Electrification is considered today as an essential developing factor for emerging countries. However, the architecture of the power systems to be developed is strongly dependent on the geographic distribution of consumers on one side and existing infrastructures on the other side. As such, isolated microgrids are seen as efficient and economic solutions to electrify remote rural areas, i.e. areas located far from an existing transmission network and where the cost of interconnection would be prohibitive. Isolated microgrids are thus only powered by locally installed generators and are operated in stand-alone, with no connection to any transmission network. This paper presents an original methodology for the multi-period planning of isolated microgrids in a green field context. The purpose is to find when and where generators and lines must be installed to achieve the cheapest system that ensures quality and reliability of the power supply. Planning of generation and distribution are usually treated separately, both in literature and real systems. However, in this context, there is a clear opportunity to improve performances and costs of the whole system by solving these two problems in a coordinated way. Furthermore, the problem is multi-period as the system has to be upgraded over time to be able to match a growing demand throughout the planning horizon. Minimizing the total cost of the system thus requires to make investment decisions over time by making a trade-off between immediate and future investments, taking the time value of money into account. Dynamic programming is a framework for multi-period decision making that allows to find an optimal sequence of investment decisions over a given planning horizon. An original tool is presented here that proposes a dynamic-programming approach to multi-period coordinated planning of generation and distribution. A non-linear and unbalanced tri-phase representation of the network is used to account for the effect of single-phase connected loads and generator on the voltage profile. The effectiveness of the proposed method is illustrated through several case studies.