Fast depleting fossil fuels and growing awareness for environmental protection have led us to the urgency of a long-term energy planning where reduction of emissions, integration of renewable supply, and energy efficiency improvement represent the main targets of a ‘smarter’ employment of primary resources. Research is needed nowadays to drive a transient phase towards the construction of future ‘smart grids’, where multiple actors will be able to communicate with each other and efficiently adapt their production/consumption with respect to the dynamic evolution of the increasingly complex power network. In this scenario, operational management of small, local electricity networks (microgrids) and their two-way interconnection to the main grid are creating new opportunities and, at the same time, new technological challenges. Advanced control schemes are being investigated to smoothen the integration of distributed generation and to achieve optimal operation at microgrid level, through coordination and dispatching of power generation, flexible loads, and storage elements.
The residential sector is responsible for about 30% of the global energy consumption and has historically played a passive role in the unidirectional centralised power infrastructure. A residential microgrid that utilises controllable prime movers, such as gas engines, to compensate fluctuating demand and output of renewable energy would represent a fundamental step towards a more economic, efficient, and environment friendly energy infrastructure. This MSc thesis project focuses on the design of energy management systems in residential buildings where micro-Combined Heat and Power (CHP) generators are installed. Micro-CHP technology is able to produce electrical energy locally in a controllable way, having at the same time the advantage of efficiently employing by-product heat to satisfy thermal demand of the building where it is located. The purpose of our work is an economic analysis regarding the profitability of investment in distributed energy resources for Dutch households and a subsequent investigation about the benefits that advanced control techniques would lead to microgrid operation on the long run. For this reason, specific case studies are built based on real data of thermal and electric consumption, which have been collected through smart meters in various Dutch houses. Two different versions of the microgrid are considered: a first case only involves micro-CHP and thermal energy storage, whereas a second one is expanded to include solar panels.
Advanced techniques employed for supervisory control of power flows in microgrids generally aim to take into account relevant information about the consequences of choosing specific actions, by considering future predictions of system evolution. Model Predictive Control (MPC) is a well-known, established and widely used control technique that is often considered as a natural approach to adopt in microgrids. Its main strength is the ability to turn a control problem into an optimisation problem; therefore the capability of including operational constraints arises naturally. However, high volatility of small-scale demand and intrinsic stochasticity of renewable energy supply lead to the hard challenge of integrating appropriate forecasting models into the decision-making strategy. When deterministic approaches relying on the certainty equivalence paradigma are applied in residential microgrids, frequent violations of thermal comfort constraints occur due to poor prediction accuracy of the stochastic
processes involved. The possibility to explicitly take into account the uncertainty affecting the controlled system extends the effectiveness of the predictive control strategies, at the cost of increased complexity. Therefore, suitable probabilistic formulation of the forecasting models for stochastic processes and subsequent control strategies in the MPC framework are studied in our work. Different stochastic approaches recently studied in the scientific literature, i.e. scenario based and tree based, are implemented and compared for the defined case studies. Their performance is evaluated in terms of economic savings, primary energy consumption, and violation of thermal comfort constraints for the households.
The results of our work show the profitability of investment in residential microgrids for average Dutch households willing to share the installation of distributed energy resources in multifamily buildings, even in absence of government subsidies. Moreover, the employment of predictive strategies for local generation scheduling results in slightly improved performance with respect to traditional rule-based controllers. The poor prediction accuracy of demand forecasting on small spatial scale still represents the main difficulty to overcome in order to fill the gap with the theoretical potential benefits of ‘optimal’ predictive strategies. However, in the investigated context, the need for a stochastic framework is motivated and highlighted
with respect to the usage of deterministic tools due to the large variance of uncertainty in system dynamics.