I’m trying to generate a random sequence of outages for a set of industrial plants.

An individual plant can be either online, or offline. I need to generate a time-series for each plant showing it’s availability (0 = offline, 1 = online).

Each plant has characteristic data giving its mean outage duration (in days, usually around 3), and the mean fraction of time it is offline in any given duration (in %, around 5%).

I have a solution using native Python, which draws a series of outage lengths and days between outages, both from geometric distributions. However, this is quite slow when generating many series. Is there a way I can use Pandas to speed this up?

mean_outage_length = 3
forced_outage_derating = 0.95
# Forced Outage Derating is percentage of time you would expect a plant to be online allowing for unplanned outages only

outage_length_mu = np.log(1 - 1 / (mean_outage_length + 1))
between_outages_probability = (1 - forced_outage_derating) * (1 / (mean_outage_length + 1)) / (forced_outage_derating)
        
if between_outages_probability >= 0.9999999999999999:
    between_outages_mu = -40
else:
    between_outages_mu = np.log(1 - between_outages_probability)

sum_of_days = 0
days_in_between_outages = []
outage_lengths = []
days_in_simulation_window = 1000        

while True:
    outage_length = math.floor(np.log(1 - random.random()) / outage_length_mu)
    days_until_next_outage = math.ceil(np.log(1 - random.random()) / between_outages_mu)
            
    if first_pass:
        first_pass = False
        if random.random() > forced_outage_derating:
            days_until_next_outage = 0

    sum_of_days += (days_until_next_outage + outage_length)

    if sum_of_days > days_in_simulation_window:
        break

    outage_lengths.append(outage_length)
    days_in_between_outages.append(days_until_next_outage)
        
    if sum_of_days - outage_length <= days_in_simulation_window:
        days_in_between_outages.append(days_until_next_outage)
        outage_lengths.append((days_in_simulation_window - (sum_of_days - outage_length)) + 1)

Expected Output:

Given a set of N plants, with each plant having a mean outage length, and expected fraction of time online, return a set of outage lengths and durations between outages.

e.g.

Input:

Simulation Duration: 100 days

Plant A: Mean outage length – 5 days, fraction of time online – 80%

Plant B: Mean outage length – 3 days, fraction of time online – 90%

Output:

Plant A: Outage Durations – [3, 7, 6, 4], Duration between outages – [15, 25, 30, 10]

Plant B: Outage Durations – [2, 5, 3], Duration between outages – [35, 15, 40]