Sonasoft Corp. announced the successful completion of their proof of concept with Delaware Electric Cooperative (DEC). The PoC leveraged NuGene, Sonasoft’s AI Bot Engine, to give accurate demand forecasts. The PoC demonstrated that NuGene can deliver annual cost savings of up to $1.3m. As a result, DEC has green-lighted a move to production starting June 1st. DEC doesn’t want to trigger LC events unnecessarily. However, they also need to try and avoid unplanned CP events, which can cost over $1m in the peak summer season. In this PoC, Sonasoft NuGene was tasked with creating an AI bot to deliver 3 key predictions: Long-range (1-2 week) predictions of likely peak and non-peak days: Short-term (24 hours) predictions of the precise peak type and timing; A specific recommendation for when to trigger an LC event. The overall aim was to reduce unnecessary LC events whilst ensuring DEC never missed a CP event. In order to do this, NuGene uses accurate real-time weather variables including temperature, humidity, sunlight intensity, wind velocity, atmospheric pressure and storm paths, along with historical data of electric consumption, to accurately forecast electric consumption. NuGene trained the forecasting bot on DEC’s historical data and then validated it with the data for 2019. It was able to deliver a 12% improvement in peak predictions compared to DEC’s existing methodology. It was also able to reduce unnecessary LC events by 5%. The model was able to deliver R2 (aka R-squared) scores of between 0.921 and 0.947. R2 measures how accurately the model is predicting reality. To put this in context, many advanced AI models manage scores of less than 0.9. `Moving forward, DEC will use Sonasoft’s AI solution to accurately predict peak usage and minimize the number of load control events. This will allow the utility company to further reduce costs, increase member satisfaction, and deliver reliable power to its members.