No abstract

No abstract

No abstract

Probabilistic electricity price forecast (EPF) systems represent a fundamental tool to achieve robust production scheduling and day-ahead bidding strategies. However, most EPF methods, including recently proposed deep learning based techniques, are still targeting point predictions, following the common Gaussian assumption. In this work, we propose a novel probabilistic EPF approach based on the integration of a Gaussian Mixture layer, parametrized by a Recurrent Neural Network with Gated Recurrent Units, including an L1-norm based feature selection mechanisms. The network is conceived to approximate general conditional price distributions through learning. Moreover, we developed a multi-hours prediction approach exploiting correlations and patters both in hourly and cross-hour contexts. Experiments have been performed on the Italian market dataset, showing the capability of the proposed method to achieve accurate out-of-sample predictions while providing explicit uncertainty indications supporting enhanced decision making.

No abstract

Conventional Life Cycle Assessment (LCA) that relies on static coefficients is usually based on yearly averages. However, the impacts of electricity supply vary remarkably on an hourly basis. Thus, a company production plan is reassessed to reduce selected LCA impacts due to electricity consumption. To achieve this, the company will need a forecast of hourly LCA impacts due to electricity consumption, which can be directly forecast with the Direct Forecasting (DF) approach. Alternatively, the Electricity Technological Mix Forecasting (ETMF) forecasts the electricity production of the technologies in the mix and subsequently linearly combines it with unitary LCA impact indicators. Here, we assessed different machine learning models to forecast two LCA impact indicators for the consumption of electricity in the Italy-North control zone. The feed-forward neural network (NN) with the ETMF approach was the best perfomer among the assessed forecasting models. In our dataset, recurrent neural networks (RNNs) performed worse than feed-forward neural networks. Due to its better forecasting performance, the ETMF approach was preferred over the DF approach. This was due to its flexibility and scalability with easy updates or expansion of the selected forecast indicators, and due to its ability to assess technology-specific errors in the forecasting. Finally, we propose to adopt the correlation of LCA impact indicators within the dataset to select indicators while avoiding unconscious burden-shifting.

No abstract

No abstract