Industrial equipment failures are often apparent in the multivariate time series, and timely detection of these anomalies is critical for production safety and economic loss avoidance. The relationships of variables in multivariate time series are intricate and interactive and have a significant impact on the detection performance. However, existing methods struggle to adequately capture the temporal features of time series data and effectively address the intricacies of their multivariate interdependencies, leading to the loss of fine-grained features. To solve this issue, in this paper, we propose a Temporal Dependencies and Multivariate Correlations Integrating (TDMCI) method for time series anomaly detection. Specifically, to fully extract temporal features, we propose a Multidimensional Feature Extraction Block (MFEB) to extract local and global temporal context features respectively. We then propose a Global-local Temporal Feature Aggregation Block (GTFAB) to enable local temporal features to fuse global temporal features, resulting in higher-resolution feature representation. In addition, we propose a Multivariate Information Interaction Block (MI \(^2\) B) for multivariate information relationship extraction to enhance the interaction between multivariate for anomaly detection. The F1 scores of TDMCI on the SWaT and PSM datasets are 96.95% and 97.55%, respectively, demonstrating competitive performance with baseline anomaly detection methods.