Dr. Watcharin Sarachai
Sign in
complete

CM-SolarTS: A Chiang Mai Solar Photovoltaic Power Forecasting Dataset for Time Series Analysis

Introduces a dedicated solar photovoltaic power dataset from Chiang Mai and studies how different temporal feature-encoding strategies affect forecast accuracy across CNN, LSTM, and AR-LSTM models, finding that the best encoding approach depends heavily on the model family used.

CM-SolarTS: A Chiang Mai Solar Photovoltaic Power Forecasting Dataset for Time Series Analysis

Overview

Solar photovoltaic power forecasting is essential for balancing renewable energy grids, but public datasets from tropical, high-irradiance regions like Northern Thailand are scarce. This project introduces CM-SolarTS, a dedicated solar PV power dataset collected in Chiang Mai, and investigates how different temporal feature-encoding strategies affect forecast accuracy. The study compares CNN, LSTM, and AR-LSTM model families and finds that the best-performing encoding approach depends heavily on which model family is used, offering practical guidance for building solar forecasting pipelines.

Authors

Watcharin Sarachai, Chirawan Ronran

Publication

2026 Joint International Conference on Digital Arts, Media and Technology with ECTI Northern Section Conference on Electrical, Electronics, Computer and Telecommunication Engineering (ECTI DAMT & NCON), pp. 145–150, 2026, IEEE.

Time series forecasting chart
Time series forecasting chart

A classic time series chart illustrating the kind of temporal patterns that feature-encoding strategies must capture for accurate forecasting.