After sustaining a record pace since March, natural gas storage injections have been slowing dramatically and are projected to fall below the 5-year-average rate over the next few weeks. While weather has factored heavily into the swing in storage activity, increased baseload demand for gas in the power sector has amplified the effects of weather anomalies and electricity demand seasonality on overall gas demand. As a result, gas demand volumes have diverged from historical levels on a temperature-adjusted basis. Today, we examine the changing historical relationships of power burn and storage injections to weather and electricity demand.
In Part 1, we started our analysis with a look at gas storage injections relative to the 5-year average and noted the large deviations this spring. Storage injections in April were 180 Bcf above the historical average for that month. The above-average injections continued in May and June but closed the gap to less than 80 Bcf above the 5-year average in May and to about 90 Bcf higher in June. The high injection rate in part can be explained away by weather, which along with record production, contributed to a particularly weak supply-demand profile in April. The weather in April warmed faster and earlier than it normally does, with national average temperatures coming in above the 10-year rolling average for most days of the month (including as much as 10 degrees above average on April 8). The result was a sharp, 17-Bcf/d month-on-month drop in residential/commercial (res/comm) demand — the 5-year-average change between March and April was about 12 Bcf/d. Power and industrial demand also fell, by 1.4 Bcf/d and 1.6 Bcf/d, respectively, for a net drop in domestic demand of about 20 Bcf/d. Then, mostly cooler-than-normal temperatures followed through May and June, when power burn for air conditioning typically ramps up, prolonging the overall bearish supply-demand balance.
But as we alluded to in the previous episode, weather alone doesn’t fully explain the above-average storage injections, since storage deviations have not only occurred on an absolute basis but also on a temperature-adjusted basis, meaning they’ve been stronger than the historical model would suggest at the same temperatures, as demonstrated in Figure 1. The blue line in the graph plots the average daily difference by storage week between our model projection for the weekly storage changes (based on historical storage activity at the same temperatures) and the estimated actual that what was reported by the Energy Information Administration (EIA). Negative deviations from the model indicate actuals were more bullish than the historical model, while positive deviations signal a more bearish trend than the historical model. As you can see, the deviations historically are volatile week to week, and a number of factors (market- and modeling-related) can cause that. However, it still provides a sense of the overall trend relative to recent years, and the abrupt increase in late March/early April (red arrow) suggests that storage injections turned distinctly more bearish on a temperature-adjusted basis than the historical model around that time. That was especially the case in April, when injections spiked to an average ~8 Bcf/d above the historical model at the same temperatures in the week of April 26. Since April, injections have remained above the model, but the deviations have trended lower, meaning reported injection rates have converged somewhat with the model relative to April.