DSM: Don’t Stop Maximizing

Deck: 

Useful analytics to improve program performance.

Fortnightly Magazine - November 2013

Big data and analytics have received a lot of press in recent years, but coverage has been light on specifics and has not focused on one of the most promising, near-term opportunities, that being demand-side management (DSM). Applying analytics to some of the newly available data can improve DSM program economics, and also doesn’t require significant up-front investment to get started.

In general, data analytics capabilities hold promise for many of the functional areas found within a utility, including operations, customer relations, and engineering. One significant and relatively new source of data comes from the millions of smart meters and advanced metering infrastructure (AMI) installed over the past five years.1 Utilities that have installed this type of infrastructure are generally motivated to leverage the volumes of data they’re now collecting. But, as with many new activities, it makes sense to start small and build success along the way.

Given the meter data that’s being collected, DSM program analysis is a good place to start. This data, along with other data such as weather and temperature readings, customer information, and even publically available tax assessor data, can be leveraged to enhance the effectiveness of a utility’s DSM program. 

Better Program Performance

Incorporating analytics into day-to-day DSM program operations can improve program economics on both sides of the equation, providing greater impacts (therefore higher benefits) and reduced costs. Analytics can improve per-customer impacts and participation rates, measurement and verification costs, energy savings forecasting – and even can aid in customer engagement, and integrating DSM into resource planning.

Targeted marketing broadly means finding the right audience for the right product. For DSM programs, it means finding potentially high-impact customers who are likely to participate. Utilities can use meter interval data in combination with other data sources, such as demographic data, to characterize their customers based on geographic location, load shape, and the likelihood of program participation. Thus, they can find the right audience for the right program. For example, program managers can help reduce demand in the areas where it’s most needed by marketing programs specifically to customers in load-constrained areas. In addition, focusing on customers who are likely to participate can reduce marketing costs. By finding the right audience, program managers can achieve both greater impacts and lower costs.

Analytics can help reduce costs incurred for evaluation, measurement and verification (EM&V). Using collected interval data to verify savings can eliminate logger equipment costs and reduce travel expenses and time spent on-site. In some utilities that have deployed AMI, program managers are already realizing these reduced EM&V costs. 

The ability to estimate and predict the impact of demand-side programs – and to align those impacts with overall supply-side requirements – marks another key benefit of demand-side analytics. Utilities often use deemed annual savings for demand-side programs to estimate and project savings into the future. By contrast, the ability to estimate the impact of programs using interval usage data for individual customers can provide more granular estimates that can be combined with participation probabilities to yield a more accurate forecast of program impacts. Likewise, having more granular forecasts of program impacts allow utilities to more accurately integrate demand-side resources into their resource plans and rate structures. The overall improvement in forecast accuracy allows utilities to not only better assess the benefits of their programs, but also determine how to improve those programs going forward.

Marketing, customer engagement, and public relations also can benefit. The analysis of interval usage data, in combination with demographic information and end-use profiles, can allow program managers to tailor marketing messages, focusing on saving money for some customers, versus saving the environment for others. Similarly, marketing messages can be customized to include general information derived from interval usage and billing data. This could be in the form of a message similar to “If you sign up for time-of-use (TOU) pricing and reduce your usage during the afternoon by 5 percent, you can save up to $15 or more off your monthly bill during the summer.” This type of messaging can be highly effective at recruiting customers, and improving program economics. 

Analytics can help avoid unnecessary marketing to uninterested customers. For example, customers with new homes probably aren’t interested in receiving advertisements for an air conditioning upgrade.

Sources of Data

Figure 1 - Analytics Enabled DSM Processes Expand With Increased Data Availability

A number of data sources are available beyond AMI interval data. Each type of data can enable specific, beneficial processes. 

Figure 1 shows a progression of available data, organized into tiers, with the types of enabled processes also identified.

Starting initially with AMI interval data, DSM program managers can begin using analytics to enable processes in Tier 1, such as identifying candidates that could benefit from TOU pricing or a direct load control (DLC) program. As additional data becomes available and is incorporated into the analysis, DSM programs can expand to include the enabled processes identified in Tier 2, Tier 3, and so on.

Other enabled processes might facilitate additional targeted marketing, a streamlining of EM&V, or more fully incorporating DSM into integrated resource plans (IRP). For example, the enabled process marked as “Identify load shifting opportunities,” from Tier 1, can lead to greater demand savings per participant through targeted marketing. Utilities can identify these opportunities by correlating customers’ daily load shapes with peak demand periods.

