With a growing penetration of renewable energy generation in the modern power networks, it has become highly challenging for network operators to balance electricity supply and demand. Residential load forecasting nowadays plays an increasingly important role in this aspect and facilitates various interactions between power networks and electricity users. While numerous research works have been proposed targeting at aggregate residential load forecasting, only a few efforts have been made towards individual residential load forecasting. The issue of volatility of individual residential load has never been addressed in forecasting. Thus, to fill this gap, this paper presents a deep learning method empowered with dynamic mirror descent for adaptive individual residential load forecasting. The proposed method is evaluated on a real-life Irish residential load dataset, and the experimental results show that it improves the prediction accuracy by 9.1% and 11.6% in the aspects of RMSE and MAE respectively in comparison with a benchmark method.

With the rapid development and widespread deployment of advanced metering infrastructures, massive amounts of fine-grained data about aggregated residential load have been collected by power network operators, further helping improve forecasting accuracy. Accurate aggregated residential load forecasting plays an increasingly important role in various interactions between power networks and electricity customers as well as in maintaining the continuous balance between electricity supply and demand. However, most existing short-term aggregated residential load forecasting methods only take into consideration the temporal correlation of historical load profiles, ignoring the potential spatial correlation between electricity consumption behaviors of adjacent residential customers. Thus, to fill this gap, this paper proposes a short-term aggregated residential load forecasting method based on graph neural networks and K-means clustering. Specifically, K-means clustering is firstly used to divide residential customers into different groups according to the similarity of their electricity consumption behaviors. Then, the spatial–temporal graph series for aggregated residential load forecasting is constructed, based on the number of groups of residential customers, the aggregated historical load profile of each group of residential customers, and the correlation between the aggregated historical load profiles of different groups of residential customers. Finally, adaptive spatial–temporal synchronous graph convolutional networks are applied to perform short-term aggregated residential load forecasting. The proposed method is evaluated on a real-life Irish residential load dataset, and the experimental results demonstrate that it can improve forecasting accuracy significantly in comparison with a number of traditional benchmark methods.