Abstract: The performance of community green spaces is a quantitative indicator to measure the reasonability of community green space distribution. As urban governance in China progressively shifts towards community-level initiatives, evaluating the performance of green spaces at a micro-scale has become a crucial approach to optimize the people-oriented layout of urban green spaces. However, current measurement methodologies and evaluation criteria of the performance of community green spaces are mainly derived from macro-scale measurement approaches, and none of them can reflect differentiation features among different communities. As a result, many problems with the application of space performance measurement methods appear, including conceptual generalization, standard confusion, and objective ambiguity, thus resulting in a significant gap between measurement results and the practical situation. To address these problems, a measurement method for green space performance integrating multi-agent simulation was proposed in this study by focusing on the measurement connotations of green space performances in communities. 
　　Firstly, evaluation parameters on both the supply and demand sides were optimized based on analysis of features formed by the individual performances of community green spaces. With consideration to the influence of spatial components (e.g., activity facilities and site conditions) on the choices of residents in green spaces on the supply side, parameters like green space attractiveness and green space scale thresholds were introduced into the traditional spatial network model. On the demand side, the calculation of probability of residents' subjective choices was introduced into the model by combining the Huff Model to simulate residents' preferences for community green spaces. Meanwhile, the moving distance of different individuals in the simulation process was expressed quantitatively using probability distribution. 
　　Secondly, a space performance measurement model was constructed by setting residents as intelligent objects. This model evaluates the distribution condition of green space performance in communities by simulating the activity behaviors of residents in the green space. The model traversed all initial points. The simulation was terminated when the "number of travelers" was zero. Finally, two types of space performance measurement indicators—namely, service population coverage of green spaces and residential accessibility—could be outputted respectively from perspectives of "capacity" and "opportunity" according to the "number of stayed residents" at the initial points, and "number of activities" at the destination points. 
　　Eventually, an empirical test based on three communities in Beijing was carried out. Results showed that compared to the traditional buffer zone method, which simply creates statistics on population coverage, the optimized method can precisely recognize population within the individual service range of green spaces that genuinely meets the utilization demand. Results change significantly. The maximum variation amplitude in the case study was 91.9% and all numerical values decreased by 64.7% on average. By analyzing accessibility results at the demand side based on the optimization method and two-step displacement method, it was found that distance is not an absolute influencing factor in accessibility evaluation of community green spaces, and attractiveness of green space more greatly influences the distribution of general community space performances. The optimization model more greatly emphasizes coupling relation and individual features of mutual influence between attractiveness and performance distribution of community green spaces at the micro-scale. Moreover, it can reveal the influence of individual supply-demand feature changes on overall performances at the urban micro-scale. It has the higher precision and sensitivity compared to traditional performance measurement methods.

Key words: community green space, space performance, two-way relationship between supply and demand, spatial model

摘要： 社区绿地空间绩效是衡量社区绿地配置合理程度的重要指标。其传统测度方法多沿用城市宏观尺度下的评测内容，故难以突出社区尺度差异化的绩效内涵。针对此问题，提出融入多智能体模拟的空间绩效测度方法。首先基于对供需两端个体特征的观测分析，在网络模型中引入居民选择概率、绿地吸引力等参数；其次融合居民空间决策行为的仿真模拟构建绿地空间绩效精细化评测模型，实现供需双向关系对绩效测度的有效反馈；最后以北京3个社区为例进行实证检验。结果表明优化模型较传统测度方法，在社区绿地空间绩效评测的精度方面有显著提升效果。

关键词: 社区绿地, 空间绩效, 供需双向关系, 空间模型