AI-Powered Decentralized Energy Trading Platform with BiLSTM-based Generation Prediction and Carbon Footprint Analysis

The increasing complicacy of centralized power trading systems exposes them to risks such as price manipulation and cyberattacks. In contrast, the decentralized and immutable characteristics of blockchain technology present a promising solution to increase the integrity and efficiency of these systems. This paper proposes a decentralized energy trading platform that integrates machine learning (ML) for energy generation prediction and a carbon footprint calculator. The blockchain component is implemented on the Ethereum platform, utilizing smart contracts written in Solidity, with testing conducted via Ganache. The energy generation forecasting leverages a Bidirectional Long Short-Term Memory (BiLSTM) model, which illustrates superior accuracy and minimal validation loss compared to Stacked LSTM and Vanilla LSTM models. LSTMs are recognized as effective tools for predicting energy generation due to their capability to capture temporal dependencies in data. To obtain real-time weather data, the Weatherstack API is employed, allowing for accurate energy generation predictions based on current conditions. The model is trained on two datasets including energy production and weather data from five cities in Spain. Additionally, the platform features a carbon footprint calculator that calculates CO2 emissions associated with various energy purchases, comparing them against emissions from fossil fuel consumption. The integration of decentralized energy trading, accurate energy generation forecasting, and a carbon footprint assessment tool concludes in a comprehensive platform hosted locally using React. This innovative approach not only improves transparency and security in energy transactions but also promotes sustainability by informing users about their environmental impact.