“Prediction is hard, especially about the future.”

- Yogi Berra

Many Gridium customers are dealing with mandates to improve their energy budgets, in some case meeting demands of 2% accuracy. Of course, improved forecasting makes good financial sense, but our analysis of weather patterns and tariff structures suggests teams should plan for even greater variance in energy spend in the years ahead. In other words, there’s a good chance your 2013 energy budget is already wrong.

Variance in energy spend broadly comes from two sources: changes to energy costs and changes to energy use patterns. For most customers, energy costs are driven by tariffed rates, which change throughout the year. In California, for example, both PG&E and SCE filed for rate increases of about 5% in June 2012. Customers who buy electricity from third-party providers can face even greater price uncertainty, depending on how much market volatility they are exposed to through their purchasing agreements. We’ll delve more deeply into the details of price volatility in a future post, but the basic outlines of the story should be familiar to any energy professional. Price risk has been an essential feature of energy purchasing since Thomas Edison’s Pearl Street station opened.

Energy use changes in response to external factors, which primarily means weather, as well as internal drivers such as occupancy, data center load, and other sources of demand. Some variance is also baked into the calendar itself, with billing periods falling across different patterns of weekends and weekdays from year to year.

The chart below shows variance for a typical Silicon Valley building by day type and by season. The effects of weather on energy use can be clearly seen.

Energy demand (kW) varies with season and type of day. The red line indicates the mean and the grey boundary the range of values experienced during one year.

Price changes interact with energy use patterns to drive even greater variance in your bills. Remember our friend the demand charge? Commercial energy is tariffed on both use (kWh) and demand (kW). For most buildings, demand charges are the ticking time bomb in your energy budget. Recall that demand is tariffed at the highest reading each billing month, and can drive 40% of the bill during summer months.

Whether that really hot day falls on a weekend or weekday can have a profound impact on summer time bills. The following chart shows the effect of ten years of simulated weather on the distribution of demand charges. The distribution follows a bell curve, and 95% of the time, demand charges will be within 20% of the mean. That is, if your typical demand charge is $10,000, then 95% of the time if will fall between $8,000 and $12,000.

Let’s look at a simulation of how demand charges vary for a building over a 10 year period of weather history. We’ve normalized this to percentile readings to show how the difference in demand charges:

Demand charge (kW) variance by simulation of 10 year weather.

The simulation shows that about 95% of the time, demand varies by under 20%. Put more simply, if your demand charge is budgeted for $10,000 a month, about a third of the time it will be $11,000 and about 15% of the time it will be $12,000. Put more simply still, for most buildings, a target of 2% forecasting accuracy is a pipe dream. Essentially, you’re trying to predict the weather twelve months in advance.

Demand charges are the ticking time bomb in your energy budget

And, as we’re all becomingly increasingly aware, weather can be extreme. Check out the tails of the demand distribution. There are a few unlucky days, including two days in which the demand charge was 90% above normal, or $19,000 in our example. Overall, about 4% of the months have demand charges more than 20% out of expectation, which means you should expect to see a large swing at least once every two years. In poker terms, the odds of demand being 20% above expectation are about the same as your odds of drawing a three of a kind.

Clearly, energy budgeting presents a challenge. Here is some practical advice to help you through the process:

  • Model uncertainty. You may need to settle on some final numbers for planning purposes, but savvy facility professionals communicate uncertainty. Share this post, use the poker analogy, communicate your inability to control weather. A good forecast show the range of likely outcomes.
  • Build in spending buffers. Variance operates on the law of large numbers. Overages in spend should eventually be matched by underages. Create a buffer account in your budget to handle weather-related variance.
  • Manage demand. Energy demand is far more volatile than energy use. Manage your demand charges by understanding when they come and planning ahead to minimize impact. Gridium’s Snapmeter tool can help you with this with payback under a month.