(715b) A Rolling Horizon Scheduling Algorithm Considering Electricity Load Tracking and Future Load Preduction
Due to the strong time-dependence of these electricity-related concerns, effective scheduling is essential in DSM, especially when a complex manufacturing system is involved. Several publications have looked at DSM for scheduling including Hadera et al. (2015) who looked at the scheduling of steel manufacturing based on a multi-contract load commitment problem using a precedence-based model. Castro et al. (2013) also studied the load commitment problem using a discrete-time Resource-Task Network (RTN) formulation. The downside of studying only a 24-hour problem stems from two major factors. The first is that they unrealistically consider the intraday and day-ahead markets by either assuming both to be active at the same time or by ignoring one of the markets. The second is that the committed electricity load in such formulations drops off towards the end of the day as the considered jobs are completed. This represents an unrealistic demand curve as in actuality production is performed on a continuous basis.
We address this problem by considering a two-day problem, combining the current-day problem of load commitment (considering the intraday market) with the future problem of load-prediction (considering the day-ahead market). The downside to such an approach is that it results in very large models that are difficult to solve. In this work, a novel rolling-horizon algorithm is proposed based on the discrete-time RTN formulation. In this algorithm, a non-uniform grid is considered. In the detailed portion of the time-grid, a fine time discretization is used to capture the details of the scheduling model. In the remaining model a coarser grid is used in order to capture the necessary details of the scheduling and electricity models, while reducing the computational complexity of the overall problem. Results show that this algorithm is able to greatly reduce the computation time of such scheduling problems, while still capturing the key aspects of the formulation. In addition, the formulation naturally lends itself to an online scheduling environment.
Castro, P. M., Sun, L., & Harjunkoski, I. (2013). Resource-Task Network Formulations for Industrial Demand Side Management of a Steel Plant. Industrial and Engineering Chemistry Research(52), 13046-16058.
Hadera, H., Harjunkoski, I., Sand, G., Grossmann, I. E., & Engell, S. (2015). Optimization of steel production scheduling with complex time-sensitive electricity cost. Computers and Chemical Engineering, 76, 117-136.