Simultaneous Stochastic Optimization of Production Scheduling at Twin Creeks Mining Complex, Nevada

"Twin Creeks is a gold mining complex that is part of Newmont’s Nevada operations. The mining complex is composed of two openpits, Mega and Vista, external ore sources and a set of existing stockpiles, all of which provide ore for an autoclave, an oxide mill and a heap leach. Mega pit provides sulfide ore for the autoclave, whereas Vista pit provides oxide ore for the oxide mill and the heap leach. Both pits operate the same mining equipment and their extraction sequence must, therefore, be optimized simultaneously. Stringent blending requirements are associated with the operation of the autoclave for the sulfide ore. Strategic blending optimization on a large scale has brought significant value to the operation by increasing synergies. This paper presents the implementation of a stochastic optimization framework for the long-term production planning at Twin Creeks that simultaneously optimizes mining, stockpiling, blending and processing decision variables. The uncertainty and variability associated with the different sources of material are incorporated in the optimization model by means of stochastic simulations. The stochastic solution generated shows substantial benefits by increasing forecasted recoverable gold, leading to increased cash flows, while reducing the risk of not achieving forecasts by increasing the probabilities of meeting production and blending targets.IntroductionMine operations consist of multiple components such as mineral deposits, stockpiles, external sources, mills and autoclaves. These components represent a mineral value chain, also known in the literature as a mining complex (Montiel and Dimitrakopoulos 2015; Goodfellow and Dimitrakopoulos 2016). The material extracted from the mines flows through the different components and is transformed into sellable products. This flow of material in a mineral value chain involves complex, nonlinear transformations that render the use of industry-standard optimizers difficult. Furthermore, the optimization of a mining complex can be formulated as a mixed-integer, nonlinear optimization program with millions of integer variables representing mining, blending, stockpiling and processing decisions, prohibiting the use of conventional nonlinear optimization approaches. Previous methods that consider the joint optimization of multiple components of the mineral value chain are available (Hoerger, Seymour and Hoffman, 1999; Chanda, 2005; Stone et al., 2005; Wooller, 2005; Whittle, 2007, 2010; Bodon et al., 2010). Although these methods add value by increasing the synergy between the different components of the mining complex, they rely on simplifying assumptions that constrain their efficacy, while ignoring a most critical aspect. The latter aspect includes the geological uncertainty and actual variability of the materials in the mineral deposits mined and are supplied to the related mineral value chain, similarly to conventional life-of-mine planning methods used for individual mines (Dimitrakopoulos, Farrelly and Godoy, 2002; Godoy and Dimitrakopoulos, 2011; Spleit and Dimitrakopoulos, 2017). Additional simplifying assumptions may include: the consideration of a limited number of components, the combination of different methods, the aggregation of variables, fixed production schedules the definition of policies prior to optimization, such as cutoff- grade definition or destination of mining blocks, the linearization of nonlinear transformation functions and/or the disregard for the uncertainty of several critical parameters (Whittle, 2007). The simultaneous optimization of multiple components of a mining complex was introduced by Newmont. Hoerger, Seymour and Hoffman (1999) implemented a mixed-integer programming formulation to optimize Newmont’s Nevada operations. The model is based on the work of Urbaez and Dagdelen (1999) and simultaneously optimizes the extraction of materials from multiple deposits and the delivery of ore to multiple plants, as well as as"