Optimal Maintenance Scheduling Considering Uncertainties

The generating unit maintenance scheduling task often involves a number of options and faces many uncertainties. The goal is to choose the options which minimize costs and avoid high risk situations. The PowrSym Maintenance Evaluator (PME) is designed to evaluate maintenance options under uncertainty. A Monte Carlo risk model is used to work through a wide range of uncertainties and produce a probabilistic evaluation of maintenance schedules and options. The results of each evaluation are not just a single “expected” result but also a graphical depiction of the range of possible outcomes.

Some examples of the options in maintenance scheduling are:

· When to schedule a maintenance outage

· Length of outage

o Scope of work

o Staffing and overtime decisions

o Interaction with other unit maintenance schedules

· Schedule a single long outage or two shorter outages

· Delay a maintenance outage at the risk of increased forced outages

Some examples of uncertainties are:

• Unit forced outages
• Weather and load uncertainty
• Hydro uncertainty
• Renewable uncertainty (wind & solar)
• Extension of maintenance outage on this unit
• Extension of maintenance outage on other units

The PME has a spreadsheet-based graphical input for setting up or modifying planned maintenance schedules. The standard PowrSym data system is used for other data inputs. The PowrSym calculating engine is used to evaluate the schedule over a wide range of uncertainty in Monte Carlo mode, usually 100 or more iterations. The study length is usually one or two years. The outputs of each evaluation are available in printed form and also as spreadsheet readable files for producing graphical reports in the spreadsheet-based reporting module.

In a typical evaluation, there is a base case run with the current maintenance schedule followed by a series of option runs to evaluate various alternatives. The reporting module can produce total value reports, difference reports as compared to the base case, or difference reports between the alternatives for each of the output variables. The reports are in both expected value and uncertainty histogram form. The reports are available by week, month, and year.

Some of the available output variables are:

• Total system costs
• Power purchases and sales by category
• Fuel consumption by fuel type
• Emissions by type
• Generation by plant and/or plant type

The graph below shows resulting production costs for a one week extension of an outage. The y axis is percent chance of falling in that bracket and the x axis is system production cost increase in $100,000 increments. A simple computation would likely yield the $500,000 result, but the risk analysis yields an expected cost of $742,000 and some probability that costs could exceed $2 million. Similar graphs can be produced for changes in other outputs such as fuel consumption and emissions.