Traditional ultra-short-term photovoltaic (PV) power forecasting methods often rely on a single prediction model
resulting in limited stability and generalization capability. To overcome these shortcomings
this study proposes a novel ultra-short-term PV power forecasting approach integrating seasonal similar-day clustering with Stacking ensemble learning. Key meteorological features relevant to PV output are first identified using Pearson and Spearman correlation analyses. Historical data from an Australian PV plant are then processed through seasonal -means similar-day clustering to capture the seasonal characteristics of PV generation more accurately. Four diverse base learners—long short-term memory (LSTM) networks
convolutional neural networks (CNN)
extreme learning machine(ELM)
and support vector machine(SVM)—are combined
with extreme gradient boosting (XGBoost) serving as the meta-learner. The hyperparameters of XGBoost are optimized via the particle swarm optimization (PSO) algorithm to construct the final Stacking ensemble model. Simulation results verify that the proposed method delivers enhanced forecasting accuracy and stability under diverse seasonal and meteorological conditions.
GUPTA P, SINGH R.PV power forecasting based on data-driven models: a review[J]. International journal of sustainable engineering, 2021, 14(6): 1733-1755.
DAS U K, TEY K S, SEYEDMAHMOUDIAN M, et al.Forecasting of photovoltaic power generation and model optimization: a review[J]. Renewable and sustainable energy reviews, 2018, 81: 912-928.
HUANG Q, WEI S Y.Improved quantile convolutional neural network with two-stage training for daily-ahead probabilistic forecasting of photovoltaic power[J]. Energy conversion and management, 2020, 220: 113085.
PAN M Z, LI C, GAO R, et al.Photovoltaic power forecasting based on a support vector machine with improved ant colony optimization[J]. Journal of cleaner production, 2020, 277: 123948.
WU T, HU R F, ZHU H Y, et al.Combined IXGBoost-KELM short-term photovoltaic power prediction model based on multidimensional similar day clustering and dual decomposition[J]. Energy, 2024, 288: 129770.
ZHANG M Y, HAN Y, WANG C Y, et al.Ultra-short-term photovoltaic power prediction based on similar day clustering and temporal convolutional network with bidirectional long short-term memory model: a case study using DKASC data[J]. Applied energy, 2024, 375: 124085.