New reports project U.S. electricity demand to grow 25% by 2030, while the IESO projects demand in the province of Ontario to grow 70% by 2050. The growth projections are driven by expected demand from AI and Crypto related data centers as well as increasing electrification. New Congressional proposals, if enacted, will dramatically cut renewable energy project subsidies and tax credits. At the same time, energy regulations seem to change daily. With this much change and uncertainty, reliable forecasts of costs, Locational Marginal Pricing, Locational Marginal Emission rates, asset valuations and derived benefits have become an increasingly problematic and risky exercise.
There are many available sources of power market forecasts. But are they reliable enough to manage the financial risk associated with investment decisions based on demand, pricing and regulatory changes. Nobody has a crystal ball that can predict the impact of every new regulation or market anomaly. However, Investment-grade forecasts are possible in each individual energy market with the curation of realistic scenarios, high-quality data, best-in-class modeling and simulation technology, and timely forecast updates.
At PowerIntel, our power industry experts create the most reliable forecasts of possible market conditions and encapsulate these projections into scenarios. A scenario is a forecast of what we think will happen during the specific time frame of concern to the stakeholder. PowerIntel scenarios include projections for fuel costs, generation mix, load growth, the impact of policy changes and other related market projections.
PowerIntel uses these scenarios to drive an expansion planning model that projects the composition of each market over the time frame of concern. Then a production cost simulation is conducted to identify the economic and environmental impact of user defined generation alternatives under each scenario. Unlike projections that are derived from forecast algorithms, PowerIntel’s forecasts include the nonlinear physical interdependency between input variables and market results.
PowerIntel’s data subscription service is designed to support evaluation of both individual alternatives and portfolios consisting of a collection of proposed generation alternatives. Results can be segmented into three basic categories with the following output categories:
Operations
-
- Generation output for both portfolio and individual generating plant
- Average zonal and nodal (Locational Marginal Prices) by total portfolio
- Capacity factor by individual generating plant
Financials
-
- Portfolio gross margin
- Cost by type for both portfolio and each portfolio generating unit
- Revenue by type for both portfolio and each portfolio generating unit
Emissions
-
- Total CO2 emissions by generating unit and total portfolio
- Carbon displacement by generating unit and portfolio
- Average Locational Marginal Emission rate by generating unit and portfolio
Cadence is Crucial
The cadence of power market forecasts is crucial. Twelve-month forecasts are no longer useful for general planning due to the pace of changes in regulations and forecasts such as weather patterns, load requirements, fuel prices, equipment issues, outages and other events. For this reason, it is necessary to have access to fresh market forecast data updated every quarter. The best quarterly forecasts will incorporate both public and private information curated from the forecasting service’s research efforts and validated by market experts and participants.
At PowerIntel, we update our market databases, scenarios and input parameters as well as rerun our expansion planning and production cost models quarterly. To ensure the fidelity of our modelling, we continuously exercise a validation process in which we compare each new quarterly forecast with the previous one and adjust our process to ensure the continuous fidelity of our models.
Getting Physical
Reliable market forecasts must be based on a physical model that reflects the impact of any changes in the physical flow of power and the requirements of forecasted load entities. This type of expansion planning forecast requires a simulation model that can create a physical representation of what the new generation mix would look like to service the forecasted load. Would it best be served by renewables, coal, combined cycle gas, nuclear or a combination thereof? Forecast models that rely on machine-learning algorithms and AI techniques might be good for short term forecasting but do not adequately capture the effects of changing market conditions. This can only be done by a physically based Expansion Planning Model.
Expansion Planning Models predict what the load and generation landscape will look like in five, ten or twenty years. This is an extremely complex and non-linear modeling exercise. It must combine the resources of the best simulation models with the subjective input of highly experienced market experts to create the most reliable results.
Reliable power cost and pricing forecasts also require a Production Cost Model that simulates the physical performance of each market. By simulating market operations against forecasted load requirements, PowerIntel produces investment grade economic and environmental results that include hub pricing, Locational Marginal Pricing (LMP) and Locational Marginal Emissions (LME). Investment-grade results for 20-plus years require a physically based simulation of each individual market on a year-by-year basis. Simply using forecast algorithms, even those utilizing machine learning and AI, will not capture the constantly changing physical nature of each market and will produce unreliable results.
Less than best-in-class forecasts can result in serious reliability and liability issues, inaccurate asset valuations, poor generation expansion and retirement planning and risky business decisions.