Abstract: To address the issue of differences in the adaptability of hydro-wind-solar power simulation performance to modeling sequence lengths, a joint stochastic simulation model is first established to generate runoff, wind power, and photovoltaic output data sequences of various lengths. Then, three traditional data-driven models (RF, SVM, GKR) and six deep learning models (LSTM, GRU, CNN, BiLSTM, iTransformer, TCN) are adopted to establish hydro-wind-solar power simulation models, respectively. The responses of model simulation performance to different modeling sequence lengths are compared and analyzed. Taking the hydro-wind-solar base in the upper Yellow River as a case study, the results show that: the SVM model achieves high simulation accuracy, with an average test set R² of 0.878. For runoff and wind power, the optimal data sequence length is 70 years; for photovoltaic power, most models perform best at sequence lengths of 70 years and 100 years. This study can provide a reference for model selection in practical engineering, and offer optimal data length reference values under various data conditions.