The present paper proposes a multi-agent simulation model for analyzing the peak-shaving efficiency of natural gas TOU (time-of-use) pricing for residential consumers. Firstly, a natural gas price sensitivity function has been explored according to gas consumption characteristics of residential users, which establishes the residential users demand response model from the perspective of consumer psychology. Secondly, a TOU pricing multi-agent simulation system has been developed, which mainly includes a government agent, a gas operator agent, and seven residential user agents of different income level. Finally, this study takes Zhengzhou city in China as an example to simulate the dynamic change process of running states in the UGPN (Urban Gas Pipeline Network) under the residential users TOU pricing policy. The simulation results indicate that the maximum peak-valley load difference can be reduced by 11.12%, given the residential users' response to TOU tariffs, by shifting load and electricity substitution. The TOU price also increases benefits of gas operators, and reduces residential consumers' energy expenditure. Furthermore, the small ratio of the residential users' consumption will lead to the low peak valley load difference, but poor peak shaving efficiency and little gas operators' benefits. In addition, the expenditure on energy changes among different kinds of consumers; that is, the high and low income households have a large increased ratio, while the middle income households have a low increased ratio. (C) 2015 Elsevier Ltd. All rights reserved.