Overview
Research introduces a data-driven dimension-reduction technique specifically designed for industrial load modeling. This approach addresses challenges associated with integrating complex industrial load models, characterized by intricate mixed-integer constraints, into economic dispatch or market clearing processes. Existing analytical dimension-reduction methods are cited as ineffective for these complex models.
Research Context
Industrial load models inherently contain intricate mixed-integer constraints. These constraints present significant hurdles for direct integration into critical energy management operations, such as economic dispatch and market clearing. The complexity of these models renders current analytical dimension-reduction techniques unsuitable for effective simplification, necessitating a new methodology.
Approach
The proposed methodology is a data-driven dimension-reduction approach. Its core mechanism involves training a dimension-reduced model. This training process specifically utilizes optimal energy usage data derived from industrial loads. The objective of this training is to enable the dimension-reduced model to best fit the original constraints of the industrial load. The implementation of this approach was carried out using an adjustable load fleet model.
Findings
The data-driven dimension-reduction approach, as implemented by the adjustable load fleet model, demonstrated superior performance. This comparative assessment was made against analytical methods. The evaluation encompassed three distinct industrial load datasets, across which the proposed method consistently outperformed the analytical alternatives.
Why This Matters
The research addresses a computational and integration challenge in industrial energy management. By offering an effective dimension-reduction method, it facilitates the integration of complex industrial load models into economic dispatch and market clearing processes. This capability is significant given the inherent difficulties posed by the models' intricate mixed-integer constraints.