I have a model which uses the historical data to predict the electricity load consumption. I want to compare my model with a baseline model to show the performance, however I can not find a good base model for this purpose. I want to ask that do you have any suggestion that which method is used in industry for this purpose? or Do you have any suggestion about that?


1 Answer 1


I´m pursuing a PhD in this area so far I can tell that:

Industry is mostly based on robust yet simple models: AWS offers a solution that as far as I can tell uses ARMA based models (AWS Forecast)

ETAP uses probably a Markov Model that also includes information about weather for the forecast (ETAP Forecast)

Scheider probably uses a ARMA-Based Model alongside some Markov it also can use weather information ( Schneider Forecast )

Also I've talked with Small retailers them provide ARMA Based and few talk about Machine Learning Models.

In Research there is a complete different story.

Some Suggestions use CNN + LSTM (CNN 1, CNN 2)

But personaly I prefer models based on Multiresolution decomposition + LSTM like (MRA + SVR + LSTM and MRA + LSTM)

So far I've not found a common dataset for benchmark because most researchers uses their own collected data and also have not found a baseline model. Nevertheless most of them use the Mean Absolute Percentage Error as Metric and try to reach a MAPE < 5.0%

Best Regards and keep in touch. My Github

  • $\begingroup$ David Rodriguez Thank you so much for your great comments. $\endgroup$
    – sadcow
    May 4 at 18:32
  • $\begingroup$ It has been my pleasure to help you out! $\endgroup$ May 5 at 4:22
  • $\begingroup$ I am working on the smart meter data and this software and methods usually work on the feeder level data. I have one question, there is any way that you understand that ETAP uses the Markove or scheider uses ARMA-based? Thanks $\endgroup$
    – sadcow
    May 5 at 18:26
  • $\begingroup$ I also work with meters (SEL/ELSPEC/SCHNEIDER) and asked to engineers [Providers] ("It uses Past data for future data", but when I asked "Does the model uses ML/AI techniques like LSTM or XG" they answered, "So far we know, no, it doesn't use ML") Nevertheless they presented some special software in SCADAS that it does uses ML and AI for faults (Schneider). Also my guess is due to the MAPE obtained in Papers compared with providers and Markov because there is a lot of literature of models that use Weather and Series. Its my guess and I still try to obtain their architectures for research. $\endgroup$ May 6 at 19:21
  • $\begingroup$ Thank you so much. $\endgroup$
    – sadcow
    May 7 at 2:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.