Service restoration is a problem with multiple objectives and constraints, when the distribution systems with DGs (distributed generators) come up large area blackout. In this article the model of multiple objective service restoration about distribution systems with DGs has been built, considering users’ priority, the action times of hand switches or tele-switches and network losses after service restoring. Multi-agent GA can be used to search the Pareto optimal solution set for the multiple objectives service restoration problem. Agent's competitive and selflearning actions reflect the principal of survival of the fittest and diversity. It can reduce complex rate and improve efficiency using challenge match method to form Pareto solution to store populations. At last the calculation example manifests the quickness and precision of the multiple objective service restoration for blackout based multi-agent GA, and it is better suit for the distribution system than single objective method.