Computationally efficient and scalable optimization approaches for integrated energy management systems

• Main Author: Dan, Mainak
• Other Authors: Arvind Easwaran
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Electrical and electronic engineering::Control and instrumentation::Control engineering; Engineering::Electrical and electronic engineering::Electric power
• Online Access: https://hdl.handle.net/10356/159808
• Institution: Nanyang Technological University

Sustainable and cost-efficient energy management has become imperative to deal with the proliferation of energy consumption, the rising prices of fossil fuels, and atmospheric degradation. In this context, an energy management system (EMS), which utilizes energy and manages energy conversion efficiently, and warrants integration of multiple energy vectors and development of demand response program is needed. In this thesis, I have considered a unified and holistic energy system that captures the coupling and interaction among generating units and storage components, and flexible load models, generally found in urban buildings. However, to reduce the complexity for development of EMS and to understand the system behavior, a systematic approach has been considered by dividing the problem into two major parts - (i) operational optimization of the generating units in the presence of fluctuating demand and dynamic market price; and (ii) consumers' behavior and interaction for demand response in the presence of varying prices and adjustable generation. This thesis has proposed three novel algorithms to solve the aforementioned problems while considering scalability and the computational expense. The first two algorithms, named as scenario-based branch and bound (SB3) and scalable two-level (Sc2L) target the operational optimization problem of the generating units. On the other hand, the consumers' response and demand-side management problem have been handled by proposing a cooperative game theory-based distributed algorithm. The energy system operational problem is found to be a mixed-integer nonlinear program (MINLP), a non-convex NP-hard problem with intermittent renewable generation and fluctuating demand. The investigation proposes a computationally inexpensive scenario-based control algorithm (a.k.a SB3 algorithm) on a moving horizon framework to address the limitation of existing meta-heuristic and mathematical optimization algorithms. SB3 partitions the complex non-convex MINLP problem into convex sub-problems (a.k.a scenarios) by relaxing the nonlinear and non-convex constraints. The solutions of these set of problems are then used to warm-start a hybrid meta-heuristic algorithm (named as CE-mGA) that exhaustively searches various scenarios in the presence of non-linear and non-convex constraints. The set of feasible scenarios is decided by the branch and bound technique considering the binary constraints present in the operational model formulation. However, this approach, in spite of its meritorious performance, possesses scalability issues for real-time implementation as it warrants computational resources for exhaustive search. To address such limitation, in the next approach, a novel two-level framework is devised. In the lower level, the binary constraints are relaxed and the nonlinear and non-convex operational constraints are convexified in order to obtain a quadratic programming problem. In the upper level an iterative approach is adopted, where a sequence of linear programming problems is solved by reimposing integral conditions of the binary variables and by designing a linear surrogate for the objective function. Both the proposed algorithms are assessed on a test-bed at a commercial building equipped with multi-energy components. The simulation results demonstrate the bene ts, effectiveness, and limitations of both the proposed approaches. Although the generating units' operation is handled by a centralized optimization approach, the consumers' behaviour and demand-side management are solved by resorting to a game theory-based distributed optimization approach. Noncooperative game theory approaches are often common strategies adapted to understand the interaction among the self-interested consumers under different pricing methods, the overall system welfare is often compromised. In this thesis, I have investigated a cooperative approach among such consumers under an abstract pricing mechanism governed by the inverse supply function of the generation facility. A distributed algorithm that requires a minimal information flow between the consumers and the market coordinator is proposed in this investigation. The problem is formulated using the Nash bargaining approach wherein, the consumers bid their consumption pro le based on the market price. In pursuit, I have proved that there exists at least one cooperative solution which provides a better (Pareto efficient) utility outcome for all the participating consumers than non-cooperative Nash equilibrium. Simulation results of the proposed approach for solving the cooperative the problem have shown significant improvement in terms of consumers' utilities over Nash equilibrium for the numerical example considered in this investigation.