Applications of tabu search parallel metaheuristic for stochastic long-term production scheduling

In open pit mine planning, the mine deposit is discretized into mining blocks, where the size of these mining blocks is defined by the mine’s extracting equipment capacity and selectivity. Mining blocks are removed from the ground at different periods and sent to various destinations to be processed, stockpiled or dumped. Long-term production scheduling with multiple destinations is used in the mining industry, to provide guidelines for this extraction process, deciding the mining period and the destination policy that should apply for each mining block. The destination policy aims to optimize where to send the extracted material, in order to maximize the discounted cash flow according to the system capacity. Stochastic long-term production scheduling with multiple destinations includes the uncertainty associated with the grade’s material in the optimization process, by maximizing the net present value, while reducing the risk of not meeting the different production targets at each destination. For deposits represented by a large number of mining blocks, the optimization leads to very complex and large mathematical programs, this cannot be solved to optimality using exact methods such as Branch and Bound. In this thesis, stochastic integer programming formulations are used to integrate the uncertainty directly into the optimization of the long-term production scheduling problem, and Parallel Tabu Search metaheuristics are presented as an approach to provide nearby optimal solution, in a reasonable amount of time. Two different approaches are presented here for the destination policy during the optimization process, based on the economic value of each block. The first approach uses a fixed destination policy, which sends each block to its more profitable destination before the optimization, whereas the second one considers optimizing the policy simultaneously within the optimization process of a life-of-mine schedule.The first part of this thesis, Chapter 3, presents three different implementations of parallel Tabu Search metaheuristics to solve a previously existing stochastic integer program, designed to provide optimal solution for the life-of-mine production schedule with multiples destinations, under geological uncertainty and under a fix destination policy. The first two methods allow a more extensive search of the solution space, the first using several independent Tabu Searches, whereas the second allows communication between the different Tabu Searches to broadcast information. The third method aims to provide a more intensive search by exploring different local area simultaneously, starting from a single solution. An application to a deposit of about 70,000 mining blocks is shown to assess the ability of all methods to generate a schedule with minimized deviations in practical amount of time.In the second part, Chapter 4, a stochastic integer program that jointly optimizes the destination and the year of extraction for each mining block is presented. A parallel multi-neighbourhood Tabu Search implementation is used to approximate the optimal solution of this formulation. The approach considers optimizing simultaneously both the destination and the period of extraction of each mining block by defining different types of neighbour solutions to explore. The computational complexity added by considering simultaneously extraction and destination variables is reduced by the use of a load balancing strategy to distribute the work equally among the different processors. An application at a deposit of about 100,000 mining blocks is made to show the ability of the method to generate a schedule where the production targets are met and the NPV is maximized in a practical amount of time.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.