This technology provides observability of behind-the-meter distributed energy resources, such as distributed photovoltaics and electric vehicles, using nothing more than the recordings from smart meters.
Behind-the-meter solar photovoltaics are a significant contributor to the total energy contributed by solar photovoltaics (PV). In the US, the latest estimates place distributed PV generation at approximately one-third of the total.[1] While the estimates of the future percentage of distributed PV relative to total vary from ~25% in the US[2] to 45% globally,[3] the total PV generation is expected to become a large contributor to total energy supply, resulting in growing reliance on poorly observable generation sources. This presents a significant challenge in power system planning and operation, because power system transients may precipitate sudden en masse disconnection of distributed PV assets, which leads to compounding other transient-induced problems. For illustration, in the August 2019 UK outage, an estimated 500MW of distributed generation was lost in the event, which accounted for 26.6% of the cumulative generation loss of 1878MW.[4]
Any potential loss of generation must be accounted for in operating reserves. To accurately select the level of reserves, the power system planners and operators need better visibility into the expected contribution from distributed energy resources. This is challenging because, in most cases, the energy produced by behind-the-meter resources is consumed locally, so the metered consumption only shows the net load.
Our proprietary technology uses image processing and transfer learning to discover the presence of behind-the-meter distributed PV and to measure its energy contribution using solely the readings from smart meters. We are currently working towards increasing efficiency of calculations and extending the technique to capture other load properties such as detecting the presence and charging patterns of electric vehicles and use of energy storage (both physical and virtual) to systematically modify load patterns.
We are interested in partnering with US-based utilities to help us test the accuracy of our methodology in the diverse set of climates. For more information, please contact us.