Non-intrusive load monitoring (NILM) is the main method used to monitor the energy footprint of a residential building and disaggregate total electrical usage into appliance-related signals. The most common disaggregation algorithms are based on the Hidden Markov Model, while solutions based on deep neural networks have recently caught the attention of researchers. In this work we address the problem through the recognition of the state of activation of the appliances using a fully convolutional deep neural network, borrowing some techniques used in the semantic segmentation of images and multilabel classification. This approach has allowed obtaining high performances not only in the recognition of the activation state of the domestic appliances but also in the estimation of their consumptions, improving the state of the art for a reference dataset.

The inclusion of photo-voltaic generation in the distribution grid poses technical difficulties related to the variability of the solar source and determines the need for Probabilistic Forecasting procedures (PF). This work describes a new approach for PF based on quantile regression using the Gradient-Boosted Regression Trees (GBRT) method fed by numerical weather forecasts of the European Centre for Medium Range Weather Forecast (ECMWF) Integrated Forecasting System (IFS) and Ensemble Prediction System (EPS). The proposed methodology is compared with the forecasts obtained with Quantile Regression using only IFS forecasts (QR), with the uncalibrated EPS forecasts and with the EPS forecasts calibrated with a Variance Deficit (VD) procedure. The proposed methodology produces forecasts with a temporal resolution equal to or better than the meteorological forecast (1 h for the IFS and 3 h for EPS) and, in the case examined, is able to provide higher performances than those obtained with the other methods over a forecast horizon of up to 72 h.

In this article, a nowcasting technique for meteorological radar images based on a generative neural network is presented. This technique’s performance is compared with state-of-the-art optical flow procedures. Both methods have been validated using a public domain data set of radar images, covering an area of about 10⁴ km² over Japan, and a period of five years with a sampling frequency of five minutes. The performance of the neural network, trained with three of the five years of data, forecasts with a time horizon of up to one hour, evaluated over one year of the data, proved to be significantly better than those obtained with the techniques currently in use.