Model not finding optimal portfolio / Optimized vs Reported Cost Inconsistent

[grgmiller]

When running scenarios, we noticed a potential issue where it seemed like the model was not finding the optimal solution. In one scenario, the model was selecting a certain amount of geothermal, and in the second scenario, when we limited the amount of geothermal that could be built, the model was finding a portfolio that seemed to be cheaper than the original portfolio, which seemed to suggest that the model was not finding the global optimum.

However, digging into the outputs of these two scenarios, I found that the issue is not an issue with the solver finding the global optimum (phew! that would have been harder to fix), but that the costs being optimized are not the same as the costs being reported. So for the geolimit scenario, the optimized costs are actually higher than the base scenario, but the reported costs in the summary_report are lower due to how these costs are calculated.

The inconsistency is in how the congestion costs are calculated:

• In the objective function, the congestion cost reflects the difference between the DLAP price and Pnode price for all generation (Generation_t * (DLAP_t-Pnode_t))
• In reality, we only pay the DLAP price for load, so if we are over-procured, the Pnode revenue from the excess generation is not balanced by a corresponding DLAP cost. Thus, in the summary report, instead of reporting the optimized congestion cost, we report the Delivered energy cost (DLAP_t Load_t) and the Pnode revenue (Pnode_t Generation_t)

Because Pnode revenue is a negative cost, we had implemented it this way because we were concerned about the model over-procuring a resource if the average Pnode revenue was greater than the average PPA cost (thus giving the generator a net negative cost). However, we may instead just want to implement some sort of pre-check when generating the input files to make sure that the Pnode revenue wouldn't be greater than the PPA cost.

Other inconsistencies/nomenclature to note as differences between the reported costs (in "") and optimized costs (italicized):

• "Energy Contract Cost (Annual)" = GenPPACostInTP + HybridStoragePPAEnergyCostInTP
• "Storage Capacity Cost (Annual) = TotalGenCapacityCost
• "Storage Energy Arbitrage (Annual)" = StorageNodalEnergyCostInTP

Proposed Fixes:

• [x] in generate_input_files.py, implement a check that calculates the generated-weighted average Pnode revenue and PPA cost for each generator and raises an error if Pnode revenue > PPA cost. Should this prevent the inputs from being generated until fixed, or is there another check that should be implemented? Is it possible that the Pnode revenue would be greater than the PPA price on purpose?
• [x] If a generator actually has a net negative price, the model would want to build as much of it as possible, even if it leads us to be over-procured during specific times or seasons. Is there a way for us to assign a penalty value to overprocured energy to limit this? Perhaps the penalty is the uncertainty in the Pnode price so that we overprocure only if the net negative price is essentially guaranteed?
• [x] Add the DLAP load cost back into the objective function. It had been removed because it was a constant cost (assuming the total amount of load was constant), but once we implement more robust demand-side resources, this will no longer be the case.
• [x] Instead of optimizing CongestionCostInTP, include the Pnode revenue term in the objective function instead
• [x] Determine what modifications need to be made for energy storage and hybrid resources so that we are not double counting Pnode revenue from either
• [x] Update generate_report.py and summary_report.ipynb to reflect these new variables and calculations.

[grgmiller]

Updates needed to summary report:

• [x] remove references to CongestionCostInTP, Generator Delivery Cost, Congestion Cost
• [x] Add DLAPLoadCostInTP and GenPnodeRevenueInTP to cost table
• [x] Remove calculations referencing HybridStoragePPAEnergyCostInTP

[grgmiller]

Currently working to update the summary report for the "hourly_90" scenario in test_PCE model run. Before we finish updating this, we need to figure out how we are going to handle the calculation of congestion costs for storage charging.

[grgmiller]

We may also want to implement a check to see if the RA resell value would incentivize overbuilding. If the resale value for 1MW of capacity is greater than the net energy cost (PPA-Pnode), then we could have an issue

• [ ] implement overbuild validation check function in generate_inputs that checks for this.
• [ ] Also consider resale of RECs