Table 4 – Layer characteristics for Scenario A. Stochastic parameters display the mean with standard deviation in parentheses. Parameter Layer 1 Layer 2 Layer 3 Layer 4 Deposit size (tonnes) 50000 50000 50000 50000 P2O5 composition 32 (0.5) 30 (0.5) 26 (0.5) 28 (0.5) MgO composition 0.8 (0.22) 1.2 (0.22) 2.0 (0.22) 1.6 (0.22) Table 5 – Profit ratio relative to the MILP benchmark over 100 14-day simulations for Scenario A. Approach P10 (%) P50 (%) P90 (%) MCTS 95.9 103.8 118.6 ggMCTS 90.7 97.7 107.0 Figure 3. Daily profit comparison over the P50 14-day simulation. Figure 4. Pile inventory over the P50 14-day simulation for the MCTS solver. 4.2 Scenario B: Strict Client Contracts The layer characteristics are the same as in Scenario A (reported in Table 4), except now each layer is considered to be in a separate mine. (This arrangement is equivalent to extracting from all four layers of one mine on each day.) A fixed amount of 500 tonnes is produced by each mine every day. This is to guarantee that there will always be enough feedstock regardless of the circumstances to theoretically ensure client contract fulfillment. Table 6 shows the profit ratio relative to the greedy MILP benchmark for the MCTS and ggMCTS solvers. Both the MCTS and ggMCTS solvers perform better than the benchmark in most simulations, with MCTS achieving a higher profit than greedy in 72% of simulations and ggMCTS achieving a higher profit in 89% of simulations. However, similar to the situation in
RkJQdWJsaXNoZXIy MTM0Mzk2