Model predictive control – a digital transformation initiative at Vale Long Harbour

Abstract

Vale Long Harbour produces nickel, copper, and cobalt from a hydrometallurgical process. With the plant approaching nameplate capacity, several digital transformation initiatives are underway including model predictive control (MPC). 25 potential applications were identified; estimated returns on investment for each application gave a ranking used to phase implementation. Currently, applications have been implemented on two autoclaves, two boilers, and a thickener. The autoclave applications control temperatures, volume flows, stream qualities, by manipulating feed, recycled leach solution, dilution, and chloride flows. The qualities are inferred using process data and machine learning-based predictive analytics, updated using assays on a four-hourly basis. Dynamic models were derived from manual steps, updated from automated tests. The boiler application controls excess oxygen in the flue gas. The aim is to save fuel and operate in an environmentally friendly manner. Emphasis has been on the management of change and proper training. Preliminary benefits from the applications are extremely good. A common criticism of advanced applications is that benefits erode over time; Vale recognizes these challenges involved and will implement a sustained benefit approach.