Frequently Asked Questions

What is the time series granularity for the energy data in Discover?

All of the energy data in Discover is available at the monthly and hourly level.

EIA provides monthly-level energy data, how does Discover have hourly-level energy for those projects that use EIA data?

For operating projects outside of ERCOT, which rely on EIA data, Discover derives hourly energy by apportioning the monthly energy value across the hours for that month based on hourly modeled output.

How does Discover handle anomalous generation for operating projects, such as curtailment, where the project isn’t producing energy in line with weather availability?

Periods of “anomalous” generation are useful data to account for and surface because they represent real features of a project’s performance–if a wind project was shut down due to icing then this is an important event to reflect because it’s a reality the project faced. Wherever observed energy appears in Discover, that data is a direct passthrough of what was reported and will contain features like this example *. The other opportunity to reflect this event is in the model output. One way to include this feature is in the observed energy time series the model uses for training; however, including this type of anomalous generation in training will run the risk of skewing model output, so for that reason it needs to be flagged and removed from training. The other opportunity to include anomalous generation in the model output is after training in the form of availability losses, which is the method Discover uses today. Discover uses the project-specific anomalous generation to derive an availability loss percentage that is applied across all hourly model observations. This method allows this generation to still be reflected as a feature of the project’s performance, but in a way that doesn’t directly interfere with model quality.

* As mentioned in an answer to an earlier question, Discover uses monthly observed energy for projects outside of ERCOT and derives hourly energy. This means that the monthly-level energy value will reflect what was actually reported, though the hourly-level energy may differ from how the project actually performed because it’s derived.

How does Discover handle ramp periods for operating projects?

“Ramp” data refers to the period of time between a project’s first date of energy production (COD), and when the project is fully operational. This period usually lasts 1-5 months, depending on fuel type (wind/solar), nameplate capacity, and turbine/panel production delays. For wind, our wind to power model is used to determine the length of a ramp period by comparing output from our energy model to what the project produced according to ERCOT or EIA data, and flagging months where modeled energy is significantly higher than actual energy. Solar is less variable, so the ramp period is deemed over once the month-over-month variability stabilizes (seasonally adjusted). Once ramp periods are flagged, they are filtered out during model training to avoid deteriorating model output quality.

Why is my chart empty?

There can be a number of reasons why your particular chart may be empty. A few common cases are listed below:

Historical data has been released incrementally by region. The further back in history you view, there may be time periods for which data was not yet available by region. For instance:
- In 2008, MISO, PJM, SERC, NYISO and ISONE regions are released.
- In 2009, CAIDO and WECC regions are released.
- In 2010, ERCOT data is released.
- In 2013, SPP data is released.
- in 2014, NWPP data is released.
Forecasted data was released incrementally by region. If you are viewing forecasted data for a project outside a supported region, there will be no data to display. The following are the regions we currently support:
- ERCOT
- PJM
- CAISO
- SPP
- MISO
- ISONE
- NYISO

Why can’t I find my project?

Discover relies on EIA report for operational projects. This reporting can lag between 1-12 months and requires internal verification prior to it appearing in Discover. If there is a specific project you are looking for and cannot find, please email us at [email protected].

What is meant by shape value?

“Shape” captures the alignment between project hourly price and hourly generation by describing generation weighted (As-Gen) price relative to the ATC price. When shape is expressed as a percentage, it represents a ratio of the As-Gen price to the average price of power. When in units of $/MWh, shape represents the difference between the generation weighted price of power and the ATC price.

See details about these shape scalar calculations here.

Negative shape occurs when a project underproduces when prices are high while overproducing when prices are low because the generation weighted price is then lower than the ATC price. Positive shape represents higher amounts of generation while prices are low paired with lower generation during high price hours. Projects in areas with a positive correlation between price and generation reflect a positive shape value (i.e. coastal ERCOT wind projects). A shape value of zero would correspond to a project with no intermittency, producing the exact same hourly generation throughout the year (which is not common for wind and solar systems that the weather impacts). Hourly market drivers (increased renewable generation generally decreases market prices) and seasonal/diurnal correlations between average generation and the price of power affect shape as well.

