Energy Analytics Company Aims for More Effective Smart Meter Readings-Project Olympus - Carnegie Mellon University

Tuesday, June 14, 2016

Energy Analytics Company Aims for More Effective Smart Meter Readings

More and more, homeowners around Pittsburgh are equipped with digital data on their energy use gathered by smart meters. Duquesne Light Co. and West Penn Power Co. have both been installing more of the devices this spring.

Residential customers who have the meters can log into their online account and pull up a line graph that estimates daily usage, usually in measurements taken every 15 minutes. The conventional wisdom goes that the more informed customers are about their energy consumption, the more efficient their use will be.

A Pittsburgh energy analytics company is taking that notion to the next level — by harnessing 900 times more data, and for every electric appliance a homeowner runs. Early studies have shown that gathering that level of usage data, which is based on each second that appliances are humming along, is far more accurate than what utilities are looking at now.

Pioneering the new way of assessing home energy use is EEme LLC, a Carnegie Mellon spin-out based in Shadyside. EEme is trying to make energy efficiency more accessible for homeowners by making smart meter readings more effective, said Enes Hosgor, its founder and chief executive.

“Behavorial demand response is emerging as a relevant energy resource,” Mr. Hosgor said.

Mr. Hosgor is in the market for what is called home energy “disaggregation” — that is, pulling apart the individual loads in any home and assessing how much electricity flows through each plug.

While measurements can be taken by individual meters attached to each power outlet, those devices can be expensive and burdensome to install.

A less intrusive method is achieved through statistical approaches. Algorithm-based measurements offer an easy method of surveying home energy use based on individual appliances because they do not require any additional hardware, said Carrie Armel, a research associate at Stanford University’s Precourt Energy Efficiency Center.

Previous studies that rely on appliance-level data typically use a meter at the plug level, which would be difficult for utilities or efficiency companies. That’s where the one-second snapshots are valuable: to improve the algorithm, Mr. Hosgor said. What’s more, recent studies testing the algorithms’ accuracy have shown it to be close to actual consumption.

Last year, Mr. Hosgor’s company worked with research group Austin, Texas-based Pecan Street to produce results from what was thought to be the largest disaggregation test yet — a comparison of the results turned out by EEme’s algorithms against the circuit-level, real-time data being collected from 264 homes in Austin.

In April, Pecan Street again tested EEme’s technology against 10 homes, this time using 77 weeks of one-second interval data. The algorithms were able to parse the electricity load of dishwashers, dryers, refrigerators, HVAC systems and electric vehicle chargers within a 10 percent error rate.

Instructions from utilities to customers on how to reduce usage “will generally prove more impactful if it provides specific, targeted suggestions on use of individual appliances that are relevant to the customer as opposed to general messages such as ‘use less energy,’” the report read.

There are some barriers. While smart meters are generally capable of registering data by the second, it often requires reconfiguration that takes many different forms depending on the brand of smart meter, Mr. Hosgor said. He also estimated that “close to nobody” in the utility industry is developing second-level smart meter data.

Mr. Hosgor said he’s agnostic to how data is used; he just wants to get it into as many hands as possible. Broad adoption of his efficiency measurements could depend on policy. Ms. Armel recommended mandates on utilities that allow customers to share the usage data with research institutes and third parties.

“Lack of real life data has been one of roadblocks for algorithm developers,” Ms. Armel said. Read more»

By: Daniel Moore