Multi-view clustering enhances clustering performance by leveraging multiple views to address the limitations of single-view data. However, in distributed settings, discrepancies in data distribution, quality, and quantity across clients can negatively impact feature representation from multi-view data, leading to instability in clustering results and degraded clustering performance. To tackle this, we propose a Differentiated Dual-supervised Deep Distributed multi-view clustering model (D4MVC). Each client locally constructs deep autoencoders and clustering models for each view, extracting useful features and generating preliminary clustering results. The server then selectively generates global supervisory signals based on the sample distribution and encoded information from clients, which are distributed back to the clients. Clients adaptively combine local and global supervision in a dual-supervision framework, capturing optimal global supervision while considering local data characteristics. Extensive experiments on six datasets demonstrate that our approach significantly improves clustering performance in heterogeneous distributed scenarios.