Search

CN-121707077-B - Industrial coal blending optimization method, system and program product based on multi-agent cooperation

CN121707077BCN 121707077 BCN121707077 BCN 121707077BCN-121707077-B

Abstract

The invention provides an industrial coal blending optimization method, system and program product based on multi-agent cooperation, and relates to the technical field of industrial intelligent control and data processing. The method comprises the steps of obtaining production state data of a current time step, wherein the data comprise coal inventory, physicochemical indexes and raw material resource weight coefficients, inputting the data into a mechanism calculation module, calculating a reference proportioning vector by using a linear programming algorithm with the aim of minimizing comprehensive resource consumption indexes, inputting the data into a strategy correction intelligent body, outputting a correction action vector aiming at the reference proportioning and a confidence factor of a current working condition, and dynamically weighting and fusing the reference proportioning and the correction action. The double-layer coupling architecture with mechanism bottom protection and AI synergy is adopted, the confidence factor is used as a safety valve, and on the premise of ensuring production safety, the problems of poor nonlinear adaptability, feedback lag and the like of the traditional coal blending model are solved, and the global optimization of industrial raw material configuration is realized.

Inventors

  • LUO YUEQING
  • LI XINRONG
  • AN ZIWEN

Assignees

  • 华院计算技术(上海)股份有限公司

Dates

Publication Date
20260508
Application Date
20260211

Claims (9)

  1. 1. The industrial coal blending optimization method based on multi-agent cooperation is characterized by comprising the following steps of: step S1, acquiring production state data of the current time step, wherein the production state data comprises stock quantity of each coal type, physical and chemical indexes of raw coal, weight coefficients of raw material resources, quality constraint of target products and environmental characteristic data; s2, inputting the production state data into a preset mechanism calculation module, and calculating to obtain a reference proportioning vector meeting the quality constraint of the target product by using a linear programming algorithm and minimizing the comprehensive resource consumption index as an objective function ; S3, inputting the production state data into a pre-trained strategy correction agent, and outputting a correction motion vector for the reference proportioning vector Confidence factor for current operating conditions ; Step S4, according to the reference proportioning vector Correcting motion vector Confidence factor Weighting and fusing to generate final proportioning vector And the final proportioning vector is issued to a coal blending executing mechanism; And S5, acquiring actual output quality data after a hysteresis period, calculating a reward value based on the actual output quality data, and updating the strategy correction agent.
  2. 2. The industrial coal blending optimization method based on multi-agent cooperation according to claim 1, wherein the strategy modification agent comprises an input layer, a mask layer and an attention mechanism layer; in step S1 and step S3, when there is a missing value or an abnormal value in the raw coal physicochemical index in the production status data, the mask layer sets the corresponding feature value to zero, and the attention mechanism layer redistributes the weights of the remaining valid data to generate the corrected motion vector 。
  3. 3. The multi-agent synergy-based industrial coal blending optimization method of claim 1, wherein in step S4, the final proportioning vector is generated The calculation formula of (2) is as follows: ; Wherein Normalize is a normalization function for ensuring that the sum of the ratios is 100%, the confidence factor The strategy corrects the agent output to approach 0 when the fluctuation of the production state data is detected to exceed the preset threshold value or the sensor fails The system is degenerated to a control mode that dominates the mechanism calculation module.
  4. 4. The multi-agent cooperation-based industrial coal blending optimization method according to claim 1, wherein in step S5, the calculation formula of the reward value includes a quality penalty term, a consumption optimization term, and a run stability term: ; The consumption optimization term Based on the difference value between the reference proportion consumption index and the final proportion consumption index, carrying out normalization calculation through a hyperbolic tangent function; The quality penalty term Applying a second order squaring penalty when the actual yield quality data exceeds the target product quality constraint; The steady operation item And calculating a negative feedback value based on the Euclidean distance between the final proportioning vector of the current time step and the final proportioning vector of the last time step.
  5. 5. The industrial coal blending optimization method based on multi-agent cooperation according to claim 1, wherein in step S5, the obtaining actual output quality data after the hysteresis period specifically includes: the method comprises the steps of establishing a time-aligned data buffer area, storing production state data and correction motion vectors into the data buffer area at a time T, and after obtaining actual production quality data at a time t+T, backtracking and calling data at a corresponding time T from the data buffer area to calculate a reward value and update parameters.
  6. 6. The industrial coal blending optimization method based on multi-agent cooperation according to claim 1, wherein the mechanism calculation module internally comprises a technical index matrix for representing the contribution rate of each raw coal to the index of the blended coal; Calculating residual errors between the actual output quality data and the quality estimated by the mechanism calculation module based on the current parameters; and based on the residual error, reversely correcting a technical index matrix in the mechanism calculation module by using a gradient descent method, and updating a nominal value in the technical index matrix to be a working condition effective value.
  7. 7. The industrial coal blending optimization method based on multi-agent cooperation according to claim 1, wherein in step S3, the corrective action vector is Is constrained by a maximum allowable adjustment amplitude to limit the degree of deviation of the policy modifying agent from the reference proportioning vector.
  8. 8. An industrial coal blending optimization system based on multi-agent cooperation, which adopts the industrial coal blending optimization method based on multi-agent cooperation as claimed in any one of claims 1 to 7, and is characterized by comprising the following steps: The data acquisition module is used for acquiring production state data of the current time step, wherein the production state data comprises stock quantity of each coal type, raw coal physicochemical indexes, raw material resource weight coefficients, target product quality constraint and environmental characteristic data; the mechanism calculation module is used for running a linear programming algorithm, minimizing the comprehensive resource consumption index as an objective function, and calculating according to the production state data to obtain a reference proportioning vector ; The strategy correction module is used for operating a strategy correction agent based on the deep neural network and outputting a correction motion vector according to the production state data Confidence factor ; The decision fusion module is used for fusing the two data according to the formula Calculating a final proportioning vector, and controlling the coal feeder to execute through an execution control interface; And the feedback updating module is used for collecting actual output quality data after the hysteresis period, calculating a comprehensive rewarding value to update the strategy correction module, and calculating a quality prediction residual error to update the internal parameters of the mechanism calculation module.
  9. 9. A computer program product, characterized in that the computer program product comprises computer program code which, when run on a computer, causes the computer to implement the industrial coal blending optimization method based on multi-agent collaboration according to any one of claims 1 to 7.

