That was using 15 minute usage, maybe 5 minute.AustinScubaAg said:current data does not have a high enough sample rate to do better than provide a shape. The more the sample rate increases the more information that can be extracted.texagbeliever said:I have seen data methodologies like that applied to residential customer usage in Texas. The takeaway is and has always been this is the general shape. The data is too random to be properly normalized so it becomes extremely difficult to extract details like you are talking about. Only the strictest of OCD time people will live and operate in a way to glean that information off of them.AustinScubaAg said:This is signal theory 101. The only question is sample interval. At 15 min the data will be noisy. If the sample rate is switched to 1 min the data would be super easy to filter. This does not take a super skilled person.texagbeliever said:It would be really tough. Light, random appliance usage, electronics, A/C kicking on and off, ceiling fans, garage door opening, the residents moving out, residents WFH, etc.AustinScubaAg said:Actually this is fairly easy to do since appliances have specific power draw over time. It would be fairly easy to train a neural network to detect the usage patterns of specific appliances. The only question is if using a 15 minute average can give enough fidelity to get that specific.texagbeliever said:DD88 said:
3. Wrong a smart meter doesn't have access to individual appliances.
It would be easy to determine major appliances like dish washers or washing machines by combing the water + electricity data.
Basically things change frequently enough you would struggle to pin down exact things. Sure you know the family runs their dishwasher, Washing machine and Dryer, but you already knew that. This would be a TON of effort for almost no return. Also the competency and skills required would far exceed what one would suspect for a government employee.