Forecasting monthly solar PV generation using ARMA and seasonal ARIMA models, including simulation of seasonal processes and autocorrelation structure analysis.
This project explores how autoregressive models capture both short-term dependencies and seasonal patterns in renewable energy generation. Through analytical and simulation tasks, it studies model stability, stationarity, and prediction performance for solar power data.
The repository contains the work for Assignment 2: ARMA Processes and Seasonal Processes, completed as part of a time series analysis course.
It includes theoretical exercises on AR(2) models, seasonal model simulations, and practical forecasting of solar power generation.
| File | Description |
|---|---|
assignment2_2024.pdf |
Original assignment instructions and theoretical background. |
assignment2_ex1.R |
Stability and autocorrelation analysis of an AR(2) process. |
assignment2_ex2.R |
Seasonal AR model applied to monthly solar power forecasting and model validation. |
assignment2_ex3.R |
Simulation of ARMA and seasonal time series, with ACF and PACF visualization. |
datasolar.csv |
Monthly solar PV generation data (year, month, power). |
Rplots/ |
Plots of time series, ACF, PACF, and forecast results. |
time-series-analysis-as2.Rproj |
RStudio project configuration. |
-
Models:
- AR(2) model — used to study stability, invertibility, and autocorrelation.
- Seasonal AR(1) model — applied to predict 12-month ahead solar generation.
- Simulated seasonal ARMA processes — analyzed through ACF/PACF structures.
-
Techniques:
- Stationarity and invertibility checks
- Model validation via i.i.d. residual tests
- Forecasting and prediction interval estimation
- Simulation of stochastic time series
-
Tools:
R, ggplot2 (visualization), xtable (LaTeX tables), base R time series functions.
- @s233239 (zoewr)
- @kongehund
- @fenfen22 (Fenfen)
- DTU Course material: Time Series Analysis (2024).