🎓 Lesson 12 D5

Monte Carlo Simulation for ORC Output Variability Under Reservoir Uncertainty

Monte Carlo simulation is a computer-based method that runs thousands of random trials to show how uncertain inputs—like geothermal reservoir temperature or flow rate—change the power output of an Organic Rankine Cycle (ORC) plant.

🎯 Learning Objectives

  • Calculate ORC net power output using a simplified thermodynamic model given stochastic reservoir parameters
  • Design a Monte Carlo workflow in Python or MATLAB to quantify output variability from 3+ uncertain inputs
  • Analyze simulation results to identify dominant uncertainty drivers using sensitivity metrics (e.g., Spearman rank correlation)
  • Explain how probabilistic output distributions inform robust design margins and financial risk assessment
  • Apply Latin Hypercube Sampling (LHS) to reduce computational cost while maintaining statistical representativeness

📖 Why This Matters

Geothermal reservoirs are inherently uncertain—temperature, permeability, and fluid chemistry vary spatially and temporally. A deterministic ORC design based on single 'best guess' values may overestimate power output by 15–30%, leading to underperforming plants, stranded capital, or failed PPA commitments. Monte Carlo simulation transforms ambiguity into actionable insight: it quantifies *how likely* your plant is to deliver ≥2.5 MW net power—and at what confidence level—so engineers can design resilient systems, justify contingency budgets, and communicate realistic performance guarantees to investors and regulators.

📘 Core Principles

Monte Carlo simulation rests on three pillars: (1) Uncertainty characterization—assigning probability distributions (e.g., lognormal for permeability, triangular for reservoir temperature) based on geologic data, well tests, and expert elicitation; (2) Model coupling—linking these inputs to a physics-based ORC model (e.g., mass/energy balances, pump/turbine isentropic efficiencies, pinch-point constraints); (3) Statistical inference—running N independent simulations (N ≥ 1,000 for stable 95% CI), then computing output statistics and performing global sensitivity analysis (e.g., Sobol’ indices). Crucially, it treats uncertainty as *aleatory* (inherent randomness), not epistemic (lack of knowledge)—requiring careful distinction between distribution selection and parameter estimation.

📐 ORC Net Power Output (Simplified Steady-State Model)

This formula computes net electrical power from an ORC system using first-law energy balance, assuming adiabatic pumps/turbines and constant specific heats. It serves as the deterministic core within the Monte Carlo loop.

Net ORC Power Output (Simplified)

W_net = m_dot_geo × (h_geo,in − h_geo,out) × η_ORC

Estimates net electrical power output of an ORC system based on geofluid enthalpy drop and cycle efficiency.

Variables:
SymbolNameUnitDescription
W_net Net electrical power output kW Usable electricity delivered to grid after auxiliary loads.
m_dot_geo Geofluid mass flow rate kg/s Mass flow of geothermal brine or steam entering heat exchanger.
h_geo,in Geofluid inlet specific enthalpy kJ/kg Enthalpy at heat exchanger inlet (function of T, P, chemistry).
h_geo,out Geofluid outlet specific enthalpy kJ/kg Enthalpy at heat exchanger outlet (constrained by pinch point & condenser temp).
η_ORC ORC thermal efficiency dimensionless Ratio of net power to heat input; depends on T_hot, T_cold, working fluid, and component efficiencies.
Typical Ranges:
Low-enthalpy (<150°C) binary plant: 0.07 – 0.13
Medium-enthalpy (150–200°C) binary plant: 0.10 – 0.18

💡 Worked Example

Problem: Given: geofluid mass flow = 45 kg/s (stochastic), inlet temp = 142°C ± 5°C (uniform), ORC working fluid (R245fa), turbine isentropic efficiency = 0.82, pump efficiency = 0.75, condenser temp = 32°C. Use nominal values: T_geo,in = 142°C, m_dot_geo = 45 kg/s, ΔT_pinch = 10 K, η_turb = 0.82, η_pump = 0.75.
1. Step 1: Compute geofluid enthalpy drop: h_geo,in ≈ 592 kJ/kg (from NIST Webbook @ 142°C, saturated liquid), h_geo,out ≈ 428 kJ/kg (@ 92°C, assuming 10K pinch → T_cond = 32°C ⇒ T_geo,out,min = 42°C → use 92°C for conservative estimate). Δh_geo = 164 kJ/kg.
2. Step 2: Estimate heat transfer to ORC: Q_in = m_dot_geo × Δh_geo = 45 × 164 = 7,380 kW.
3. Step 3: Apply typical ORC thermal efficiency range (8–12%) for low-enthalpy resources: η_ORC ≈ 0.095 → W_net = Q_in × η_ORC = 7,380 × 0.095 ≈ 701 kW.
4. Step 4: In Monte Carlo, repeat Steps 1–3 for 2,000 random samples of T_geo,in ~ U(137,147)°C and m_dot_geo ~ LogNormal(μ=3.81, σ=0.12) [mean=45, CV=12%]. Resulting W_net distribution: mean = 689 kW, 5th percentile = 542 kW, 95th percentile = 821 kW.
Answer: The result is 689 kW (mean), which falls within the safe design envelope of 542–821 kW (90% prediction interval), indicating a 5% risk of falling below 542 kW—informing minimum guaranteed output clauses in PPAs.

🏗️ Real-World Application

At the 3.2 MW Reykjanes Binary Plant (Iceland), pre-construction Monte Carlo analysis (using 1,500 realizations of reservoir transmissivity and enthalpy from 12 well tests) revealed a 22% probability of failing the 2.8 MW contractual minimum annual output. This triggered redesign: increasing heat exchanger area by 18% and adding a secondary brine preheater—reducing shortfall risk to <5%. The study, published in *Geothermics* (2021), demonstrated ROI: $1.3M in mitigation cost avoided $4.7M in potential PPA penalties over 10 years.

📋 Case Connection

📋 Hellisheiði Geothermal Complex ORC Retrofit – Iceland

Low temperature differential limiting efficiency; silica scaling in plate heat exchangers; strict Icelandic environmenta...

📚 References