Another example is the Tier 2 process, “Enhance verification of weather related demand and energy savings.” This process can improve DSM economics by reducing EM&V costs. One component to DSM integration is improving the forecasting accuracy of DSM’s future impact on a utility’s demand. Adding granularity, in combination with geospatial detail, can enable impact forecasting improvements. 

Getting started with analytics doesn’t require millions of dollars in up-front investment. In fact, Tier 1 data analysis for a selected subset of customers can be done with spreadsheet analysis by an experienced staff analyst. (See sidebar, “Heat Slope Analysis,” for a Tier 2 example). 

The available data varies regionally, but all U.S. utilities should have access to weather data and tax assessor data. The National Oceanic and Atmospheric Administration (NOAA) stores weather data which is readily available on its website. County offices generally store tax assessor data and provide access to that data for purchase.

On the other hand, such categories of data as interval usage, or building and end-use surveys, are the responsibility of the utility to collect and store. While most AMI and automatic meter reading (AMR) meters collect interval usage data, utilities need to leverage online, telephone, or hard-copy surveys to collect building and end-use data from customers. To collect more advanced interval data, utilities typically need to install more sophisticated meters.

Developing Your Capability

Figure 2 - Heat Slope Analysis: Single Customer Example

As the saying goes, “Rome wasn’t built in a day.” Developing a utility’s analytical infrastructure or the accompanying business processes also will take time.

Many in the utility industry use terms like “big data” or “cloud computing” when referring to the multitude of different ways interval usage data and other information can be used to improve utility operations. Those terms can be intimidating to someone outside of an IT department, but getting started on demand-side analytics doesn’t have to be as complicated or as expensive as it sounds.

Finding both the time and money to develop demand-side analytics is typically the largest barrier faced by utilities. To avoid running into such barriers, utilities should approach the development of analytical capabilities one step at a time, beginning with existing software and hardware, then upgrading those resources in a prudent way to meet the utility’s evolving requirements. At minimum, data can be extracted from an existing customer information system and analyzed using Microsoft Excel. By starting with something that requires negligible incremental investment, a utility can minimize financial risks and its analysts can more freely investigate the possible applications of existing data. After assessing the potential applications of existing data, the next step is to identify an analytical process to use as a proof of concept (PoC) and eventually as a proof of value (PoV) as shown in Figure 3. Proceeding in a step-by-step manner lowers the barrier to entry and allows time to develop the proper in-house expertise, including an added level of familiarity with the limitations and hurdles of big data. 

Proof of Concept

The initial PoC should be something that can be done on a small scale to show that it’s possible to use information that’s either currently available or can be obtained without major effort. Typically, this type of effort would be a process similar to the processes listed under Tier 1 and Tier 2 in Figure 1.

Figure 3 - Process for Growing DSM Analytics Capabilities

Once the PoC is complete, the analysis should be scaled up to the point where the benefits and costs can be documented and projected to a scale that covers the entire scope of that analysis for the utility. Documenting the benefits and costs is an important part of the process for developing analytical processes. Utility executives and regulators will want to know how demand-side analytics are affecting the utility in terms of overall costs and savings, improved customer experience as a result of participating in utility programs, and lessons learned from using data in a new way. Successfully communicating how analytics are either improving demand-side programs or how they’re providing insight into customers’ energy use is vital to the continued application and justification of demand-side analytics.

While utilities might want to expand their analytical capabilities to move beyond the PoV phase and implement processes beyond Tier 2, it doesn’t necessarily mean a large financial investment is required. For example, a utility could implement the open source statistical computing and data visualization software R2 in combination with the open source relational database MySQL.3 Since neither R nor MySQL have licensing fees, a utility could utilize both tools to further develop analytical capabilities without creating a sunk cost in software. In practice, however, most utilities will need some form of user support (particularly if they lack the appropriate in-house expertise), and will need to consider commercially licensed statistical software.

Getting started with analytics that have an initial focus on DSM offers many benefits. This area promises opportunities for greater energy and demand savings, improved customer engagement, and reduced costs. Examples of successful targeted marketing campaigns in the DSM community are beginning to appear, as program managers are learning to identify the customers who are most likely to participate in their programs as well as those customers who have the greatest potential for savings from their programs. DSM program managers also are starting to benefit from reduced measurement and verification costs as third-party evaluators take advantage of enhanced verification approaches enabled by interval data. The up-front investment to get started is relatively low and the potential benefits are great.

 

Endnotes:

1. “Smart Grid Data Analytics for Consumer Engagement.” Research Brief, 3Q, 2013. Navigant Research. See p.13. for market forecast of smart meter installed base by region.

2. http://cran.us.r-project.org/

3. http://www.mysql.com/