With the continuous increase of load in the computer room, it is difficult for traditional monitoring methods to accurately predict power changes and identify anomalies. Therefore, an intelligent monitoring system integrating LSTM and SVM algorithms is researched and designed to improve load prediction accuracy and anomaly detection ability. The experimental results show that the mean square error (MSE) of the LSTM-SVM model is reduced to 0.031, and the accuracy is 94.6%. After SVM model is combined with wavelet denoising and RBF kernel function, the anomaly detection accuracy is improved to 91.6%, and the recall rate is 89.2%. The intelligent monitoring system combined with LSTM and SVM shows high accuracy and stability in the task of load prediction and anomaly detection, which can provide strong technical support for the future intelligent operation and maintenance of the computer room. The study further incorporates a terminal–edge–cloud architecture to ensure real-time monitoring and decision-making. Comparative analysis with traditional models validates the superiority of the proposed approach. The originality of this work lies in unifying forecasting and anomaly detection, while addressing high-dimensional power data. Practical implications include improved resilience, energy efficiency, and reduced downtime for modern data centers.