Power grid outage prediction and duration estimation

Hi everyone, 

I would appreciate if someone who can offer help on the following:

Is there any specific AI tool (preferably free or trial version) from the companies or someone else that can be used for grid power outage prediction and duration estimation? On my estate, there is nuisance power trips from the grid, and standby generators crank to supply the power to the estate. Again, generators stop when the grid supplies the power after running few minutes at the moment. I am looking to get the best average duration even if grid power supply is back to keep generators running. This will help avoid parts such as MCCB, motor mechanism and others of generators failure too frequently. It would also save carbon footprints etc. 

It can be based on past trends such as outage time, duration and months etc

Parents
  • I'm not aware of an AI tool but if you're looking for duration might not traditional statistical analysis be sufficient for this purpose? Assuming you have the historic data, at least in the first instance you can probably get "good enough" information from a few hours playing with the numbers. Personally I would trust this more than an AI anyway.

    If you don't have the data then I would start there. Neither traditional approaches nor AI will give you an accurate insight without it; outages will be entirely dependent on site-specific factors. If you don't already have something as well as making a point to note start/stop times it might be that your network operator is able to provide information, or you might be able to piece it together from metering information, relay timestamped info, server logs etc.

    It is worth noting in any analysis that you are probably experiencing several different types of outage, from transient dips to planned maintenance work, which have independent causality, frequency and duration.

Reply
  • I'm not aware of an AI tool but if you're looking for duration might not traditional statistical analysis be sufficient for this purpose? Assuming you have the historic data, at least in the first instance you can probably get "good enough" information from a few hours playing with the numbers. Personally I would trust this more than an AI anyway.

    If you don't have the data then I would start there. Neither traditional approaches nor AI will give you an accurate insight without it; outages will be entirely dependent on site-specific factors. If you don't already have something as well as making a point to note start/stop times it might be that your network operator is able to provide information, or you might be able to piece it together from metering information, relay timestamped info, server logs etc.

    It is worth noting in any analysis that you are probably experiencing several different types of outage, from transient dips to planned maintenance work, which have independent causality, frequency and duration.

Children