Abstract:In recent years, hybrid Solar-Wind energy system has emerged as a viable solution to achieve sustainable energy generation and alleviate the burden on the power grid. However, enhancing the system configuration to balance energy production and consumption remains a challenging task. In this study, we propose an energy forecasting methodology that leverages transformers as an AI tool to predict energy production from a hybrid Photovoltaic-Wind system in an urban environment. The methodology involves pre-processing data, training the transformer model, and generating energy forecasts. The research utilized a one-year dataset, collecting data at 10-minute intervals, encompassing various parameters like wind speed, wind direction, global horizontal irradiance, direct normal irradiance, and diffuse horizontal irradiance, all for the purpose of wind and solar energy prediction. The study additionally examined how the model's performance is impacted by fine-tuning hyperparameters and revealed that this dependency is inversely proportional to characteristics such as training data, horizon, and learning rate. Hyperparameters such as look back and epoch, on the other hand, were observed to have a direct link with the model's dependency. Multiple simulations have been conducted to fine-tune the hyperparameters of the machine learning model to improve its efficiency. The outcomes exhibited the energy forecasting methodology's effectiveness in predicting energy production for a hybrid Photovoltaic-Wind system in an urban environment. The forecasting accuracy for solar energy and wind speed reached impressive levels, with 90.7% and 90.4% respectively, across various horizons. The time resolution used for forecasting was 10 min, and the data was split into training, testing, and validation sets with ratios of 60–20–20% and 70–15–15%. Leveraging a transformer model for energy forecasting with an accuracy up to 90% holds the promise of streamlining the planning and management of hybrid PV-Wind-Battery storage systems. This, in turn, can play a critical role in advancing the adoption of sustainable and efficient energy solutions for both residential and commercial structures. The code for timeseries data forecasting is available at https://www.kaggle.com/code/alijavaid43/transformer-model .

A Novel Hybrid Transformer-Based Framework for Solar Irradiance Forecasting under Incomplete Data Scenarios

Hidformer: Hierarchical Dual-Tower Transformer Using Multi-Scale Mergence for Long-Term Time Series Forecasting

Short-Term Irradiance Prediction Based on Transformer with Inverted Functional Area Structure

Enhancing One-Day-Ahead Probabilistic Solar Power Forecast With a Hybrid Transformer-LUBE Model and Missing Data Imputation

Intra-hour solar irradiance forecasting: An end-to-end Transformer-based network

Leveraging advanced AI algorithms with transformer-infused recurrent neural networks to optimize solar irradiance forecasting

Enhanced Photovoltaic Power Forecasting: An iTransformer and LSTM-Based Model Integrating Temporal and Covariate Interactions

A novel intra-hour PV power forecasting technique based on total-sky images

Multi-step solar radiation prediction using transformer: A case study from solar radiation data in Tokyo

Photovoltaic Power Prediction Based on Irradiation Interval Distribution and Transformer-LSTM

A Sky Image-Based Hybrid Deep Learning Model for Nonparametric Probabilistic Forecasting of Solar Irradiance

Hourly Solar Irradiance Forecasting Based on Encoder–decoder Model Using Series Decomposition and Dynamic Error Compensation

Photovoltaic power forecasting: A hybrid deep learning model incorporating transfer learning strategy

A hybrid deep learning model with an optimal strategy based on improved VMD and transformer for short-term photovoltaic power forecasting

Hourly stepwise forecasting for solar irradiance using integrated hybrid models CNN-LSTM-MLP combined with error correction and VMD

Harnessing AI for solar energy: Emergence of transformer models

A Novel Improved Variational Mode Decomposition-Temporal Convolutional Network-Gated Recurrent Unit with Multi-Head Attention Mechanism for Enhanced Photovoltaic Power Forecasting

Solar Irradiance Anticipative Transformer

Solar irradiance forecasting using multi-layer cloud tracking and numerical weather prediction

Improving ultra-short-term photovoltaic power forecasting using a novel sky-image-based framework considering spatial-temporal feature interaction

Sustainable urban energy solutions: Forecasting energy production for hybrid solar-wind systems