Abstract:The intermittent nature of renewable resources like solar and wind presents challenges for small-scale energy markets and off-grid regions. Localized forecasting of these resources is crucial to incentives the attention of citizens and urban designers in adopting small-scale generation systems, i.e., hybrid or renewable energy-based microgrids, in their communities. This study employs Encoder–Decoder Sequence-to-Sequence (Seq2Seq) models with two different recurrent neural network architectures (RNN), long short-term memory (LSTM) and gated recurrent unit (GRU), to forecast wind speed and solar radiation on a location appointed to install microgrids. The models were trained using meteorological data obtained from weather stations while taking into account data on extreme weather events, such as wind gusts, to enhance the predictive accuracy of the models. Particularly relevant in the context of the increasing menace of climate change and the resulting rise in the frequency and severity of extreme weather events. An ensemble method is also employed to quantify uncertainty in the model's prediction to increase the confidence that an end-user of a microgrid can deposit in the forecasting for decision-making. We compared the Seq2Seq-LSTM and Seq2Seq-GRU models using deterministic and probabilistic metrics. Both models showed satisfactory accuracy and confidence in wind speed and solar radiation forecasting. However, the Seq2Seq-GRU model stood out for its significantly faster computation times. On average, it achieved a 20% reduction in processing time for the same dataset. The models performed best for shorter prediction horizons (1-hour) and higher lookback periods ( k = 720-hrs). We also analyzed the models' performances according to the season. The results show that the seasonal variation significantly affects the accuracy and uncertainty of wind speed forecasting, with fall and summer demonstrating the highest accuracy and winter and spring the lowest. For solar radiation, the models show greater reliability, narrower prediction intervals, and higher accuracy during winter, with opposite trends observed during summer. The deep learning framework proposed in this study can help anticipate daily surpluses or shortfalls in energy generation across different seasons, contributing to the efficient operation of microgrids.

Interpretable Probabilistic Forecasting of Imbalances in Renewable-Dominated Electricity Systems

Probabilistic Forecasting of Photovoltaic Generation: an Efficient Statistical Approach

Nonparametric Stochastic Differential Equations for Ultra-Short-Term Probabilistic Forecasting of Wind Power Generation

Towards smart energy management for community microgrids: Leveraging deep learning in probabilistic forecasting of renewable energy sources

Probabilistic forecasting of power system imbalance using neural network-based ensembles

Probabilistic Wind Power Forecasting with Hybrid Artificial Neural Networks

Decomposition aided attention-based recurrent neural networks for multistep ahead time-series forecasting of renewable power generation

An integrated method with adaptive decomposition and machine learning for renewable energy power generation forecasting

A Hybrid Approach for Probabilistic Forecasting of Electricity Price.

The Key Technology for Grid Integration of Wind Power: Direct Probabilistic Interval Forecasts of Wind Power

Probabilistic forecasting of German electricity imbalance prices

Interpretable Short-Term Load Forecasting via Multi-Scale Temporal Decomposition

A deep generative model for probabilistic energy forecasting in power systems: normalizing flows

Interpretable Transformer Model for Capturing Regime Switching Effects of Real-Time Electricity Prices

Short-term wind-PV-wave power probabilistic prediction based on an improved sequence-to-sequence attention model

Probabilistic Short-Term Wind Power Forecasting Based on Deep Neural Networks

Probabilistic Forecasting of Real-Time Electricity Market Signals via Interpretable Generative AI

Forecasting Intraday Power Output by a Set of PV Systems using Recurrent Neural Networks and Physical Covariates

Electricity load forecasting based on an Interpretable ProbSparse attention mechanism

Combating Uncertainties in Wind and Distributed PV Energy Sources Using Integrated Reinforcement Learning and Time-Series Forecasting

An Adaptive, Data-Driven Stacking Ensemble Learning Framework for the Short-Term Forecasting of Renewable Energy Generation