Abstract
This paper presents three ensemble learning models for short term load forecasting. Machine learning has evolved quickly in recent years, leading to novel and advanced models that are improving the forecasting results in multiple fields. However, in highly dynamic fields such as power and energy systems, dealing with the fast acquisition of large amounts of data from multiple data sources and taking advantage from the correlation between the multiple available variables is a challenging task, for which current models are not prepared. Ensemble learning is bringing promising results in this sense, as, by combining the results and use of multiple learners, is able to find new ways for current learning models to be used and optimized. In this paper three ensemble learning models are developed and the respective results compared: gradient boosted regression trees, random forests and an adaptation of Adaboost. Results for electricity consumption forecasting in hour-ahead are presented using a case-study based on real data from an office building. Results show that the adapted Adaboost model outperforms the reference models for hour-ahead load forecasting.

Abstract
Recent changes in the energy sector are increasing the importance of portfolio optimization for market participation. Although the portfolio optimization problem is most popular in finance and economics, it is only recently being subject of study and application in electricity markets. Risk modeling in this domain is, however, being addressed as in the classic portfolio optimization problem, where investment diversity is the adopted measure to mitigate risk. The increasing unpredictability of market prices as reflection of the renewable generation variability brings a new dimension to risk formulation, as market participation risk should consider the prices variation in each market. This paper thereby proposes a new portfolio optimization model, considering a new approach for risk management. The problem of electricity allocation between different markets is formulated as a classic portfolio optimization problem considering market prices forecast error as part of the risk asset. Dealing with a multi-objective problem leads to a heavy computational burden, and for this reason a particle swarm optimization-based method is applied. A case study based on real data from the Iberian electricity market demonstrates the advantages of the proposed approach to increase market players? profits while minimizing the market participation risk.

Abstract
Current approaches for risk management in energy market participation mostly refer to portfolio optimization for long-term planning, and stochastic approaches to deal with uncertainties related to renewable energy generation and market prices variation. Risk assessment and management as integrated part of actual market negotiation strategies is lacking from the current literature. This paper addresses this gap by proposing a novel model for decision support of players' strategic participation in electricity market negotiations, which considers risk management as a core component of the decision-making process. The proposed approach addresses the adaptation of players' behaviour according to the participation risk, by combining the two most commonly used approaches of forecasting in a company's scope: the internal data analysis, and the external, or sectorial, data analysis. The internal data analysis considers the evaluation of the company's evolution in terms of market power and profitability, while the sectorial analysis addresses the assessment of the competing entities in the market sector using a K-Means-based clustering approach. By balancing these two components, the proposed model enables a dynamic adaptation to the market context, using as reference the expected prices from competitor players, and the market price prediction by means of Artificial Neural Networks (ANN). Results under realistic electricity market simulations using real data from the Iberian electricity market operator show that the proposed approach is able to outperform most state-of-the-art market participation strategies, reaching a higher accumulated profit, by adapting players' actions according to the participation risk.

Abstract
This paper proposes an optimal bidding strategy for autonomous residential energy management systems. This strategy enables the system to manage its domestic energy production and consumption autonomously, and trade energy with the local market through a novel hybrid interval-stochastic optimization method. This work poses a residential energy management problem which consists of two stages: day-ahead and real-time. The uncertainty in electricity price and PV power generation is modeled by interval-based and stochastic scenarios in the day-ahead and real-time transactions between the smart home and local electricity market. Moreover, the implementation of a battery included to provide energy flexibility in the residential system. In this paper, the smart home acts as a price-taker agent in the local market, and it submits its optimal offering and bidding curves to the local market based on the uncertainties of the system. Finally, the performance of the proposed residential energy management system is evaluated according to the impacts of interval optimistic and flexibility coefficients, optimal bidding strategy, and uncertainty modeling. The evaluation has shown that the proposed optimal offering model is effective in making the home system robust and achieves optimal energy transaction. Thus, the results prove that the proposed optimal offering model for the domestic energy management system is more robust than its non-optimal offering model. Moreover, battery flexibility has a positive effect on the system's total expected profit. With regarding to the bidding strategy, it is not able to impact the smart home's behavior (as a consumer or producer) in the day-ahead local electricity market.

Abstract

Abstract
This paper proposes a novel Case Based Reasoning (CBR) application for intelligent management of energy resources in residential buildings. The proposed CBR approach enables analyzing the history of previous cases of energy reduction in buildings, and using them to provide a suggestion on the ideal level of energy reduction that should be applied in the consumption of houses. The innovations of the proposed CBR model are the application of the k-Nearest Neighbors algorithm (k-NN) clustering algorithm to identify similar past cases, the adaptation of Particle Swarm Optimization (PSO) meta-heuristic optimization method to optimize the choice of the variables that characterize each case, and the development of expert systems to adapt and refine the final solution. A case study is presented, which considers a knowledge base containing a set of scenarios obtained from the consumption of a residential building. In order to provide a response for a new case, the proposed CBR application selects the most similar cases and elaborates a response, which is provided to the SCADA House Intelligent Management (SHIM) system as input data. SHIM uses this specification to determine the loads that should be reduced in order to fulfill the reduction suggested by the CBR approach. Results show that the proposed approach is capable of suggesting the most adequate levels of reduction for the considered house, without compromising the comfort of the users.