Extracting Customer Value from Your Energy Forecast

By Bob Champagne

As customers try harder and harder to squeeze more and more from the energy dollar, the industry is being bombarded with new technology solutions. From distributed technologies like solar and wind, to a myriad of end use technologies and IOT integrations that are helping customers manage their demand. As an industry, most of our technology dollar is either spent on alternative sources of supply or changing the way our customers use energy.

But what about the vast majority of our energy that is still provided and delivered through conventional means? There are energy technology companies that believe that little is being done to address the 50-80% of our energy bill that is being spent procuring and delivering electrons in its most conventional sense. The industry may have lost its focus on energy, its core product – what drives it, why it costs what it does, and what can be done to lower its price.

The value of a good forecast

The energy industry has spent millions if not billions on customer data – acquiring it, processing it, analyzing it, and leveraging it across its customer service processes. Yet, still very little is known about when and how our customers use energy, and what they are likely to do in the future.
For most industries, especially ones in highly competitive climates, such data is the lifeblood of their business. Hotels live and die by their ability to predict occupancy levels. Same with airlines, although some appear to be learning slower than others. And for today’s digital communications companies, the ability to predict usage data and balance it across their networks is becoming central to their P&L. In most competitive industries, cut-throat price competition requires that businesses avoid idle capacity and keep their production as close to capacity while minimizing the risk of being short.

As utilities become more competitive and the pace of change accelerates, one would expect the same level of focus to be placed on precision forecasting as a mission critical priority. But the reality is that the movement here has been surprisingly slow.

At a high level, utilities have done an effective job at forecasting our peak loads at a reasonably good level and ensuring continuity of supply. And in large measure, our regulators and utilities remain adequately insulated from large price fluctuations during periods of uncertainty.  With the exception of some rare but notable events like polar vortex, continuity of energy supply has been reasonably good, and prices have been to a large extent predictable.

Whilst we are reasonably insulated from major price swings and supply shortfalls, the cost of that assurance is substantial. Today, the risk of forecasting error is mitigated through financial hedges and other types of “supply insurance” they procure from customers in the form curtailment and demand response resource- essentially paying for the right to interrupt supply during a shortfall. In regulated markets, the cost of risk mitigation is passed back to customers in the form of rate hikes or overly generous fuel adjustment mechanisms. In some cases, even penalties and price premiums caused by failed sufficiency tests can even be passed onto customers.

To a large degree, this cost has become an accepted reality, and customers and regulators have been comfortable (either out of ignorance or tolerance) paying a high premium to avoid that risk. But as price pressures mount, and markets continue to open to competitive forces, this will likely be a key battleground for successful competitors.

The trend toward bottoms up forecasting

Several leading energy companies are radically changing their approach to energy forecasting – one that combines bottoms up load analytics with machine learning technology to eliminate large amounts of uncertainty in their energy forecast. Rather than building forecasts from the top down using system-level production data, these organizations are building their forecasts from the customer upward, which allows the organization to integrate thousands of variables to better understand what’s driving consumption and why.

Most of these organizations have seen significant reductions in their energy forecasting, particularly in periods that are highly exposed to price volatility. Average reductions in MAPE (Mean Absolute Percent Error), an industry accepted measure of forecast variance, have ranged from 20-40%. That’s a massive improvement and one that can lead directly to lower energy purchasing and hedging costs which in turn spells lower rates for customers and higher margins for the utility.

Perhaps more importantly, these organizations now have a level of transparency into their customer consumption patterns that was previously unattainable. For example, by understanding specific customer load patterns, and what’s driving them, the utility is able to target products and programs that would be most beneficial. Prices can be reconfigured to better match the customers’ cost to serve and impact on margins. Forecasts can be used to predict customer bills on the fly and communicate with the customer to help avoid bill shocks. Consumption patterns can also provide useful insights to help inform customer acquisition and retention efforts.

Forecasting data can also be used by asset managers to improve overall infrastructure reliability. The ability to develop and monitor energy forecasting at the feeder or individual asset level can help inform both operating decisions and investment considerations.

Making the transition

As companies make the transition from top down to bottoms up energy forecasting, three common characteristics are emerging,

DATA ACQUISITION AND CONDITIONING

Most companies that are applying bottoms up forecasting consider access to customer consumption data to be key to their success. A seamless connection to customer consumption, either through internal MDM environments or LDC interfaces can significantly improve chances of success and increase efficiency. Also important is access to interval consumption data. While full deployment of smart meters is not necessary to implement bottoms up forecasting, it often helps to have a sample of data on which to establish and test initial forecasts and build proxy
signatures where appropriate.

DISAGGREGATION AND PREDICTIVE MODELING

Key to bottoms up load forecasting is a quick and cost effective mechanism for load disaggregation of demand and clear transparency into the drivers and sensitivities involved. While there are multiple approaches to load disaggregation and analysis, they vary considerably in terms of cost, complexity, scalability and the value produced for the business. For example, core analytics platforms provide multi-level load disaggregation analytics through a simple SaaS based interface without the need for any end use or sub-metering equipment.

ANALYTIC CONSISTENCY/ SCALABILITY

Most of today’s organizations have multiple approaches to energy forecasting whether they know it or not. These different and often conflicting approaches usually emerge inside of business functions to solve specific business problems that the top down forecasts simply can’t support. Shifting to bottom-up opens new possibilities because of the ease through which it can be re aggregated and adapted for different purposes. Approaches like that provide one integrated source forecasts that utilize the same data and standards regardless of the downstream application for the forecast. That means fewer inconsistencies and conflicts, significant savings produced through eliminating redundancy, and increased confidence in the overall forecast.

Capitalizing on success

The true test of an energy forecast lies in its ability to generate results that are sustainable and drive value in the business. Finding and exploiting new uses for the energy forecast is becoming a hallmark of many innovators across the industry. The savings produced in the trading and risk areas are just the tip of the iceberg. The real innovations are emerging around the core in areas like pricing, product marketing and even customer engagement which all can extract immediate value from a true bottoms up forecast.

As we navigate deeper into the journey of predictive energy analytics, and its downstream applications through AI and other emerging technologies, it’s important that we not lose sight of the value that is right under our noses.

There is enormous value in the data we have today if we are smart enough to use it. But that will require changes in our perspectives and business processes to harness and capture those new sources of value. The possibilities are truly endless.

______________________________________________________________________________

Bob Champagne is the CMO & SVP of Innowatts