Multimodal Optimization Forecasting Model Based on Intelligent Fuzzy Interval Reconstruction

Breaking through the limitations of current research, integrating data reconstruction into the interval forecasting methods for electric load represents a significant leap forward. This study proposes a joint electric load interval forecasting system that combines data recovery and fuzzy prediction, effectively overcoming the data quality issues in current electric load interval forecasting. Unlike mainstream methods, this system employs an interval construction approach at the input stage, which effectively addresses the issues of data gaps and noise present in real load data. An innovative intelligent variational fuzzy information granulation technique is introduced, which significantly improves the quality of the input interval data. Moreover, the combination of models not only enhances the system’s interpretability but also provides more comprehensive feature extraction capabilities, thereby improving forecasting performance. Building on this foundation, an improved multi-objective artificial bee colony algorithm has been developed and compared with several mainstream algorithms, resulting in a significant enhancement in the forecasting accuracy and stability of the combined model. Testing and comparative results on the electric load dataset from Queensland, Australia, demonstrate that the system achieves top-tier forecasting performance in both point and interval forecasting. The QSD data is derived from historical data recorded in the Australian electricity market.