So, I downloaded the 2025 real-time and day-ahead data and started digging. The scale is different here. I found roughly 1,200 unique transmission lines that were causing binding constraints in the grid.
Getting the raw data is the easy part. Making it usable is the real work. I had to map the constraint names to actual start and end substations and then query OpenStreetMap to extract the physical line geometries.
But geometry isn't enough. You need the physics. Since I don't have the operator's private asset database, I did the next best thing: I made my best (engineering) guess and traveled across Texas in Google Street View.
I think I did this year's vacation to Texas already by literally spending hours virtually driving down highways and dust roads to verify bundle configurations for the key corridors. I likely virtually drove past our CEO Georg Rute's home in Austin a dozen times, but I was too busy looking up at the transmission towers to stop and say hello.
It might sound a bit obsessive, but that is the only way to get the the main conductor assumptions right without insider info.
Long story short, I managed to clean the data and narrowed it down to the top 300 most critical lines by congestion cost to import into Gridraven Claw. Just like with the PJM experiment, I am now collecting the data to start running the conductor-finder brute-force tool to reverse-engineer the likely conductor types and operating temperatures that match the historical flow.
The good part? We are now effectively monitoring over $1 billion in annual congestion value continuously, with hourly updates and 60-hour forecasts. I have added a map to show all the lines we monitor in ERCOT and a sample 345kV line run by CPS Energy with forecasts.
It is nice to see what can be done with open data. Not sure if that is useful at all commercially yet, but at least it is great to see how the thermal models work and how well they correlate with actual market data
I am not sure yet if Part 4 will be about tackling another ISO or finally showing the validation results from the conductor finder, but let’s see. A lot of good ideas pending.
