Cellular Neural network was introduced in 1988, and then the applications have been researching and developing. This paper presents a CNN architecture for solving set of partial differential equation describe phenomena happening in a bay with parameters water level and velocity in two directions. There are 4 parts: 1st introduction; part 2nd describe problem model; part two analyzes, finds CNN algorithm, then propose CNN hardware architecture, proof the stability and equivalence of CNN partial difference differential (CPDDE) to original PDE model, simulate in some case using FPGA chip; the conclusion part evaluates method and gives some trends for developing.