Search engines are typically used to acquire desired knowledge. Other than return the exact match of the query, users also want the search engine to help them do some knowledge exploration, i.e., getting some related by not exactly matched results. In this work, we introduce a novel knowledge graph-enhanced search engine aimed at facilitating knowledge exploration. Through query reformulation, we obtain a set of knowledge-enhanced subgraphs, whose matches are then returned as the “explored” results. This approach effectively broadens the knowledge domain of the original query, offering users a more diverse set of relevant information. To achieve these objectives, we vectorize the knowledge-enhanced subgraph set and select a fixed-size subset in the subgraph vector space. This ensures that the subset covers the expanded knowledge domain of the original query subgraph while maintaining relevance to it. Additionally, we propose a new clustering-based distributed result diversification algorithm to further enhance our approach. Compared to traditional result diversification algorithms, our method significantly reduces runtime while maintaining relatively high effectiveness. Experimental evaluations conducted on real-world datasets validate the effectiveness and efficiency of our proposed method. In the same distributed environment, our algorithm ensures a quality evaluation loss within 3.5%, while achieving a 1.7-2.0x improvement in time efficiency compared to traditional algorithms. Case studies further demonstrate the practical utility of our approach in real-world queries. Furthermore, our method can serve as a query preprocessing module, benefiting state-of-the-art dense retrievers in terms of both in-domain and out-of-domain results, thereby enhancing the comprehensiveness and diversity of search outcomes.