Energy insights and energy disaggregation are not new things in the energy market. Many utility companies have tested and invested in energy disaggregation services after realising the significance of keeping their customers aware, informed, and happy. But with so many disaggregation services providers out there, a prevalent challenge is to adopt the appropriate services that will produce accurate insights results. From a data science perspective, one of the most prolific reasons for inaccurate energy services is the low granularity of available energy measurements in the market today. Additionally, even while the volume of energy data might be available, the technical know-how in processing them might not be there yet to ensure accurate disaggregation results.
Read on to learn how NET2GRID processes Smart Meter data of all granularities to be an industry standout.
1. Trained models based on real-time information
Energy disaggregation is the result of analysing the energy data input by machine learning algorithms and models. The most commonly-used data sets that are available in the market today are lower-granularity 15-minute smart meter data coming from household meters. But either having 15-minute smart meter data coming from 50 household meters or 15-minute data from 5 million household meters don’t make a big difference, as it doesn’t help in training the appliance-level detailed Machine Learning models as if more granular data were in place. That’s because the amount of information contained in the 15-minute data is still limited. In contrast, NET2GRID’s data repository mainly consists of the more granular 1-min Real-Time data. That allows training of the 15-minute algorithms with a more considerable amount of detail and learnings previously gained from the Real-Time data, thus, making these models much more insightful and deep than the competition can.
2. Largest end-use events dataset amongst competitors
But it’s not only the Smart Meter data that contributes to a disaggregation service’s accuracy. NET2GRID’s models trained on 1-sec data are benchmarked against ground-truth data of more than 50k end-use events (i.e. data coming from the usage of home appliances like EVs, washing machines, ovens, heat pumps & more). This amount has aided in dealing with many appliance scenarios, such as the same appliance that operates in different modes or the same appliance type operating in a similar mode. For example, different modes of a singular washing machine or various types of washing machines executing the same activity but operating in different manners. A ground-truth data set comprised of appliance data that was collected from households across the globe, making the algorithms capable of producing results for different countries and contexts.
3. Nailing energy disaggregation accuracy up to 90%
NET2GRID has 10 years of experience in the energy field with top energy leaders and can provide highly accurate disaggregation services. It is a one-stop-shop company that develops and provides hardware and software services and can access and process ground truth Smart-Meter and Real-Time data. The above practices, along with the benchmarking performed on the models, can assure that not only the capability of analysing smart meter data of all granularities is there but also that the disaggregation models on Real-Time data are up to 90% accurate, a percentage amongst the highest overall in the market today. This allows NET2GRID to measure, for example, the performance of solar panels without added interfaces in the inverter or hardware in the home - just by analysing the available Smart Meter data.
Would you like to learn more about NET2GRID? Request a demo here.