimizing operation of the thermal power generating unit. In the current industrial field, the uniformity control of coal powder distribution at the outlet of a coal mill mainly adopts a traditional control method, and the core comprises fuzzy control, an Adaptive Neural Fuzzy Inference System (ANFIS), proportional-integral-derivative (PID) control and the like. The fuzzy control converts continuous input variables such as coal powder flow rate, concentration and the like into a fuzzy set through a membership function, a control rule is constructed based on expert experience, actual control quantity is output after processing by a Mamdani reasoning method and a gravity center method, ANFIS is introduced into a neural network on the basis of the fuzzy control, the control precision is improved through an optimized control rule, PID control realizes closed-loop adjustment by means of a proportional-integral-derivative link, dynamic balance of coal powder distribution is maintained, and the method provides basic technical support for coal powder distribution control under conventional working conditions. The traditional control method has obvious defects in practical application, is difficult to adapt to the control requirements of complex working conditions, has limited control precision, has insufficient regulation capability on pulverized coal distribution deviation, has larger powder quantity distribution deviation of a diffusion type pulverized coal distributor, has further increased deviation under a variable load working condition, has difficult control of steady state deviation in an ideal range when PID control fluctuates in coal quality, has obvious control lag in an ANFIS model under dynamic working conditions such as deep peak regulation and the like, has long adjustment time in the coal type switching process, is easy to cause the problems of overlarge deviation of outlet temperature of a boiler burner and the like, has weak generalization capability, has easy failure condition in a fuzzy control rule base when coal quality characteristics (such as ash content and moisture content) change greatly, has obviously reduced prediction precision of the ANFIS model, has greatly reduced fineness percent of coal and is difficult to adapt to the actual running requirements of various working conditions. Disclosure of Invention The application provides a pulverized coal distribution uniformity control method, a pulverized coal distribution uniformity control device, electronic equipment and a storage medium. The problems of weak model generalization capability, limited control precision and lack of dynamic self-adaptive adjustment of a system caused by insufficient data adaptability and domain difference in the related technology can be solved. According to a first aspect of the present application, there is provided a pulverized coal distribution uniformity control method, including: Acquiring a historical operation data set and a current monitoring data set, and performing feature alignment processing on the historical operation data set and the current monitoring data set; Taking a historical operation data set as a source domain, taking a current monitoring data set as a target domain, adopting an integrated transfer learning framework to train a control model, wherein the contribution of the source domain and the target domain to model training is balanced by dynamically adjusting sample weights, and adopting an integrated learning strategy to improve the robustness of the model; predicting key state parameters of the target process by using the trained control model to generate a pr