My friend you can use Artificial Neural Network to make predictive model based on previous information using data of similar city that its information was per-designed.

You may wish to consider the following steps with respect to the development of pipe diameter prediction model considering water supply. First of all, obtain the population data for all the communities of interest. Project the population of these communities using the design life. Then, determine the water demand for each community by multiplying the population by the per capita per day demand of water (you can use the standard set by the national agency or the WHO). Factor the demand by the peak daily and peak hourly coefficients in order to obtain the total demand for the respective communities. Carryout extended hydraulic simulations for the different communities using EPANET. Optimize the design to meet requisite pressure requirements as set by national or international agency. Fit the parameters of pipe diameter (dependent variable), populations (for the different communities), pipe lengths (for the different communities) etc. Sepaprately consider fitting the data using linear, power, exponential approaches. Whichever yields the highest coefficient of correlation can be subjected to sensitivity analysis. The less sensitive variables should be dropped while the sensitive parameters are used in the development of the final model. However, it must be noted that the developed model may only worked for communities with similar characteristics.

You can also develop your own using Hardy cross method.