Real-Time Economic Optimization of Storage

Presentation Description: The value of storage and renewables plus storage can be significantly increased through analytics that continuously optimize operations through a combination of AI techniques. With application of reinforced learning for bid optimization and forecasting future market conditions with supervised learning, AI has become a key ingredient to maximizing value. This presentation will address analytic techniques to improve the economic intelligence of ISO bids along with direct override control signals to further advance storage economics. The presentation will show a case study in ERCOT of the optimization techniques and AI structure to increase storage project revenues by 30% over traditional approaches. The AI methods we apply will address techniques for next-day and next-hour probabilistic price spikes. The dynamic optimization structure creates economically optimal offer strategies that maximize revenue given the price forecasts while managing the state of charge of the system.

Key Findings:
• Predictive price analytics add most value through a probabilistic framework rather than an absolute price forecast
• Real-time price spikes can be predicted with significant bearing on operational strategy
• Advanced optimization algorithms are the cornerstone to pair battery attributes with market dynamics
• Profitability can be increased by more than 30% through advanced analytics