Benchmarks for the coal processing and blending problem

In this paper we present a challenging problem that many decision makers in coal mining industry face. The coal processing and blending problem (CPBP) builds upon the traditional blending problem known in operations research (OR) by including decision variables around coal processing, novel constraints as well as arbitrary user-defined profit functions which express price bonuses and penalties. The added complexity turns the traditional blending problem into a challenging black-box optimisation problem. We give an informal and mathematical description of this problem and present nine real-world problem instances as benchmark. Finally, we provide preliminary results for solving the problem by using a Genetic Algorithm (GA) and compare the results with those from a commercial Linear Programming (LP) solver. The results show that the GA significantly outperforms the LP solver in many problem instances while being marginally worse in others.