Nostradamus is a neural network, short-term demand and price forecasting system, designed for electric and gas utilities, system operators and power pools, cooperatives, energy marketers, and gas pipelines.
With Nostradamus, organizations can improve forecast accuracy by tracking forecasts with actual values as well as update new information as it becomes available. The neural network algorithm learns the pattern of load or price changes from numerous weather variables as well as day of week, time of year, and holidays by learning from historical data.
Nostradamus can be fully integrated with existing systems and operational within one week (including training), with no need to involve IT staff. It includes:
- Neural network technology – recognizes relationships and applies this knowledge to produce accurate daily and hourly demand and price forecasts.
- Flexible data integration – for creating a customized model set-up, forecast execution, and results processing.
- Rules-based error handling – to identify, cleanse, and manage data in a way that can still produce accurate forecasts.
- Load duration with corresponding forecast – for ensuring that the newly constructed model properly meets the peaks and valleys of the respective load patterns.
- Actual vs. forecast – plotted over time gives the finest detail needed to clearly isolate specific points in time where the new model is not measuring up.
- Error by timeframe – can indicate outlying months, days, or hours, as well as isolate where adjustments need to occur.
- A scatter plot – of actual data versus forecasts can depict outer points immediately.