Scenario settings
This is a comprehensive overview of all available Scenario settings on Tensor Cloud. Although the number of Scenario settings is smaller than those of Assets or PPAs, there is considerable complexity in the time-based manner in which Scenario assumptions are made.
For an introduction to what Scenarios are, see our introduction to Scenarios.
Scenario name and description
In the Scenario settings dialog you can press the edit button next to the name of the Scenario to edit its name, and type into the textbox below to add a description. The length of the description is limited to 400 characters.
You cannot edit the name or description of the Tensor Baseline default Scenario. After creating your workspace, make sure to add a new Scenario that fits the needs of your organization and those of your investors.
Scenario Parameters
Each Scenario has parameters in three major categories: Macroeconomics, electricity price, and operations.
Macroeconomic Parameters
Inflation Rate
This is the inflation rate of the overall economy. Currently, the inflation rate is only used for determining the price of PPAs with inflation escalation. Going forward, Tensor Cloud might leverage your inflation assumption for determining future electricity prices, or for discounting cash flows.
Currently, Tensor Cloud provides the Japanese Cabinet Office forecast until 2026 as a built-in assumption. Note that since the time horizon of the Cabinet Office data only extends until 2026, inflation rates past 2026 are assumed to be flat.
Electricity Price Parameters
Your assumptions about future electricity prices are the most important value driver of assets outside of fixed-price PPAs or FIT subsidy. Trends of both average system price and price mean average deviation are modeled as flat in the Tensor Baseline Scenario.
Here is how Tensor Cloud calculates future prices (forward price curve) across grid zones and 30-minute time slots:
- We start by ingesting past actual data of the JEPX day-ahead and intraday prices from April 2016 to today
- We then disaggregate the past data into its daily, weekly and yearly seasonality, and its overall trend since 2016 which we remove
- Using the disaggregated past seasonality as training data, we train a machine learning model to predict future seasonality but without a trend
- Lastly, we apply the predicted future seasonality to your future average system price and price mean average deviation assumptions to arrive at a blended forward price curve
The above approach shields you from the complexity of having to predict future daily, weekly or seasonal price movements at 30-minute resolution, and allows you to focus on long-term trends in electricity prices and price volatility instead. We have purposely designed our price prediction model this way, as it mirrors a conventional techno-economic modeling process built around rough yearly estimates.
Tensor Cloud continuously updates it's forecasting models based on the latest market data. Also, predictions get replaced by actuals as they become available. That means you might see slightly different results when re-running an asset simulation. You can read more about the non-deterministic way of asset simulation on Tensor Cloud here.
Average system price
Future electricity prices are expressed as annual price averages on the day-ahead markets in Japanese Yen per kWh. Similar to other parameters, you build your assumption by specifying the average annual price for each grid area for any year until 2066.
Tensor Cloud does not require you to build separate assumptions for the intraday market as there is a strong and predictable correlation between intraday and day-ahead prices. Tensor Cloud infers this correlation from past market data and automatically models the intraday market at 30-minute resolution based on your chosen day-ahead assumption.
Japanese intraday trading volumes are still far behind those on the day-ahead market, especially when compared to European countries. As trading volume on the JEPX intraday market grows under the new market regime, we will revisit scenario parameters and provide more flexibility around creating custom intraday price assumptions.
You will not be able to re-define or enrich past actual data with custom values. Here, Tensor Cloud treats the past as fixed. For example, if you were to create a new assumption today (2023), the average electricity price for 2022 would be the first value in your list and you would not be able to edit or delete it, because 2022 has already happened.
As with other parameters, your average system price assumptions will benefit from Automatic Interpolation and Flat Extrapolation.
Mean average deviation
Tensor Cloud allows you to model price volatility in each grid area by entering the annual mean absolute deviation of day-ahead electricity prices in Japanese Yen per kWh. Similar to the average system price, the intraday mean average deviation refers to the JEPX system day-ahead prices.
To better understand how Tensor Cloud uses your assumptions, let's calculate backwards. To arrive at the mean average deviation we would do the following:
- We start with the arithmetic mean of the JEPX day-ahead system price for a given fiscal year, considering all 30-minute time slots
- For each 30-minute time slot over that same fiscal year, we then subtract the price for that time slot from the arithmetic mean calculated during step 1
- Lastly, we calculate the arithmetic mean of the absolute values of the differences calculated during step 2 and round the result to 2 digits
Operational Parameters
Curtailment
Besides electricity prices, your curtailment Assumption will in many cases have the greatest impact on the financial performance of your assets. This Parameter is expressed as percentage of annual generation being curtailed.
Location-specific prediction of curtailment has been a notoriously challenging task which is why we believe that until there is more publicly available hyper-local curtailment data, annualized average curtailment makes the most sense for asset planning.
Annual curtailment values cannot be set to higher than 65%. We have set this limit to avoid distortions in the simulation results. Contact us if you have a need for assuming higher annual curtailment.
In the future, we plan to provide an option to model out curtailment in each grid zone in a more bottom-up fashion, for example, by allowing you to specify future energy storage capacity and power mix to automatically derive implied curtailment rates.
If you are already operating your own curtailment model or have purchased curtailment forecasts at sub-annual granularity from a third-party data provider or think-tank, and would like to use this data on Tensor Cloud, contact us.
Forecast MAPE
Part of solar PV asset operation is creating a daily forecast of the next day's generation and submitting this forecast to the grid operator. Any difference between the forecast and the actual electricity generation will have to be compensated by trading on the intraday market and then paying imbalance charges for any remaining difference.
