Following a successful summer 2024 pilot, Predictive Capacity Dispatch (PCD) has achieved a breakthrough in demand response reliability, using AI to predict and optimize load shed over extended periods. Uplight’s technology transforms thermostats from unpredictable grid resources into valuable, controllable capacity assets for utilities.
During the recent heat wave across the US, PCD technology deployed its first live events through Uplight’s enhanced Flex platform. On July 29th, a Southeastern utility called a three-hour PCD event, with the system predicting it could deliver a sustained 37 MW of load shift from 34,000 smart thermostats. The official M&V results confirmed that the technology works, delivering 36 MW of constant, firm, predictable load shift. This achievement moves demand response programs from helpful last-resort options to direct capacity resources in the utility Demand Stack.
Enhanced Accuracy and Consistency Through Individualized AI Modeling
What sets Uplight’s PCD apart is its sophisticated, household-level approach. Using live OEM device telemetry, AMI data, and high resolution weather data, Uplight’s improved algorithms create individualized load forecasts for each participating household. This granular modeling captures unique user comfort preferences, home characteristics, and usage patterns, enabling the system to predict future household load curves with high accuracy and optimize staggered dispatch strategies for maximum load shed consistency.
The results speak for themselves: 3% prediction error forecasted 24 hours before event dispatch, and less than 5% variability in per-hour load shift performance. This precision transforms non-steerable assets into predictable, controllable demand response resources while delivering more consistent load shed throughout each event.
Enhanced Customer Experience and Platform Capabilities
The enhanced Flex platform now offers utilities direct scheduling capabilities for PCD, providing greater control and flexibility in event management. Meanwhile, PCD’s staggered dispatch approach reduces total demand response hours for enrolled participants, significantly improving the customer experience. Recent PCD events showed an average 15% reduction in opt-out rates compared to the utility’s early season standard events, demonstrating decreased customer demand response fatigue and improved satisfaction.
The success extends beyond the July 29th flagship event. The first and second PCD events delivered similarly strong performance, proving the system’s repeatability while showcasing the AI model’s ability to learn and improve as more event data is collected, with even greater precision and reliability expected as the technology scales.
