SQC and Schneider Electric's quantum chip boosts household energy forecasts by 20%

[Rooney]

Australian company Silicon Quantum Computing and Schneider Electric have moved to the second stage of the Australian government's Critical Technologies Challenge Program. They have been awarded A$3.6 million to expand a project that uses quantum hardware to improve energy forecasting, working with UNSW Sydney. It is one of the more practical quantum stories of the week, because it involves a real business problem with measurable results

The hardware is SQC's chip called Watermelon, which the company describes as an atomically engineered, quantum enhanced AI chip. Rather than running the whole model on a quantum computer, it generates quantum features that are fed into the model alongside normal classical features. The aim is a richer predictive model for next day energy demand across household systems. It is available through the cloud or as hardware, including a turnkey data centre deployment

The numbers from the first stage are interesting. Over a 12 month period, the team reported an average 20 percent improvement in forecasting accuracy against the classical benchmark, with gains of up to 41 percent. Stage 2 will expand the modelling to hundreds of Australian homes and integrate Watermelon directly into Schneider Electric's AI workflows for production use

SQC founder and chief executive Michelle Simmons said quantum processors would work alongside CPUs and GPUs to deliver real world gains, which is the hybrid approach most of the industry now talks about. Colette Munro, Schneider Electric's Pacific Zone president, pointed out that solar panels, electric cars and home batteries have created new levels of complexity in the energy system. Forecasting demand when homes can produce and store their own power is a much harder problem than it used to be

I would like to know exactly what the classical benchmark was, since a 20 percent gain over a weak baseline means less than one over a strong model. Even so, this is the kind of practical test quantum needs. Does anyone here have solar and batteries at home? Would better forecasting make a difference to your bills?

SQC and Schneider Electric Advance Quantum-Enhanced Energy Forecastingthequantuminsider.com

[Nicola47]

The benchmark question is the first thing I wondered too. Beating a basic statistical model is easy, beating a well tuned neural network is not. Without that detail, 20 percent is hard to judge. Peer reviewed results would settle it