Local co-location pattern (LCP) mining is an important branch of spatial co-location pattern mining, which aims to discover co-location patterns that prevalently co-occur in local regions. The LCPs can reveal the implicit association relationships among spatial features in local regions but not on a global scale. Existing LCP mining algorithms cannot effectively identify artificially divided local regions, and it is difficult to set an appropriate prevalence threshold for selecting prevalent LCPs. To address these problems, we propose a novel top-k LCPs mining algorithm based on weighted Voronoi diagram (called Top-k LCPM-WVD). This algorithm effectively identifies LCP distribution regions formed by human factors using the weighted Voronoi diagram, and proposes a top-k mining framework to efficiently mine the k most prevalent LCPs within regions. Extensive experimental evaluations are conducted on both real-world and synthetic datasets. Compared to the existing state-of-the-art algorithms, our proposed Top-k LCPM-WVD algorithm yields more interpretable LCPs with good efficiency.