Static budget-headroom plans
The static-plan comparison asks what happens when an agency cannot re-plan every year. It evaluates existing frontier plans against a common actual annual budget and the same annual cost, productive-hours, and capacity factors used by the rolling-policy experiment.
This is candidate selection from the existing frontier, not robust optimization. The candidate’s nominal budget is the budget used to build its treatment schedule; the actual budget is the common 60,000 CAD/year constraint against which all candidates are tested.
Network result
8,737 ha, 59 segments, 2027-2031.
20 annual-factor episodes, seed 7.
Actual budget 60,000 CAD/year; actual capacity 6,000 hours/year.
Budget-risk limit for the starred candidate: 5%.
Candidate |
Nominal budget |
5-year cost P10-P90 |
Mean risk |
Budget violation |
|---|---|---|---|---|
milp-budget-20000 |
20,000 |
88,293-110,903 |
0.1741 |
0.0% |
milp-budget-30000 * |
30,000 |
132,246-166,121 |
0.1009 |
0.0% |
milp-budget-45000 |
45,000 |
198,832-249,729 |
0.0526 |
10.0% |
milp-budget-60000 |
60,000 |
265,469-333,438 |
0.0294 |
95.0% |
milp-budget-80000 |
80,000 |
353,128-443,639 |
0.0142 |
100.0% |
milp-budget-100000 |
100,000 |
442,405-555,672 |
0.0093 |
100.0% |
The 30,000 CAD target is the lowest-risk candidate within the 5% budget-risk limit. The 45,000 CAD target has lower risk but breaches the actual budget in about 10% of episodes. The 60,000 CAD balanced plan has lower risk still but breaches in 95% of episodes because it commits a schedule that uses essentially the whole nominal budget.
Decision trade
The static candidates leave budget headroom by building to a lower nominal budget:
30,000 target: 132,246-166,121 CAD five-year cost, risk 0.1009, no budget violation.
45,000 target: 198,832-249,729 CAD, risk 0.0526, 10% budget violation.
Rolling policy: 298,123-299,398 CAD, risk 0.0296, no budget violation.
Rolling gives much lower risk than the static candidates at a higher treatment cost. The static 30,000 target is the low-cost option, not the risk-minimizing option.
Limits
Candidate evaluation, not robust optimization or a probabilistic guarantee.
The actual budget is fixed at 60,000 CAD/year for this comparison.
Twenty episodes and independent annual factors are a demonstration sample.
The MILP is exact only for the declared linear risk state at 59 segments.
No field calibration, outage constraint, safety certification, or asset-owner approval.