Improving advanced M&V by incorporating real time unplanned event identification on deep-level mines

Abstract

Baselines and budgets play a crucial role in the monitoring of energy-saving initiatives by measurement and verification (M&V) specialists. While budgets are primarily designed for cost planning, baselines serve as historical benchmarks to discern improvements over time. The advent of advanced metering infrastructure has revolutionized the assessment of progress in real-time, giving rise to advanced M&V processes. This evolution includes the integration of data forecasting techniques, such as machine learning and regression models. However, existing models often focus solely on comparing results to a constant baseline, neglecting the impact of operational changes like unplanned events, which are scarcely addressed in the literature. This study aims to develop a predictive energy consumption model that accommodates operational changes and provides practical guidelines for handling unplanned events in deep-level mines. The approach employs the Moving Average (MA) equation to predict monthly energy consumption based on historical data. This prediction serves as a baseline, compared with a second prediction incorporating actual energy consumption data for each day. The resulting error is monitored to detect changes in operations, utilizing the guidelines as a reference. The developed model and guidelines were applied to two deep-level mining systems lacking advanced M&V implementation. The model successfully identified unplanned events in both cases, significantly reducing the time spent compared to traditional M&V processes. In case study A, where a water pipe issue went unnoticed for 97 days, the model could have expedited Improving advanced M&V by incorporating real time unplanned event identification on deep-level mines the identification and correction process by approximately 70 days, representing a 72% improvement. In case study B, where compressed air control valve failures were masked by another initiative, leading to a delayed investigation three months later, the prediction model could have saved 82% of the timeline. The combined potential savings on both case studies amount to R1.6 million, underscoring the critical need for real-time unplanned event identification in deep-level mines through the application of advanced M&V principles.