Enterprise-wide Optimization for Industrial Demand Side Management

In the light of increasing volatility in electricity price and availability, demand side management (DSM), which refers to the active management of electricity consumption, has become crucial for the economic performance of power-intensive industries. Due to its time-sensitive nature, DSM is a challenge for industrial plants; however, it can also be an opportunity if sufficient process flexibility is available, which can be leveraged to take advantage of financial incentives provided by various electricity markets. The goal of this work is to develop systematic decision-making tools for industrial DSM at the enterprise level. We identify and address four major challenges: (1) accurate modeling of operational flexibility, (2) integration of production and energy management, (3) decision-making across multiple time scales, and (4) optimization under uncertainty. We develop a discrete-time mixed-integer linear programing (MILP) model that integrates detailed production scheduling and electricity procurement from various sources. The proposed model is proven to be computationally efficient, which can be attributed to the mode-based formulation and the incorporation of a special type of process surrogate model, referred to as Convex Region Surrogate (CRS). In a CRS model, the feasible region is given by the union of convex regions in the form of polytopes, and for each region, the corresponding cost function is approximated by a linear function. For the construction of CRS models, we propose a data-driven algorithm that can be applied to data obtained from the real process or from simulations. Using the proposed integrated scheduling model as a basis, we optimize decisions regarding load shifting, inventory management, electricity procurement, energy storage, provision of interruptible load, etc. The framework is further extended to the supply chain level by also integrating distribution decisions, resulting in a multiscale production routing problem (MPRP). In order to solve large instances of the MPRP, we propose an iterative MILP-based heuristic algorithm that obtains high-quality solutions in reasonable computation times. A strong focus of this work lies on the treatment of uncertainty, which occurs in many forms in industrial DSM problems. We consider uncertainty in product demand, electricity price, and grid contingency events. These uncertainties all have different characteristics and affect the process in different ways; hence, we consider them using different modeling strategies, namely stochastic programming and robust optimization. We emphasize the consideration of risk, which is incorporated into the stochastic programming model using the conditional value-at-risk. In the proposed robust optimization models, we reduce the level of conservatism by implementing appropriate budget uncertainty sets and incorporating recourse decisions in the form of linear decision rules. Computational challenges are addressed by applying reformulations and decomposition strategies. Finally, we examine for linear systems the relationship between flexibility analysis and robust optimization. The effectiveness of the proposed methodologies is demonstrated in several case studies, many of which consider industrial test cases with real-world data provided by Praxair.