High memory usage in OpenLCA 2.4.0 when running Monte Carlo simulations via Python (olca_ipc)

Question

Hi everyone, 

I’m running Monte Carlo simulations in OpenLCA(2.4.0) using Python (olca_ipc 2.4.0, olca_schema 2.4.0), and I’m experiencing extremely high memory usage. 

Even with 200 iterations, OpenLCA memory can reach 25 GB. I also noticed that commenting out the following lines prevents memory growth:

for i in range(num_iterations):

# Uncommenting these lines causes memory to grow quickly

# result.simulate_next()

# result.wait_until_ready()

xs[i] = result.get_total_impact_value_of(impact_category_ref).amount

Questions:

1.  Does anyone know why "simulate_next( )" causes such rapid memory growth? 
2.  Is there a recommended way to run many Monte Carlo iterations in Python within openLCA while keeping memory usage manageable?

Many thanks in advance!

Siyue

Codes:

FU = 1.0 # functional unit

simulation_plan = {

"biodiesel production, from soybean": 500,

"biodiesel production, from rapeseed": 259,

}

 

impact_method = client.find(o.ImpactMethod, "IPCC 2013 GWP 100a")

impact_category_ref = o.Ref(

ref_type=o.RefType.ImpactCategory,

id='0529022f-23cd-36e5-bbba-3aff792d439c'

)

 

all_results = []

 

for system_name, num_iterations in simulation_plan.items():

print(f"\nMonte Carlo for: {system_name} ({num_iterations} iterations)")

product_system = client.find(o.ProductSystem, system_name)

 

setup = o.CalculationSetup(

target=product_system.to_ref(),

impact_method=impact_method.to_ref(),

amount=FU

)

xs = np.empty(num_iterations, dtype=float)

 

result = client.simulate(setup)

result.wait_until_ready()

 

for i in range(num_iterations):

# Uncommenting these lines causes memory to grow quickly

# result.simulate_next()

# result.wait_until_ready()

xs[i] = result.get_total_impact_value_of(impact_category_ref).amount

 

result.dispose()

 

df_partial = pd.DataFrame({

"product_system": [system_name] * num_iterations,

"iteration": np.arange(1, num_iterations + 1),

"impact": xs

})

all_results.append(df_partial)