Optimizing Cloud Computing Energy Consumption Prediction Using Convolutional Neural Networks with Bidirectional Gated Cycle Unit

This paper proposes a model that combines a Bidirectional Gated Recurrent Unit (BiGRU) with Convolutional Neural Network (CNN) optimization for predicting cloud computing power consumption. Experimental results show that the model significantly improves prediction accuracy, with the Root Mean Square Error (RMSE) of the training set decreasing from 0.6 to below 0.2, and the loss value dropping from 0.08 to 0.02, eventually stabilizing. These results indicate a notable enhancement in the model’s prediction performance for cloud computing power consumption. The model achieves an $\mathrm{R}^{2}$ of 0.59 on the training set and 0.54 on the test set, demonstrating reasonable predictive capability, though the slightly lower performance on the test set suggests some overfitting. Despite this, the model presents a viable approach for cloud power consumption prediction. Future research will focus on further optimizing the algorithm and improving the model’s generalization ability to enhance its accuracy and stability in real-world applications. This study offers valuable theoretical insights and technical guidance for optimizing cloud computing resource management and energy consumption.