ENSAE : Hidden Markov Models & Sequential Monte Carlo Methods

On particle filters applied to electricity load forecasting

Based on the article "On particle filters applied to electricity load forecasting"
Code : Antoine Grelety, Samir Tanfous, Zakarya Ali

Code Structure :

• Data treatment
  • Import Temperatures.ipynb : take care of temperatures data
  • utils.ipynb : add daytypes columns to RTE electricity load data
  • /data : folder containing all the data
• Parameters initialization
  • gibbs-parameters_init_v*.ipynb : Application of Gibbs Sampling to perform parameters initialization
  • elec_forecast/gibbs_sampling_model.py : Python class with Gibbs sampling methods
• Models evaluation
  • particle_filtering_*.ipynb : We use particle filter to estimate the model parameter and perform predictions
  • elec_forecast/particle_filter.py : Python class with Particle Filter methods (PMMH, Prediction, Resampling)
  • Robbins-Monro
    • parameter_estimation_robbins_monro.ipynb : notebook dedicated to parameter estimation via Robbins-Monro algorithm
    • elec_forecast/rm_estimation.py : Python class with Robbins-Monro algorithm methods