ElectricUtility dispatch for Outage Scenarios Looks Inaccurate

[boxpower]

Modeled a grid-tied microgrid with PV, ElectricStorage and Generator via the reOpt API. The input for "ElectricUtility" included an outage of 4000 hours starting at a time step of 3077.

When looking at the "ElectricUtility" object in the "outputs" section, I expected to see a huge block of zeroes in the "electric_to_load_series_kw" array due to the fact that I am modeling an outage during that time and hence, the grid should not be supplying electricity to the load during that outage period.

However, the array doesn't seem to indicate any outage period at all. Is this a change that can be made to the array to make it easier to visualize the outage periods? Thanks

Attaching the inputs to the reOpt API and the output json:
outputs.json
inputs.json

[kurt-verdant]

I'm just a user so take my input with a grain of salt. I arrived at the conclusion that the "electric_to_load_series_kw" array in the "ElectricUtility" object in the "outputs" section does not contain information about the outage.

The numerous arrays in the "Outages" object in the "outputs" section have length equal to the outage length. For example, the last ten values in "pv_curtailed_series_kw" in the "Outages" object are:

If you want an 8760 that includes the results found in the "Outages" object, I believe you need to combine information from that object with the non-outage subset of the "electric_to_load_series_kw" array (and likely others) in the "ElectricUtility" object (and likely others) in the "outputs" section.

There are numerous series in the "Outages" object:

In short, I believe that when outages are involved a fair bit of wrangling is typically required to get results combined into a single, 8760-row, rectangular, table. If I'm incorrect I hope someone with more knowledge than me will chime in.

[adfarth]

Great question! In REopt, there are two distinct ways you can model outages:

1. Using the "Single Outage Modeling Inputs": outage_start_time_step and outage_end_time_step.

• When you use these input, the grid purchased electricity is set to zero during the outage timesteps, as you were expecting to see in the results. The financial metrics also assume that there is no grid-purchased electricity during those timesteps. The critical load is also spliced into the load profile during the outage duration.

2. Using "Multiple, Stochastic Outage Modeling Inputs": ElectricUtility | outage_start_time_steps, outage_durations, and outage_probabilities, Financial | value_of_lost_load_per_kwh, and Site | min_resil_time_steps.

• When you use these inputs, the outages are modeled but are not spliced into the electric load. As @kurt-verdant mentioned, the outage dispatch series are available under the Outages results key.
• With these inputs, the model will choose to meet the critical loads only as cost-optimal. This trade-off depends on cost of not meeting load (see Financial | value_of_lost_load_per_kwh) and the costs of meeting load, such as microgrid upgrade cost (see Financial | microgrid_upgrade_cost_fraction), fuel costs, and additional DER capacity. To ensure that REopt recommends a system that can meet critical loads during a defined outage period, specify this duration using Site | min_resil_time_steps.
• When using multiple outage modeling, the expected outage cost will be included in the net present value and lifecycle cost calculations (for both the BAU and optimized case). You can set Financial | value_of_lost_load_per_kwh to 0 to ignore these costs. However, doing so will remove incentive for the model to meet critical loads during outages, and you should therefore consider also specifying Site | min_resil_time_steps. You can alternatively post-process results to remove lifecycle_outage_cost from the NPV and LCCs.

[boxpower]

Thank you so much for the detailed response!

Quick question about "Multiple, Stochastic Outage Modeling Inputs":
I run a simulation with two outage time_steps and corresponding outage_durations.
"ElectricUtility": { "outage_start_time_steps": [5, 3077], "outage_durations": [24, 4000], "outage_probabilities": [0.5, 0.5] }

When reviewing the critical_loads_per_outage_series_kw array in the Outages object, I see two arrays, which I initially assumed would be showing the critical loads for each of the outage scenarios I simulated.
However, each of those arrays contain two arrays:
"critical_loads_per_outage_series_kw": [ [ [1,2 ... 8759, 8760],[1,2 ... 8759, 8760] ], [ [1,2 ... 8759, 8760],[1,2 ... 8759, 8760] ]]
Not sure how to interpret this array structure.
Does each array in the "critical_loads_per_outage_series_kw" array represent the two different probabilities in the "outage_probabilities" input and then the critical load used to model each outage time_step for a given probability?
There's a similar array structure for the PV, Storage, Generator etc dispatch arrays in the Outages output.
Can you please clarify what "scenario" each of the arrays represent, so that I can accurately reconcile them with the rest of the outputs?

Thanks!

outputs.json
inputs.json

[adfarth]

The inputs you've supplied will result in REopt modeling both a 24 and 4,000 timestep (e.g. hour) outage occurring at both time_step 5 and 3077, both with 50% probability.

All of the outage timeseries results are 3-dimensional arrays with dimensions corresponding to the outage durations, the outage start time, and the time step in the outage. For example, results["Outages"]["critical_loads_per_outage_series_kw"][s,t,ts] is the critical load in outage scenario (duration) s, starting at time step t, at time step ts in that outage. To access the timeseries for the outage of duration 24, starting at timestep 3077, you'd use results["Outages"]["critical_loads_per_outage_series_kw"][1,2,:]. (We will update the REopt.jl documentation to make this more clear).

Note that arrays all extend to the outage of the longest duration.

[boxpower]

Thank you so much for the clarification!