A clustering-based energy consumption evaluation method for process industries with multiple energy consumption patterns

The production systems in process industries are confirmed to be tremendously energy-consuming, and the trust in promoting their energy efficiency has become a concern, with its precondition being to evaluate the real-time energy consumption. A widespread evaluation method is to develop a global model that employs energy audit techniques, whereas they are always carried out with few appreciations of multiple energy consumption patterns, and the utilization of energy consumption auxiliary information. To address the challenge, a two-stage clustering-based-energy consumption evaluation method is proposed for process industries in this study. Specifically, a novel structure of the fuzzy clustering method is designed with a mixture of unsupervised and semi-supervised learning stages that leverages the weighted information to independently address energy consumption patterns. Then energy consumption predictions are estimated for potential energy-optimized control. The key performance indicators of energy consumption are calculated for each pattern, and the final evaluation grade will be achieved through the fuzzy synthetic evaluation method. According to the experiment results, the proposed method delivers better evaluation results against baselines with more accurate clustering; it may provide a new thought for energy consumption evaluation and is confirmed to enable practitioners to acquire the potential benefits in engineering.