Optimization and Scalability of Blockchain Enabled Demand Response Smart Contracts using Sharding and Neural Networks

Utilities have a need for demand reduction in response to energy events that occur throughout the year such as extreme weather or other heavy energy usage periods. In addition, utilities must mitigate the increased variability of load flow on the grid due to increased renewable energy generation being added. Demand response (DR) is one solution utilities use to address these issues. However, DR is primarily for large energy consumers. As such, smart contracts on a permissioned blockchain are proposed in order to enable smaller residential energy customers to participate in demand response mitigation and peer-to-peer energy trading without the need for a trusted third party. Thus, increasing the available pool of demand reduction participants with minimal additional management. A demand response smart contract is created on Ethereum using the Remix – Ethereum IDE. LSTM (Long Short-Term Memory), a deep learning based Neural Network, is used to predict the future energy trading prices based on past power purchase agreement (PPA) prices, which is then used in our bidding model for P2P energy trading. This allows both prosumers and consumers to make informed decisions for the energy trading based on potential profit from revenue generated less transaction costs. The blockchain is sharded in order to improve scalability especially around throughput due to the large number of IoT devices and associated data that will be added to the blockchain.