Improving allocation models in smart grid planning frameworks using novel optimization schemes and planning mechanisms

Abstract

The inception of smart grids has increased the popularity of integrating multiple Distributed Energy Resources (DERs) and Flexible Alternating Current Transmission System (FACTS) units. This interest is due to promoting sustainability through boycotting fossil fuels to using Renewable Energy Sources (RES). These units also offer an economic advantage since DERs are installed close to service load areas, having no need to transfer large pools of energy over long transmission lines. The interest in Electric Vehicles (EVs) has also peaked, given their capability to mitigate carbon emissions and improve grid performance. However, these units require careful integration since they can cause adverse effects such as voltage instability and extreme power loss. Therefore, smart grid planning frameworks are developed to find optimal allocation schemes for DER and FACTS units, including EV charging stations while considering their technical, economic, and environmental benefits. These frameworks are equipped with uncertainty models, constraints, and power flow algorithms, which increases their time complexity burden. Hence, an efficient optimization algorithm is necessary.

This research makes unique contributions, toward hybridizing metaheuristic algorithms for the optimal allocation of multiple DER and FACTS units, including EV charging stations in planning frameworks. In contrast to the current application of hybrid metaheuristic algorithms, which uses high computational time, this thesis presents the decomposition method for the allocation problem and assigns each algorithm to a problem. As a result, the optimization scheme becomes more computationally efficient.

EV charging can also potentially cause grid instability. To reduce the effect of this drawback, previous studies have worked on the optimal placement of EV charging stations in distribution networks. However, these studies have not considered the random behaviour of EV charging. An oversight of random EV charging can cause unprecedented peak times, negatively affecting grid stability. Hence, there is a need to model the distribution of EV charging. To handle this task, this research developed an EV charging coordination model. This model considers a normal distribution of EV charging and uses a reinforcement learning technique to find optimal locations of EV charging stations.

Necessary for dealing with bias and representational issues in multiobjective optimization (MOO), which are ongoing issues in planning frameworks, a game-based MOO is developed to minimize bias between smart grid players (or decision-makers) — EV aggregators and distribution network operators in a centralized EV charging scheme. Furthermore, a novel category-based MOO scheme is developed to tackle the difficulty of representing many objectives. The results show a minimal bias among similar to dissimilar objective functions, particularly technical objectives.

The final part of the research proposes an adaptive-dynamic planning mechanism and is validated on varying distribution networks, including 15-, 33-, 69-, and 118-bus distribution test networks. It is shown that the proposed mechanism yields a more extensive solution space, making it possible to find better solutions than in conventional planning mechanisms.