Under the dual challenges posed by global climate warming and the accelerating pace of urbanization, the urban heat island effect has emerged as an increasingly severe environmental issue, exerting profound impacts on urban thermal comfort, ecological sustainability, and public health. In this context, the cold island effect of urban green spaces—referring to their capacity to mitigate local air temperatures through ecological and microclimatic regulation—has garnered significant attention for its potential to alleviate urban thermal stress, enhance urban climate resilience, and contribute to the creation of livable, low-carbon urban environments. However, despite its recognized value, there remains a lack of systematic planning approaches and analytical models to quantify and optimize the spatial configuration of green spaces in a way that fully leverages their collective cold island functionality. Against this background, this study took the urban green space system as its principal research subject, and sought to explore, from the perspective of spatial planning and ecological network theory, the mechanisms, structural patterns, and optimization strategies that could enhance the overall cold island effect of such systems at the city scale. Building upon the foundational ecological mechanism of “networking” —which optimizes ecosystem performance through the integration of functional elements and the strengthening of inter-patch synergies—this study introduced the innovative notion of a “cold island network”. This concept posits that the cold island effects of individual green space patches are not isolated phenomena, but rather function as interconnected spatial entities whose synergistic interactions can significantly reinforce the systemic cooling performance of urban green spaces. In order to translate this conceptual framework into a planning-oriented methodology, the study first constructed a predictive model of green space cold island extent, using patch size as the core explanatory variable. The model, developed within a planning discourse that prioritizes operability and policy relevance, revealed that the spatial extent of cold island effects demonstrates an inverse tangent functional relationship with green patch size, thereby providing a quantitative basis for defining and measuring the potential spatial cooling radius of individual green spaces, which in turn visualized the spatial distribution of the green spaces’ cold island effect. Subsequently, using the green space system planning of Chongqing High-tech Zone as a case and employing tools from complex network analysis, this research constructed a cold island effect network model by conceptualizing green space patches as “nodes” and regarded the spatial overlap of their cold island extents as “edges”. Through a multi-scale analytical framework encompassing the system level, the subgroup level, and the individual node level, the study identified several key structural characteristics of the existing cold island network in the study area. These included a generally lowlevel of overall network integration, significant divergence in connection efficiency among the three major subgroups, and a notable absence of highcentrality nodes—except for the Zhaishanping subgroup—which are typically essential for maintaining network coherence and maximizing the spatial coverage of cold island effect from green spaces. In the final stage of analysis, the constructed cold island network was overlaid with the current spatial distribution of heat islands in the study area, thereby enabling the identification of critical cooling “blind spots” —areas where cold island effects are absent. Based on this integrated spatial diagnosis, the study proposed a set of three targeted planning strategies aimed at optimizing the spatial layout and functional connectivity of the green space system, with the overarching goal of enhancing its systemic cold island performance. These strategies included: 1) constructing a regionally structured green network that improves macro-level spatial continuity and functional integrity; 2) strengthening intra-group connectivity to weave ecological links across cooling blind spots; and 3) regulating the morphological characteristics and spatial orientation of green patches in order to amplify cold island effect at the node level. By introducing a novel framework that combines predictive modeling with network-based spatial analysis, this study does not only offer a new theoretical and methodological lens for understanding and enhancing the cooling performance of urban green spaces, but also provides practical support for planners and policymakers tasked with mitigating urban thermal risks, and presents valuable references for the development of related standards in green space planning and design.