Power plant owners are legally obligated to report planned and unplanned power plant outages under European REMIT regulation. Collection and interpretation of this data has become critical for power market analysis. As plant owners have little visibility on mid- and long-term future outages (obviously they cannot already announce future unplanned outages), the value of this data as a forecast of power plant availability can be challenged. When faced with such incomplete data, how can we still use historic and scheduled plant availability curves to get an understanding of the future availability of supply?
Historic Actuals curves account for both planned and unplanned events, whereas forward-looking Schedule curves only include planned events (with the exception of the on-going unplanned events that started before the first point of the Schedule curve). The impact of unplanned events is thus not reflected in the Schedule curves, leading to an overestimation of the implied power plant availability forecast. The quality of Schedule curves decreases the further you go out along the curve, as plant owners report fewer and fewer outages through their REMIT platforms (with Schedule availability generally even reaching 100 percent), see Figure 1.
To analyse and illustrate some of the issues that arise, we consider the example of French nuclear. This has been chosen due to the key role played by the technology in this market and the very strong seasonality in maintenance profiles.
The actual availability of French nuclear generally fluctuates between 60 and 95 percent (see Figure 2). Maintenance most often takes place during summer time, as EDF aims to keep nuclear availability at the highest possible levels during the winter months when power demand is highest. A quick look at the currently scheduled power plant availability for the coming years shows a curve fluctuating at a higher level, between 70 and 100 percent. Due to the lack of unplanned power plant outages in the Schedule, as well as yet-to-be-planned scheduled maintenance, the predictive quality of the Schedule curve is clearly decreasing over time.
In order to assess the quality of the Schedule availability curve, we can compare the Schedule curve generated on 1 January 2016 (incorporating with all the known events for 2016 as of 1 January 2016) with the Actual availability in 2016, see Figure 3.
For the first three to four months, the Schedule curve of 1 January is an accurate forecast of the upcoming nuclear availability. Then, we see the quality progressively decrease as unplanned events (1 211 messages about unplanned events affecting French nuclear reactors were published by RTE in 2016, see Figure 1) and newly scheduled maintenance are announced. In Figure 4, we annotate this evolving relationship between the two curves as it develops over time.
Despite this divergence, the Schedule curve contains very interesting information on weekly variations. Most of the signification shifts in nuclear availability, which took place in 2016 (such as those seen in early July and at the end of August), were already captured in the Schedule curve generated six months earlier.
For longer-term analysis, the Schedule curve needs to be enriched by a statistical analysis of the unplanned events to quantify their contribution and to correct the Schedule power plant availability curve for this missing component. Figure 5 shows the performance of such an approach (revised Schedule in blue vs. actuals in black).
Statistical indicators have been computed on historical availability curves. They result in offsets varying with maturity, applied to the Schedule availability curve to enrich it, thus improving forecast quality. It is clear that such a combination of planned events with historical data can lead to a reliable forecast of actual out-turn availability.
Reported plant availability is critical for power market analysis. Being able to collect, clean, map, and interpret outage messages requires a combination of strong IT capabilities, proven market expertise, and rigorous quality control procedures. However, these are not the only challenges that analysts face. To achieve a good availability forecast, historical plant outage messages and reported outages must be combined in clever ways.
Gain unparalleled insights into European power plant availability with Genscape's PowerRT and EPSI Platforms to understand and identify patterns and trends to strengthen your power market analysis. To learn more, visit the PowerRT and EPSI webpages.