Predicting Gold Mine Surface Cooling Systems Energy Consumption Kabelo Donald Lomko Department of Electrical and Electronic Engineering Science University of Johannesburg Johannesburg, South Africa kbleng@gmail.com Khmaies Ouahada Department of Electrical and Electronic Engineering Science University of Johannesburg Johannesburg, South Africa kouahada@uj.ac.za l Hailing Zhu Department of Electrical and Electronic Engineering Science University of Johannesburg Johannesburg, South Africa kasha0306@gmail.com Abstract— An artificial neural network (ANN) was utilised to predict the energy consumption of the fridge plants of a mine’s surface cooling system. Predictive accuracy of 96, 89% was achieved. The maximum and minimum predicted energy consumption on the fridge plants was found to be 17 MW, 12 MW, respectively, which is fairly close to the real-time energy consumption of the machines. This model was implemented under automated load shift conditions to reinforce a hypothesis of this research, which is that demand side management (DSM) initiatives can be augmented by accurate predictive models. Accurate predictive models will ensure effective cooling system planning, sufficient machine maintenance, effective cooling system operation, optimal mine energy allocation, and energy management on the mine cooling systems, particularly its fridge plants/chillers. As the mining industry traverses towards automation of its DSM initiatives, intelligent systems have to be implemented for full automation to be achieved, and this research sought to make a contribution to that aspect. An ANN was found to outclass multiple linear regression, thus, ANNs were found to be better models for integration into DSM projects. Finally, the number of fridge plants that need to operate were determined based on the predicted energy consumption. The number of fridge plants that operated during Eskom’s morning and evening peak periods was 4 and 1, respectively. This was found to be better than the traditional mode of operation whereby the entire number of fridge plants (6) operate all day. Keywords— mine cooling systems; demand side management; load shift; artificial neural network; fridge plants/chillers I. INTRODUCTION At the heart of economic development lies the need for a sustainable, reliable, and affordable energy supply. Due to the energy-intensive nature of the industrial and mining sectors of the economy, a strain is placed on the South African power grid. In 2013, it was estimated that industrial and mining sectors of the economy consume 37% of the total energy produced in the world. A strain on the power grid threatens the sustainability and reliability of electricity supply [1]. Capital is an important part of increasing the capacity necessary to generate electrical energy to meet the demand. This need induced the power utility, Eskom, to increase electricity tariffs. Research indicates that an increase in electricity costs has been higher than that of the gold prices [1], [2]. The increase in electricity prices is experienced mostly by the more energy-intensive industries. The mining sector, in particular, has been experiencing the effects of escalating prices [1]. This indicates that the gold mining industry will, at some stage, not be economically viable [3]. Thus, efficiency in the operation of a mine is pivotal because of the exorbitant electricity costs. This can be achieved by the modernisation of mining, DSM initiatives, and accurate energy consumption prediction. The mining industry is gradually migrating from the inefficient, energy-wasting mode of operation to a more energy-efficient and energy cost-saving mode of operation. At this stage strides have been made in achieving automation of mine cooling systems; however, the automation is not yet complete. This research explores the inherent gap and suggests that full automation will require the integration of current DSM automation projects with intelligent systems that can predict consumption. These intelligent systems will ensure that future consumption on the mine fridge plants is known and as a consequence, running schedules will be pre- determined. This will ensure that effective cooling system planning, sufficient machine maintenance, effective cooling system operation, optimal mine energy allocation, and energy management on the mine cooling systems is achieved automatically. II. ENERGY CONSERVATION ON MINE SURFACE COOLING SYSTEMS, DSM INITIATIVES, AND FORECASTING METHODS A. Energy conservation on mine surface cooling systems The deepest mines in the world are found in South Africa. AngloGold Ashanti’s Mponeng gold mine is the deepest in the world. At the culmination of the year 2012, Mponeng mine was operating at a depth of 3.9 km. After 2012, further digging has occurred that has caused the mine to become the only mine in the world to break into the 4 km mark [2]. At such mining depths, cooling is important because Virgin Rock Temperatures (VRTs) escalate to temperatures in the region of 67 °C [3]. At such temperatures, mining activity becomes hazardous, thus, sufficient water and air cooling are pivotal in deep-level mines [4]. Deep-level mining requires sufficient cooling as discussed, and the associated energy requirements increase with depth [4]. The energy consumption per unit of gold produced is directly proportional to the mining depth. A large part of the mining energy goes into its cooling systems because of the depth, size and increased temperatures [5].