# Randomized outage and scenario generator with variance reduction

**Introduction**

FIDES performs high-fidelity reliability analysis of power systems using Monte Carlo simulation. To increase computational efficiency and statistical coverage, it employs Latin Hypercube Sampling (LHS) to generate representative outage schedules for power system assets that deliver reduced variance converging in less number of iterations. These schedules are based on customized probability distributions for time between failures (TBF) and time to recover (TTR), capturing the stochastic nature of outages across generation, storage, and transmission components.

---

**1. Randomized Outage Schedules with LHS**

FIDES uses LHS to generate stratified samples of outage events that more efficiently explore the uncertainty space compared to purely random sampling. Each simulation iteration includes sampled values of:

* **Time Between Failures (TBF):** duration of asset uptime before failure,
* **Time To Recover (TTR):** duration of outage before restoration.

These samples are generated independently for each asset type:

* Generation units (thermal, renewable, hydro)
* Storage systems (batteries, pumped hydro)
* Transmission lines

Each asset is assigned custom distributions (e.g., exponential, Weibull) based on historical reliability data.

---

**2. Generator Outage Modeling**

For each generator an outage schedule is created by alternating TBF and TTR periods, defining a binary availability vector. This sequence is repeated across the simulation horizon for each facility.

---

**3. Storage and Transmission Outages**

For storage units and transmission lines, a similar procedure is used. The outages are characterized by durations and recovery times that impact charging/discharging ability and power flow capacity, respectively. Binary availability indicators and are applied to enforce outage states in dispatch models.

---

**4. Renewables and Demand Uncertainty**

The uncertainty of the generation levels of renewables and of demand is represented by modelling a set of scenarios representing the different values of the demand and renewable generation. Each scenario is given a probability and the results are sampled considering the given probability.

---
**5. Scenario Generation and Aggregation**

Each LHS draw corresponds to a complete system-wide outage scenario. FIDES runs hundreds of such scenarios in parallel using parallel computing. The outputs include:

* Expected Energy Not Supplied (EENS)
* Detailed system operation

---

**Conclusion**

FIDES leverages Latin Hypercube Sampling and custom outage distributions to simulate high-resolution, statistically robust reliability scenarios. Its modular and parallelizable design enables integration with detailed dispatch models and supports system adequacy assessments for diverse asset portfolios.