Utilities face tremendous pressure to reduce electricity demand and encourage residential energy efficiency. The residential sector accounts for 38% of total electric usage – more than the commercial or industrial sectors.
Unfortunately, 83% of residential customers have very limited knowledge of how they use electricity. Compounding this general lack of energy education, consumers lack the motivation to take action because it has historically been difficult to show consumers how their individual energy choices can make a difference.
“The utilities that succeed in educating their residential customers may be the best positioned – not only to survive but also to thrive in the new customer-centric electricity landscape that is now taking shape”, says Utility Dive in their residential customer survey.
To realize the savings associated with increased energy efficiency, utilities need to find new ways to engage and motivate residential customers to save energy. While 76% of utilities agree that consumer education is a higher priority for their business today, only 2% believe they’re doing a good job at it.
Where Software-Based Disaggregation Falls Short
Disaggregation breaks the entire house electrical usage into appliance-specific usage. For example, instead of only knowing that your home used 27 kilowatt-hours on Tuesday, disaggregation breaks that 27 kWh down into usage by specific appliances. With access to appliance-level energy usage information, consumers can see exactly how they are using energy.
However, as we will see below, our ability and desire to act on this information is based on the accuracy of the information itself. Without accuracy, consumers will distrust the data and disengage.
Of the companies employing disaggregation methods in the market, most use a software-based approach, also known as Non-Intrusive Load Monitoring (NILM). Using machine learning algorithms, estimates are developed for appliance-specific usage by using only whole house consumption as an input. This appears to be an appealing option on the surface given that no plugs, sensors, or additional hardware are required to generate usage estimates.
In practice, software-based disaggregation has not been successfully deployed on a wide scale despite the fact that research began in 1982. Many companies offer energy management solutions that integrate software-based algorithms because additional hardware is not required. However, no utility has yet to deploy accurate software-based disaggregation to a large group of residential customers. This may be because systematic reviews show that providing information alone, as with the software-based approaches, fails to decrease energy usage.
While the absence of hardware and the associated costs would make software-based methods seem like an attractive option for disaggregation, but the following characteristics limit its viability as a large-scale energy saving tool:
- Software-based disaggregation is not highly accurate. A recent study found that three industry-leading vendors that employ software-based approaches had error rates between 40-50% for clothes dryers and 30-80% for refrigerators (and these are just two examples!). Software-based algorithms yield high error rates because they rely solely on household usage data. This is especially problematic because low accuracy leads to low consumer trust and therefore lower engagement. For example, software-based methods may fail to detect an individual dryer cycle, which would cause the customer to lose faith in the solution itself.
- Software-based disaggregation is often neither real-time nor granular. Many software-based disaggregation algorithms only report usage at the daily or monthly levels. Aggregate usage over time offers substantially less customer engagement opportunities than a minute-by-minute real-time breakdown of appliance usage.
- Software-based disaggregation is not actionable. While these methods provide estimates of usage associated with each appliance, software-based approaches do not provide a clear path from appliance-specific energy use to savings. For instance, these methods may tell the consumer that the air conditioner used 6 kWh yesterday, but what actions can he take to improve that? Is the air conditioner working harder than expected to maintain a comfortable temperature given the outside temperature? Is the air conditioner wasting energy by running when the house is empty? Has the air conditioner’s performance deteriorated in the past month, indicating maintenance is required? The consumer simply doesn’t have enough information to act.
Load-ID: Hardware Drives Accuracy and Action
Powerley has developed a novel, scalable approach, called Load-ID, that surpasses the accuracy of software-based disaggregation methods. The key component to elevate accuracy is hardware. Through this hardware-assisted approach, the solution delivers accuracy in real-time and with finer granularity. Load-ID utilizes machine learning algorithms that leverage data from smart thermostats as well as connected appliances and devices.
Why is this important? Powerley’s machine learning algorithms achieve substantially lower error rates and can guide users to use less energy through personalized, data-driven coaching. It has the potential to drive consumer actions to increase energy efficiency because, unlike software-based disaggregation methods, the approach is:
- Highly accurate. Error rates are only 2-3% compared to the 20-40% of standard software-based algorithms. Powerley’s Load-ID algorithm will detect every appliance cycle, without any false positives or false negatives that would reduce consumer trust.
- Real-time and granular. The Load-ID algorithm allows customers to view appliance-specific usage every minute. This enables consumers to easily identify energy waste, such as an HVAC system working to cool an empty home.
- Actionable. The highly accurate results will allow for intelligent energy coaching. Given the granularity, it can be used to detect deteriorating performance of appliances and energy waste. The platform empowers the user to take action directly through the mobile app, even going as far as automating energy savings through personalized scheduling capabilities.
Powerley’s hardware-assisted Load-ID algorithm goes beyond pure software disaggregation methods to elevate accuracy which, in turn, is immensely more effective at impacting consumer behavior to increase energy efficiency.
Software-based Disagg | Hardware-Assisted Load-Id | |
---|---|---|
Error Rates | 20-40% | <3% |
Granularity | Daily | Seconds |
Real-Time | No | Yes |
Actionable | No | Yes |
In the next article, we’ll explore how both utilities and consumers can achieve energy efficiency goals through a hardware-assisted Load-ID methodology. We will also show how a real-time connection to energy usage data can redefine the consumer relationship with energy and the utility.