Demand response integrated market operation with renewable energy applying AI technique

Abstract

With the objective towards building a sustainable future power grid that is more efficient and economical, various efforts on the development of smart grid technologies have been made. On one hand, the goal to reduce carbon emission has pushed forward the deployment of renewable generation, which brings clean and low operation cost energy to grid. However, the fluctuations caused by the inherent intermittency of renewable energy also brings new challenges on the system balancing. On the other hand, rapid development of demand response and energy storage technologies bring new options to accommodate more renewable energy resources in the grid. At the same time, artificial intelligence based techniques emerges and they are suitable to be applied for acquiring deep and precise knowledge of renewable energy. In this thesis, various work has been conducted on the abovementioned topics to investigate their applications in the power system. Firstly, literatures associated with demand response integration in the electricity market framework are reviewed. A classification of these works are categorized into general market integration and modelling, integration in the wholesale market and retail market and finally, market operation of demand response with renewable energy. Secondly, a day-ahead market clearing model considering price-responsive demand bidding is proposed and simulated on the power system to study the impact of integrating demand response on the economical and operational aspect of the system. Thirdly, a new framework that integrates the demand shifting with the generation scheduling is proposed with the bidding of demand shifting formulated, through which the energy consumption of customers can be maintained. Results demonstrates that with the load reduction can take place during the peak hours, and can be recovered during the valley periods, which reduces the peak-valley difference. Price spikes during the peak hour can also be mitigated. Next, the impact of solar power integration to power systems is studied. On this topic, firstly an artificial neural network based approach is proposed to estimate solar power when ground measured data become unavailable. With the use of high resolution data of solar power output, the proposed neural network model can achieve accurate estimation. Comparison of the estimation result between different locations shows that the proposed model can keep the estimation error within a reasonable range. However, it is also evident from the analysis that performance of the neural network model can still be impacted by the climate condition of the specific location. Secondly, an optimal scheduling model is proposed that considers the scheduling of battery and thermal generation under solar power penetration. Actual solar power output data of the summer season is applied to the proposed model for simulation. Furthermore, to approximate the stochastic nature of solar power, a method is proposed to randomly generate solar power profiles and Monte-Carlo simulations are carried out on the proposed model. It is demonstrated that integrating solar power can help to reduce system operation cost, balance the peak load during the noon, and saves battery discharging energy for balancing evening peak load.