Description

Industrial coal blending optimization method, system and program product based on multi-agent cooperation Technical Field The invention relates to the technical field of industrial intelligent control and data processing, in particular to an industrial coal blending optimization method, system and program product based on multi-agent cooperation. Background The key point of the industrial coal blending is that the industrial coal blending is an important link in industrial production such as thermal power, coking and the like, and the industrial coal blending is characterized in that on the premise of meeting the constraint of the quality index of a target product, the coal types with different sources and different physical and chemical properties are proportionally combined so as to realize the balance of production stability and resource utilization efficiency. However, in actual production, the coal quality is obviously affected by the difference of mine points, the piling condition and the change of operation conditions, and the coal blending process has the characteristics of multiple constraint and nonlinearity and time variation. The coal blending scheme in the prior art depends on manual experience or an optimization method based on a mechanism model, the manual experience method has strong subjectivity and poor stability, the mechanism model method is generally based on linear or weak nonlinear assumption, and is difficult to adapt to coal quality fluctuation, parameter drift and complex working condition change, so that an optimization result is easy to deviate from a target in actual operation. Accordingly, we present an industrial coal blending optimization method, system and program product based on multi-agent collaboration, the above information disclosed in the background section is only for enhancing understanding of the background of the present disclosure, and therefore it may include information that does not form the prior art that is known to one of ordinary skill in the art. Disclosure of Invention The invention aims to overcome the defects of the prior art and provide an industrial coal blending optimization method, system and program product based on multi-agent cooperation, which solve the technical problems in the prior art. In order to achieve the above purpose, the present invention provides the following technical solutions: the industrial coal blending optimization method based on multi-agent cooperation comprises the following steps: step S1, acquiring production state data of the current time step, wherein the production state data comprises stock quantity of each coal type, physical and chemical indexes of raw coal, weight coefficients of raw material resources, quality constraint of target products and environmental characteristic data; S2, inputting the production state data into a preset mechanism calculation module, and calculating to obtain a reference proportioning vector meeting the quality constraint of the target product by using a linear programming algorithm and minimizing the comprehensive resource consumption index as an objective function ; S3, inputting the production state data into a pre-trained strategy correction agent, and outputting a correction motion vector for the reference proportioning vectorConfidence factor for current operating conditions; Step S4, according to the reference proportioning vectorCorrecting motion vectorConfidence factorWeighting and fusing to generate final proportioning vectorAnd the final proportioning vector is issued to a coal blending executing mechanism; And S5, acquiring actual output quality data after a hysteresis period, calculating a reward value based on the actual output quality data, and updating the strategy correction agent. The strategy correction agent comprises an input layer, a mask layer and an attention mechanism layer; in step S1 and step S3, when there is a missing value or an abnormal value in the raw coal physicochemical index in the production status data, the mask layer sets the corresponding feature value to zero, and the attention mechanism layer redistributes the weights of the remaining valid data to generate the corrected motion vector 。 In step S4, the final proportioning vector is generatedThe calculation formula of (2) is as follows: ; Wherein Normalize is a normalization function for ensuring that the sum of the ratios is 100%, the confidence factor The strategy corrects the agent output to approach 0 when the fluctuation of the production state data is detected to exceed the preset threshold value or the sensor failsThe system is degenerated to a control mode that dominates the mechanism calculation module. In step S5, the calculation formula of the prize value includes a quality penalty term, a consumption optimization term, and a running stability term:; The consumption optimization term Based on the difference value between the reference proportion consumption index and the final proporti