UCLA

Two pioneer states, California and New York, have set their ambitious targets to get 100% and 70% of their electricity from renewable energy resources by 2045 and 2030, respectively. Aligned with these endeavors, currently 19.2% of the electricity in California is coming from the solar energy where the utilities are expected to add an additional 60% solar energy in the next five years. To achieve this goal, Electrification- the transition from non-electric end-use energy consumers to the electric consumers is a trend in the energy sector, as it facilitates to have access to sustainable and clean energy infrastructure. The fact that 28% of the energy in the USA is consumed by the transportation sector where 92% is provided by the fossil fuel energy motivates Electrication strongly. This revolution is specifically happening in the California state where 50% of the electric vehicle (EV) owners are living.

The increasing penetration of renewable energies, such as solar energy, as well as Electrification in the electrical grids introduce new challenges for system operation and planning. Solar energy, inherently, is intermittent and shows stochastic behavior which makes it a non-dispatchable source of electricity. Therefore, the conventional models to capture its generation profile are no longer applicable. Also according to a new report by National Renewable Energy Laboratory (NREL) [MJL18], EVs are introduced as one of the most influential elements of Electrification. EVs can drastically change the load pattern, thus their integration in the power grids is carefully observed by the electric utilities.

To mitigate the stochastic behavior of solar energy and make it a dispatchable resource, the energy storage can be utilized to capture its intermittency through the coordinated charging and discharging sequences. That is one reason why the California and New York states plan to integrate an additional 1.3 GW and 3.0 GW energy storage in their power grids by 2024 and 2030, respectively. Moreover, EVs are controllable loads that provide the flexibility and opportunity to shift their consumption profile according to the operating conditions of the power grids. Nevertheless, the deployment of energy storage is challenging as the generation profile of the solar energy should be modeled accurately, and an effective and optimal controller should be designed to coordinate the charging and discharging of the energy storage. Also designing an efficient charging load management system for EVs is a difficult task since all the involved entities in the load management decision making, such as the system operator, load aggregators, and the end customers, must be satisfied, and the safe and stable operation of the grid should be guaranteed. Accordingly, the optimal load management is a large scale problem, especially when it should be solved and repeated every several minutes for the whole power grid.

The models proposed in the literature for the solar energy generation, load, and EV charging demand modeling either can not capture their stochasticity accurately or are not computationally efficient. Therefore, the coordination methods for energy storage as well as EV loads are not effective in accommodating renewable energies and Electrification in the power grids. In addition, the energy storage and load coordination methods are not scalable and suffer from a considerable computation burden when the number of energy resource units and controllable loads in the optimal decision-making increases.

In this dissertation, (1) modeling of the solar generation, load demand, and EV charging load, (2) the integration of battery energy storage system (BESS) in the power grids, and (3) the large scale accommodation of EV loads are addressed. For the modeling, a probability model based on kernel density estimator (KDE) is proposed which, comparing to the previous models, provides a low-computation precise stochastic model. For the integration of BESS in the electrical grids, a mobile BESS (MBESS) is prototyped to capture the random behavior of the EV charging profile, reduce the charging demand cost and improve the reliability and resiliency of the charging service. The performance of MBESS is validated through the experiments in the Civic Center parking structure, in the City of Santa Monica, and it is shown that MBESS can effectively shave the peak load of the EV charging demand. To address the lack of scalability in the previous load management methods, the distributed optimization methods are used so that the optimal EV load coordination is solved through an iterative negotiation procedure. The scalability of the proposed methods is coming from the fact that each agent solves its desired problem locally while it exchanges the insensitive limited information with the others. The proposed methods satisfy all the agents and guarantee the power grid operation in the stability and safe region.

The proposed probability model is shown to improve the accuracy of the solar energy profile up to 36.7%, the load demand profile up to 5.9%, and the EV charging parameters, including the arrival time, required charging energy, and the departure time up to 26.6%, 49.3%, and 41.21%, respectively. In addition, the experiments with MBESS verifies that it not only reduces the charging cost but also provides the emergency power to the charging system in the case of failure in the power grid, which is called islanded operation. Moreover, through the numerical simulations using real data, it is validated that the distributed multi-agent based methods for the load coordination can approximately decrease the convergence time and the communication overhead by 94% while the computation burden for the distribution system operator and the load aggregators reduces significantly. Also the load coordination results validate the efficacy of the proposed frameworks in accommodating the large populations of EV loads in the distribution grids by improving the voltage prole from 45% up to 93% and reducing the peak load from 50% up to 66%. The results show that an efficient load management system is a necessity for Electrification integration without any investment on the grid capacity expansion.