This technology is an energy storage arbitrage algorithm that uses deep learning predicted opportunity value functions.
Current power systems are transitioning to more sustainable outlets that make it difficult to balance electricity supply and demand in real-time. Precise prediction of real-time energy market prices can ensure profitable energy market arbitrage. However, various methods that predict real-time price are disadvantaged with having a high mean absolute percentage error, due to the highly stochastic and confidential nature of energy market prices.
This technology is an energy storage price arbitrage algorithm that uses deep learning to predict opportunity value functions. The opportunity value function offers more stability compared to the highly unpredictable real-time price. Moreover, transfer learning is used to increase the robustness of the model, promoting high performance even with minimal data and low computational power, making it scalable for market implementation.
This technology has been validated in a case study on historical prices in NYISO.
IR CU23095, CU23109
Licensing Contact: Dovina Qu