Hello everyone, I’m working on forecasting time series data for power systems.

I now have three years of meteorological data and corresponding power system data, and I want to predict the temporal changes of power system related power flow data based on meteorological conditions.But I don’t know how to use what Pkg in Julia to achieve this probability prediction algorithm (MLJ, Flux?).Can everyone help me?

Thank you very much!

It depends.

Bayesian forecast using a probabilistic model `Turing.jl`

.

Point Estimate forecast using traditional ML algorithms `MLJ.jl`

Point Estimate forecast using a Neural Network `Flux.jl`

Bayesian forecast using a Bayesian Neural network `Flux.jl`

+ `Turing.jl`

Can these Pkg be used for probabilistic prediction of multivariate time series?

Do you have a concrete model and methodology in mind? Forecasting is a very general term.

Meteorological conditions are the main objective factors affecting the operation mode of the power system. Meteorological historical data and the corresponding power flow data are used to train my neural network and analyze the possible operation mode of the power system in the future through weather forecast.

I want to combine LSTM training with quantile prediction and other methods to make probability prediction for dimensions of power system related data, and get prediction curves of different confidence intervals of these dimensions.See the figure below:

I also studied the Flux and MLJ documentation, but there seems to be no method that can be used for probabilistic prediction.