How forecasting algorithms can include upcoming transmission outages to better predict the congestion component of locational marginal price
Introduction to Congestion
Locational Marginal Prices (LMPs) in energy markets are comprised of three components: energy (marginal system cost of producing one more unit of power), loss (cost of losses in moving the power from generation to where it’s needed) and congestion (the marginal cost of generating one more unit of power at a given node). The congestion component can be a focus for traders who manage spreads between a generator and a long-term hedge (or a load center, for utilities) and for speculative traders hoping to provide liquidity. Congestion patterns change over time as regional differences in load and generation continue to evolve, but they also change on an hourly basis as plant and transmission outages affect system constraints in Security Constrained Economic Dispatch (SCED). Savvy traders can manage those risks by incorporating transmission outages in their prediction algorithms.
How Transmission Availability Affects Prices
SCED is the process responsible for determining the optimal generation dispatch while ensuring grid reliability. A key feature of SCED is its ability to account for system constraints, which include both generation and transmission limitations. When the optimal solution is constrained by those physical limitations, the costs of those constraints are shared across the grid.
Shift factors are mathematical coefficients that quantify how changes in generation or load at a specific node impact the flow across a constrained transmission line or other infrastructure limitations. These coefficients help determine how congestion costs propagate across the system and make up the congestion component of LMPs. This means that every binding constraint in the system contributes to congestion pricing at specific nodes, shaping trading opportunities and risks.
However, predicting the impact of transmission outages on congestion is not straightforward. The direct current optimal power flow (DCOPF) solver used in SCED operates within a complex network of interdependent constraints, making it difficult to determine in advance how a given outage will affect prices. By analyzing past transmission outages and their resulting price impacts, traders can identify patterns that can be used in the future.
Key Takeaways
Security Constrained Economic Dispatch is finding the least-cost way to solve optimal power flow subject to physical constraints of the grid
When those constraints bind, the costs are expressed as a shadow price and are shared across all of the price nodes on the grid (though not evenly), called shift factors
The congestion component of LMP is the sum of all shift factors at a given price node
Do it Yourself
The following is a high-level overview of how traders can incorporate transmission outages in their machine learning pipelines.
Step 1: Preparing the Data
Historical transmission outages and binding constraints are required to link your generator or load node to upcoming transmission outages. We get the data from Yes Energy but the data is also available directly from the ISO websites or from other vendors. Transmission outages are reported with start and end dates, which must be transformed into a binary time series before they can be used programmatically. Binding constraints are reported for either day-ahead or real-time prices and in Yes Energy are already cleaned in a tidy format.
Step 2: Building an ISO-wide dataset
All reported outages (both planned and forced) can be queried alongside the binding constraints in a given ISO. Each series gets a column and users can decide how far back they want historical data, with the obvious trade-off that more data can cause significant memory challenges.
System Operator | Active Constraints | Outages |
NEISO | 404 | 1,745 |
NYISO | 271 | 13,423 |
PJMISO | 4,813 | 109,253 |
MISO | 7,665 | 133,208 |
ERCOT | 4,576 | 114,457 |
SPPISO | 9,310 | 25,925 |
CAISO | 5,744 | 15,811 |
The number of active constraints and outages reviewed by Enertel in our ISO-wide machine learning models.
Step 3: Putting it all together
Once you have a dataset of transmission outages and historical constraints, you can use various clustering algorithms like Euclidean distance, correlation or others to measure how similar a given binding constraint is to all of the reported transmission outages in the ISO. You can also work backwards and programmatically evaluate how similar an upcoming outage is to the binding constraints that have happened in the past.
Both methods require understanding how a given constraint dissipates to each of the price nodes in the system, or calculating their shift factor, which can be done with power flow analysis or with the help of a vendor.
Conclusion
Transmission outages play a critical role in shaping congestion patterns, influencing how traders manage hedges and speculate on locational price spreads. By incorporating outage data into predictive models, market participants can improve the accuracy of congestion forecasts and make better-informed decisions.
Enertel AI provides short-term energy and ancillary price forecasts for utilities, independent power producers (IPPs), and asset developers to inform trading strategies. You can view our data catalog, view a clickthrough demo of our product for operators, or request a sample of backcast data for your assets.
The data in this blog post was provided by Yes Energy.
If the methodology proposed in this post could be useful to you, don’t be afraid to reach out!
Comments