Research on Cost Prediction Model for Wind and Photovoltaic Projects Based on Deep Learning

With the growing emphasis on renewable energy globally, wind and photovoltaic projects have become key components of the energy industry. Accurate cost prediction for these projects is crucial for planning, budgeting, and risk management. However, traditional cost prediction methods struggle to handle the temporal and spatial dependencies inherent in wind and photovoltaic projects. This paper proposes a method based on Adaptive Temporal Convolutional Neural Networks (AT-CNN) to address the limitations of traditional approaches. By incorporating multi-scale convolution kernels, adaptive attention mechanisms, and adversarial training strategies, the AT-CNN model effectively captures the nonlinear changes and dynamic features of project costs, improving prediction accuracy and robustness. Experimental results show that AT-CNN exhibits significant advantages in the construction phase of wind projects, especially in capturing dynamic changes, outperforming traditional regression models. Although errors increase in the later stages of photovoltaic projects, the overall accuracy remains higher than other models. The AT-CNN model demonstrates strong adaptability and application potential in cost prediction for wind and photovoltaic projects.