How does Discover reflect the differences between weather-year price structuring and an 8760?

Discover highlights the differences between using a weather-year and an 8760 (TMY) in the Compare metric seen in the image below. The As-Gen pricing structure uses the weather-year along with concurrent prices to model future project revenue, while the 12×24 structure is the closest to a TMY. While Discover models hourly generation using weather-year data instead of 8760s, the 12×24 structure is similar to an 8760 by modeling offtake structures and masking generation intermittency slightly more than an 8760 does.

Using the compare metric in Discover, users may choose “As-Gen” and “12×24” in the price type inputs to see how different the weather-year and TMY generation actually are. The project analysis and summary stats capture the month to month volatility, and an example representing the differences between the two profiles can be seen below.

While the 8760 captures the typical seasonal and diurnal generation at the project site, there are two main problems with using it. One of these problems occurs when an 8760 is paired with non-concurrent prices, which ignores the impact of hourly wind and solar generation on market price, and can lead to project revenue overestimation in areas with high renewable penetration. Using 8760s for forward-looking price models does allow for the impact of hourly generation on market price, but the resulting prices will only be in response to the impacts of one normal weather year. This way of modeling does not account for the reality that renewable energy projects will be exposed to many different weather conditions, as well as non-typical weather years having extreme effects on power prices, as experienced by Texas in February 2021. 8760s are particularly inaccurate for variable wind patterns, but cause these revenue model errors for solar as well. Discover combats the dangers of using an 8760 by using weather condition data, pricing data, and generation data for a specific project’s timeframe and location, rather than average weather data that does not impact the market’s prices at that time.

Additionally, see REsurety’s white paper Friends Don’t Let Friends Use 8760s for more information.

What is a p-value?

P-values denote a percent likelihood of a random variable exceeding a fixed value. That random variable may be price, generation or revenue.

Examples:

A P1 price of $40 means there is a 1% likelihood of price exceeding $40.
A P99 price of $5 means there is a 99% likelihood of price exceeding $5.
A P1 generation of 100 MWh means there is a 1% likelihood of generation exceeding 100 MWh.
A P99 generation of 2 MWh means there is a 99% likelihood of generation exceeding 2 MWh.

What is the difference between average and p50?

Median, or p50, means that there is an estimated 50% chance of exceedance.

Mean is the sum of all data points, divided by the number of data points, which is the average of the data points.

What does it mean to observe a higher than average LME value for, say, a wind project?

If a wind farm has a relatively high LME value, it means that it has avoided more emissions as compared to other projects. This happens when the project generates in a location and during times with marginal generators are high-emitting. For example, this may be because of timing: the wind farm may tend to produce in hours when coal plants (which are high emitting) are on the margin. Alternatively, it may be because of location: the wind farm may be in an area where, due to transmission congestion, a local peaker becomes marginal regularly.

What does it mean to observe a lower than average LME value?

This is the opposite of a high LME value: it means that this renewable energy project has avoided fewer emissions from its generation. This is often due to transmission congestion: renewable resources behind significant transmission congestion are often curtailed, which results in low LMEs in the area (because renewables become the marginal resource). This is very evident in the panhandle of ERCOT, for example.

Notably, if load (for example, an electrolyzer or data center) is located in a location with a low LME value, it means that the load has caused (or “induced”) fewer emissions than it would have elsewhere.

Where can I find information on assumptions and other inputs applied to the Fundamentals Forecasts?

If you have access to Fundamentals Forecasts on CleanSight, you can find this information in the “Reports” tile. In this tile you will find forecast reports for both forecasted LMEs and prices, organized by ISO and the quarter when the model was updated. To see the most recent Reports, click on the “Release Date” column title twice.

Which web browsers are officially supported for CleanSight?

Google Chrome is our primary supported browser. We recommend using the latest version for the best CleanSight experience.

What about using other non-supported web browsers?

Mozilla Firefox and Microsoft Edge (latest versions) generally functional well with CleanSight. However, they are not officially supported, and you may encounter minor visual differences compared to Chrome.

Safari is not recommended with CleanSight. It has known functional and visual issues that will negatively affect your experience.

Ask a Question

If you have any questions or feedback, please contact us at [email protected].