This paper presents the application of collaborative reinforcement learning models to enable the distributed learning of energy contracts negotiation strategies. The learning model combines the learning process on the best negotiation strategies to apply against each opponent, in each context, from multiple learning sources. The diverse learning sources are the learning processes of several agents, which learn the same problem under different perspectives. By combining the different independent learning processes, it is possible to gather the diverse knowledge and reach a final decision on the most suitable negotiation strategy to be applied. The reinforcement learning process is based on the application of the Q-Learning algorithm; and the continuous combination of the different learning results applies and compares several collaborative learning algorithms, namely BEST-Q, Average (AVE)-Q; Particle Swarm Optimization (PSO)-Q, and Weighted Strategy Sharing (WSS)-Q. Results show that the collaborative learning process enables players’ to correctly identify the negotiation strategy to apply in each moment, context and against each opponent.

Electricity markets are complex environments with very dynamic characteristics. The large-scale penetration of renewable energy sources has brought an increased uncertainty to generation, which is consequently, reflected in electricity market prices. In this way, novel advanced forecasting methods that are able to predict electricity market prices taking into account the new variables that influence prices variation are required. This paper proposes a new model for day-ahead electricity market prices forecasting based on the application of an artificial neural network. The main novelty of this paper relates to the pre-processing phase, in which the relevant data referring to the different variables that have a direct influence on market prices such as generation, temperature, consumption, among others, is analysed. The association between these variables is performed using spearman correlation, from which results the identification of which data has a larger influence on the market prices variation. This pre-analysis is then used to adapt the training process of the artificial neural network, leading to improved forecasting results, by using the most relevant data in an appropriate way.

The rapid growth of energy consumption and its consequences in the last decades, made the world persuaded to energy optimization and energy management. Therefore, producers and prosumers should be equipped with the automation infrastructures to perform the management programs, such as demand response programs. Office buildings are considering as a proper case for implementing energy consumption minimization since they are responsible for a huge portion of total energy consumption in the world. This paper proposes a multi-period optimization algorithm implemented in Supervisory Control and Data Acquisition system of an office building. The developed optimization algorithm is an efficient solution considered for minimizing the power consumption of air conditioners by considering the user comfort constraints. Two determinative parameters are defined to prevent over-power reduction from certain devices. In order to respect to user preferences, priority numbers are dedicated to each air conditioner to present the importance of each device. A case study with several scenarios is implemented to verify the performance of the proposed algorithm in real life using real data of the building. The obtained results show the impacts of proposed parameters and different comfort constraints of algorithm while the main target of the optimization has been reached.

The current power systems, namely the low voltage distribution networks, have been suffering considerable changes in recent years. What appeared to be innovation trends nowadays due to technological advances and manufacturing cost reduction has become the new reality in the coming years. Thus, the growing trend of power generation by renewable sources has posed new challenges and new opportunities. Furthermore, the wide installation of “smart meters” and the interest in placing the citizens as core players into the future energy markets and systems operation improves the role of the distribution system operator. In this way, developing new and innovative methodologies to explore the potential mechanisms for providing ancillary services in distribution networks becomes of great importance, namely in low voltage levels. This research paper proposes an innovative methodology to enhance the demand response participation of small consumers and dispatchable distributed renewable energy sources flexibility as ancillary services to mitigate the voltage and congestion issues in low voltage distribution networks. A realistic low voltage distribution network with 236 buses is used to illustrate the application of the proposed model. The results demonstrate a considerable voltage profile and congestion improvements.

The current energy scenario requires actions towards the reduction of energy consumption and the use of renewable resources. In this context, a microgrid is a self-sustained network that can operate connected to the smart grid or in isolation. The long-term scheduling of on/off cycles of devices is a critical problem that has been commonly addressed by centralized approaches. In this work, we propose a novel agent-based method to solve the long-term scheduling problem as a distributed constraint optimization problem (DCOP) by modelling future system configurations rather than reacting to changes. Moreover, with respect to approaches based on decentralised reinforcement learning, we can directly encode system-wide hard constraints (such as for example the Kirchhoff law) which are not easy to represent in a factored representation of the problem. We compare different multi-agent DCOP algorithms showing that the proposed method can find optimal/near-optimal solutions for a specific case study.

This paper presents a new multi-agent decision support system with the purpose of aiding bilateral contract negotiators in the pre-negotiation phase, through the analysis of their possible opponents. The application area of this system is the electricity market, in which players trade a certain volume of energy at a specified price. Consequently, the main output of this system is the recommendation of the best opponent(s) to trade with and the target energy volume to trade with each of the opponents. These recommendations are achieved through the analysis of the possible opponents’ past behavior, namely by learning on their past actions. The result is the forecasting of the expected prices against each opponent depending on the volume to trade. The expected prices are then used by a game-theory based model, to reach the final decision on the best opponents to negotiate with and the ideal target volume to be negotiated with each of them

Building energy management systems have been largely implemented, focusing on specific domains. When installed together, they lack interoperability to make them work correctly and to achieve a centralized user interface. The Building's Reasoning for Intelligent Control Knowledge-based System (BRICKS) overcomes these issues by developing an interoperable building management system able to aggregate different interest domains. It is a context-aware semantic rule-based system for intelligent management of buildings' energy and security. Its output can be a set of alarms, notifications, or control actions to take. BRICKS itself, and its features are the innovative contribution of the present paper. It is very important for buildings' energy management, namely in the scope of demand response programs. In this paper, it is shown how semantics is used to enable the knowledge exchange between different devices, algorithms, and models, without the need for reprogramming the system. A scenario is deployed in a real building for demonstration.

Local electricity markets (LEM) are a promising idea to foster the efficiency and use of renewable energy at the distribution level. However, how these local markets will be integrated into existing market structures, and to make the most profit from them, is still unclear. In this work, we propose a LEM framework based on bi-level optimization. In the upper level, end-users aim at maximizing profits, while the lower level represents the clearing market process. Moreover, a cascade integration to the wholesale market through an aggregator that acts after the LEM has been cleared is considered. Learning strategies using only available information can be a powerful tool to take the most advantage of LEM. To this end, we advocate the use of ant colony optimization (ACO), a nature-inspired technique, similar to that employed in machine learning. By using ACO, consumers, producers and prosumers, can learn the best strategies to maximize their profits without sharing private information and based solely on their experience.