To address the issues of complex pump configurations, poor gain performance, and high costs in most current fiber Raman amplifiers (FRA), this study proposes a gain-flattened FRA design scheme based on the tree seed algorithm(TSA). By considering both average gain and gain flatness as target parameters and integrating the average power method, optimize the pump parameters (including pump wavelengths and pump powers) for first- and second-order FRAs. The evolution of signal and pump power along the transmission fiber is numerically analyzed, achieving flat amplification of C+L-band (100 nm) optical signals under different gain levels. Results demonstrate that the optimized four-pump FRAs achieve gain flatness values below 1 dB at gain levels of 8 dB, 11 dB, and 16.7 dB, with a minimum gain flatness of 0.4 dB. By comparing with existing optimization methods, it has been proven that the tree species optimization algorithm has significant advantages of high iteration efficiency and good optimization effect in the design of gain flat multi pump FRA, providing reference for the design of backward multi pump FRA.