Christos Pergamalis

The goal of the study presented in this article is to investigate all current issues related to the proper deployment of load forecasting (LF) techniques in smart grids (SGs). The latter concept has recently emerged as a potential solution to the global energy problem as well as to the ever-increasing and diverse consumer demands. To this end, more flexible dispersed production units are involved, mainly based on renewable energy sources (RESs). Another key novelty of SGs is their ability to gather information directly from consumers and production units in real time, thus facilitating optimum network planning and recovery as well as minimization of outage probability. Hence, it is important to use appropriate advanced infrastructure, which, in combination with modern telecommunication networks, will enable the full exploitation of SGs. In this context, to make the electricity system more efficient, avoid voltage and frequency imbalance issues and implement optimal production and consumption planning, LF is a vital process and plays a key role in the management of future electricity systems. Therefore, recent state-of-the art approaches in LF methods are also presented and discussed. In the same context, current limitations and proposals for future work are identified as well.

The increasing adoption of electric vehicles in residential buildings creates challenges for charging infrastructure management, particularly in pricing services to balance revenue, user satisfaction, and grid stability. Traditional pricing methods, such as fixed rates and time-of-use tariffs, cannot adapt to the dynamic nature of charging demand. We propose a reinforcement learning framework for dynamic pricing of residential EV charging stations. The framework formulates the pricing problem as a Markov decision process and employs proximal policy optimization to learn a pricing policy based on real-time conditions. The state representation includes ten features covering temporal indicators, charging loads, grid status, traffic, and weather. A multi-objective reward function balances revenue, station utilization, grid stability, and user satisfaction. The system is trained on 6878 charging sessions from a residential complex in Trondheim, Norway. Compared with fixed pricing and time-of-use pricing, the proposed method achieves an overall score of 0.569, representing improvements of 32.9% and 48.9%, respectively. Sensitivity analysis confirms that the model remains robust across different demand response assumptions. The main contributions include a custom reinforcement learning environment for residential EV charging and empirical evidence that learned policies outperform traditional pricing approaches.

The growing use of electric vehicles in the residential building sector presents new challenges in the management of the charging infrastructure, especially in deciding how to best price the use of it to balance operator revenue and user satisfaction with grid stability. Traditional pricing methods like fixed pricing rates and time-of-use tariffs cannot accommodate the dynamic nature of charging demand, which fluctuates depending on temporal patterns and weather conditions as well as user behavior. This limitation means that the resources are not used optimally and the revenue opportunities are lost during periods of high demand. To overcome this issue, we propose a reinforcement learning framework for dynamic pricing for residential electric vehicle charging stations. The framework models the pricing problem as a Markov Decision Process and uses Proximal Policy Optimization to learn a policy for setting optimal prices of private and shared charging stations according to real-time conditions. The state representation is done using ten features such as temporal indicators, current loading on the grid, grid status, traffic volume, and weather data. A multi-objective reward function is an approach to balance four objectives - revenue maximization, station utilization, grid stability, and user satisfaction. The system is trained on actual charging data from a residential complex in Trondheim, Norway. 6878 charging sessions during a 13-month period are used for training. We compare the learned policy with three baseline technologies: fixed pricing, time-of-use pricing and rule-based pricing. Experimental results show that the proposed approach reaches an overall score of 0.569, which is 32.9% and 48.9% improvements in comparison to fixed pricing and time-of-use pricing, respectively. The learned policy is able to successfully adjust the prices based on different conditions and sustain a balanced performance for all the goals. The main contributions include a custom reinforcement learning environment for residential EV charging pricing, a multi-objective reward formulation, and empirical evidence that learned policies outperform traditional pricing approaches.