These two parameters allow you to specify how accurate these generation forecasts are.
Deleting a Scenario
To permanently delete a Scenario, press the Delete Scenario button at the bottom of the Scenario settings dialog. Note that you cannot delete the Tensor Baseline Scenario.
Since deleting Scenarios can have wide-ranging consequences for your assets, you only delete a Scenario after it has been removed from all your assets.
Assumptions
As described above, each Scenario on Tensor Cloud has multiple Parameters such as inflation rate or electricity price. Of course these Parameters will have to change over time to reflect changes in electricity markets or the wider economy. To achieve this in a spreadsheet financial model, you would simply fill in a different value for each year of your model. Similarly, Tensor Cloud allows you to easily define how Scenario Parameters change over time.
An assumption on Tensor Cloud is simply a Parameter changing over time in a specific way. While each Parameter of a Scenario can have only one active assumption, you can create as many assumption as you like, re-use them in other scenarios, or use our library of built-in assumptions when creating your scenario. An assumption can have a name, can be saved, edited, and deleted.
For example, the data points below could be a possible assumption for the Inflation Rate Parameter:
| Year | Inflation rate |
|---|---|
| 2022 | 1.6% |
| 2023 | 0.9% |
| 2024 | 0.6% |
| 2025 | 0.6% |
| 2026 | 0.4% |
Since manually entering values for each year (and each grid zone for some Parameters) can be tedious and error prone, Tensor Cloud has two features that make creating assumptions easier: Automatic Interpolation and Flat Extrapolation.
Automatic Interpolation
Tensor Cloud will automatically fill in values for missing years. For example, let's say your assumption only contains two values: 1% for 2022 and 3% for 2024. In this case, Tensor Cloud will do a linear interpolation between the two values and create the following:
| Year | Value |
|---|---|
| 2022 | 1.0% |
| 2023 | 2.0% < interpolated value |
| 2024 | 3.0% |
Tensor Cloud currently only supports linear interpolation. We are evaluating other interpolation methods that would make it easier to model things like exponential growth of a Parameter.
Flat Extrapolation
Looking at the above example, what would be the values for the years after 2024? Here is where Flat Extrapolation comes in: Tensor Cloud simply takes the value of the last year and applies the same value for all years until the last year of the simulation.
The final result from above example would look like this:
| Year | Value |
|---|---|
| 2022 | 1.0% |
| 2023 | 2.0% < interpolated value |
| 2024 | 3.0% |
| ... | 3.0% < extrapolated values |
| 2066 | 3.0% < extrapolated value |
Currently, the last possible year for assumptions on Tensor Cloud is 2066. We will periodically review this and extend it into the future as-needed. Contact us if you need your assumptions to stretch beyond 2066.
How to use assumptions
You can access the list of assumptions for each Scenario parameter from the drop-down selectors on the right side of the screen. If the assumption currently selected is editable, an Edit button will appear next to the drop-down selector; if it is a Tensor-provided assumption, you will be able to view the assumption's details, but unable to edit them. Clicking the Edit button will take you to the assumption Builder
Each drop-down selector contains an option at the bottom to add your own custom assumption.
Assumption Builder
Similar to how you would build an assumption in Excel, for example, by entering values for your assumed inflation rate for each month or year of your financial model, Tensor Cloud allows you rapidly build assumptions by entering yearly values. This is done in the Assumption Builder.
How to name Assumptions
The first step when creating a new assumption from scratch is to give it a name, by pressing the Edit button next to the default name at the top of the Assumption Builder dialog.
As assumptions are shared between all members of an workspace, we recommend to agree with your colleagues on a consistent naming scheme early on. Prefixes or postfixes can work well to keep the team aligned around a common standard. For example, experiment/NAME, draft/NAME, or approved/NAME.
Adding values to assumptions
To add a new value, first choose the desired year from the year drop-down field, then enter a value and press the Add button or press enter. You can enter values in any order, Tensor Cloud will automatically sort them by year for you.
You can delete values by clicking the delete icon on the right of each line.
Instead of adding values for each year by hand, we recommend identifying major future inflection points in your model, and using Automatic Interpolation and Flat Extrapolation to fill in the blanks between. Also keep an eye on the Timeline Preview to understand your assumption over time.
Assumption starting values
For any new assumptions, the Assumption Builder will already contain values for the last year for which we have actual data. For example, when building a new assumption for solar PV curtailment, actual curtailment data from the last calendar year will be pre-populated for each grid area.
As Tensor Cloud uses the latest actual market and grid data when running asset simulations, you won't be able to edit or delete actual values from assumptions, you can only add new values for any year in the future. Note that the Assumption Builder works in Japanese fiscal years.
Grid Zones
Some Parameters such as electricity prices and curtailment are grid area specific. For these parameters, the Assumption Builder allows you to add, edit and delete values for each area individually.
Select your grid zone using the Grid Zone drop-down selector at the top of the Assumption Builder and enter your values normally. By pressing Apply to all button, you can copy the values from one grid area to all the others.
Use the Apply to all button with caution, as you might inadvertently overwrite values for grid areas you have spent time on building up.
Timeline Preview
The right side of the Assumption Builder shows you a line-chart that visualizes how the values you have entered will look like over time from today to 2066, the current maximum end time of simulations on Tensor Cloud.
Hover your mouse over the line to better understand the results of Automatic Interpolation and Flat Extrapolation over the values you have entered.