Planning in smart grids

Abstract

Thee electricity supply chain is changing, due to increasing awareness for sustainability and an improved energy efficiency. The traditional infrastructure where demand is supplied by centralized generation is subject to a transition towards a Smart Grid. In this Smart Grid, sustainable generation from renewable sources is accompanied by controllable distributed generation, distributed storage and demand side load management for intelligent electricity consumption.The transmission and distribution grid have to deal with increasing fluctuations in demand and supply. Since realtime balance between demand and supply is crucial in the electricity network, this increasing variability is undesirable. Monitoring and controlling/managing this infrastructure increasingly depends on the ability to control distributed appliances for generation, consumption and storage. In the development of controlmethodologies,mathematical support, which consists of predicting demand, solving planning problems and controlling the Smart Grid in realtime, is of importance. In this thesis we study planning problems which are related to the Unit Commitment Problem: for a set of generators it has to be decided when and how much electricity to produce to match a certain demand over a time horizon. The planning problems that we formulate are part of a control methodology for Smart Grids, called TRIANA, that is developed at the University of Twente. In a first part, we introduce a planning problem (the microCHP planning problem), that considers a set of distributed electricity generators, combined into a Virtual Power Plant. A Virtual Power Plant uses many small electricity generating appliances to create one large, virtual and controllable power plant. In our setting, these distributed generators are microCHP appliances, generating Combined Heat and Power on a domestic scale. Combined with the use of a heat buffer, operational flexibility in supplying the local heat demand is created, which can be used in the planning process, to decide when to generate electricity (which is coupled to the generation of heat). The power output of a microCHP is completely determined by the decision to generate or not. The microCHP planning problem combines operational dependencies in sequential discrete time intervals with dependencies between different generators in a single time interval, and searches for a combined electricity output that matches a desired form. To illustrate the complexity of this problem, we prove that the microCHP planning problem is NP-complete in the strong sense. We model the microCHP planning problem by an Integer Linear Programming formulation and a basic dynamic programming formulation. When we use these formulations to solve small problem instances, the computational times show that practical instance sizes cannot be solved to optimality. This, in combination with the complexity result, shows the need for developing heuristic solution approaches. Based on the dynamic programming formulation a local search method is given that uses dynamic programs for single microCHP appliances, and searches the state space of operational patterns for these individual appliances. Also, approximate dynamic programming is proposed as a solution to deal with the exponential state space. Finally, a column generation-like technique is introduced, that divides the problem in different subproblems for finding operational patterns for individual microCHPs and for combining individual patterns to solve the original problem. This technique shows the most promising results to solve a scalable Virtual Power Plant. To apply the microCHP planning problem in a realistic setting, the planned total output of the Virtual Power Plant is offered to an electricity market and controlled in realtime. For a day ahead electricity market, we propose stepwise bid functions, which the operator of a Virtual Power Plant can use in two different auction mechanisms. Based on the probability distribution of the market clearing price, we give lower bounds on the expected profit that a Virtual Power Plant can make. To control in realtime the operation of the Virtual Power Plant in the TRIANA approach, the planning is based on a heat demand prediction. It has been shown that deviations from this prediction can be ‘absorbed’ in realtime. In addition to that, we discuss the relation between operational freedom and reserve capacity in heat buffers, to be able to compensate for demand uncertainty. As a second planning problem, we integrate the microCHP planning problem with distributed storage and demand side load management, in the classical framework of the Unit Commitment Problem. In this general energy planning problem we give a mathematical description of the main controllable appliances in the Smart Grid. The column generation technique is generalized to solve the general energy planning problem, using the real-world electricity infrastructure as building blocks in a hierarchical structure. Case studies show the practical applicability of the developed method towards an implementation in a real-world setting.