This thesis researches a novel electricity demand response method and renewable energy management technique. It demonstrated the use of flow batteries and residential hot water heaters to balance wind power deviation from plan.
The electricity supply-demand balancing problem becomes increasingly more difficult. A large portion of complexity to this problem comes from the fact that most renewable energy sources are inherently hard to control and intermittent. The increasing amount of renewable energy generation makes scientists research new supply-demand balancing possibilities to adapt for the changes.
In this research wind power data was used in most cases to represent the supply side. The focus is on the actual generation deviation from plan, i.e. forecasting error. On the other hand, the methods developed in this thesis are not limited to wind power balancing.
Two major approaches were analysed - heating ventilation and air conditioning system control (mainly focused on, but not limited to, residential hot water heaters) and hybrid power system comprising of thermal and hydro power plants together with utility scale flow batteries. These represent the consumption side or the demand response mechanism.
The first approach focused on modelling the behaviour of residential end users. Artificial intelligence and machine learning techniques such as neural networks and Box-Jenkins methodology were used to learn and predict energy usage. Both joint and individual dwelling behaviour was considered. Model predictive control techniques were then used to send the exact real-time price and observe the change in electricity consumption. Also, novel individual hot water heater controllers were modelled with the ability to forecast and look ahead the required energy, while responding to electricity grid imbalance. It proved to be possible to balance the generation and increase system efficiency while maintaining user satisfaction.
For the second approach, the hybrid multi-power plant system was exploited. Three different power sources were modelled, namely thermal power plant, hydro power pant and flow battery. These sources were ranked by the ability to rapidly change the output of electricity. The power that needs to be balanced was then routed to different power units according to their response times. The calculation of the best power dispatch is proposed using a cost function. The aim of this research was to accommodate for the wind power imbalance without sacrificing the health of the power plants (minimising load variations for sensitive units).