Distributed Energy Forecasting

Forecast distributed renewable output or demand to guide day-ahead operating decisions.

Business impact

• Forecast accuracy — One case reported solar prediction accuracy improved by 31%.
• Operating cost variation versus perfect forecasts — One case reported operating-cost variation of up to 10% versus perfect forecasts.

Data requirements

• Distributed site output data (tabular) — Used to forecast photovoltaic output across multiple sites.
• Weather forecasts (tabular) — Used as input to wind and solar prediction models.
• Historical turbine data (tabular) — Used to train wind-output forecasts.
• PV-generation history (tabular) — Used in microgrid dispatch planning with forecast inputs.
• EV-load history (tabular) — Used to predict charging demand for station operations.

AI methods and techniques

• Machine learning — Used to learn forecasting patterns from site and history data.
• Neural network — Used to predict energy output from weather and turbine data.
• Multi-layer perceptron neural network — Used for photovoltaic output forecasting across distributed sites.
• EDCF algorithm — Used to forecast charging-station demand and align schedules.
• Persistence approach — Used as a baseline in the microgrid forecasting setup.
• Heuristic method — Used with forecasts to optimize microgrid dispatch decisions.

Industries

Real-world evidence

4 documented case studies on record.

Companies using this: Australian Renewable Energy Agency, SERVEO, Politecnico di Milano, Google DeepMind.