Genetic algorithms — A novel technique to optimize coal preparation plants

Abstract

A coal preparation plant typically operates with multiple cleaning circuits based on the particle size distribution of run-of-mine coal. Clean coal product from a plant commonly has to satisfy multiple product quality constraints, including product ash, product sulfur, heating value, moisture content, etc. Numerous studies in the past illustrate that the optimal yield of the plant can be obtained by operating each circuit to produce the same incremental product quality. This equal incremental product quality approach optimizes the plant yield considering only one product quality at a time. Thus, when required to simultaneously satisfy multiple product quality constraints, the process not only becomes increasingly complex and cumbersome, but also may lead to erroneous conclusions in many cases. A novel plant optimization technique was developed using genetic algorithms (GA) to maximize the overall revenue generated by a coal preparation plant by searching the best possible combination of overall yield and multiple product quality constraints. This approach is based on an evolutionary algorithm that maximizes the overall plant revenue based on a single objective function, which was developed by incorporating clean coal yield, targeted product ash content, product heating value, and product SO 2 emission potential. Comparative results discussed in this publication indicate the suitability of the proposed GA-based plant optimization approach.