In the underwater optical wireless communication-direct current bias-optical-orthogonal frequency division multiplexing (UOWC-DCO-OFDM) system, in order to ensure that the optical signal has a low peak-to-average power ratio (PAPR) at the transmitting end and can carry out long-distance low underwater, the method combining the tone reservation-least squares algorithm (TR-LSA) and companding transformation was adopted. At the same time, the optimized neural network was used to estimate the channel of the underwater environment. Based on this method, the channel equalizer was designed at the receiving end to deal with the strong attenuation of the optical signal in the underwater environment. The results show that the PAPR of the UOWC-DCO-OFDM system is reduced by 9dB, and the bit error rate is 10-3 lower when the signal-to-noise ratio is 10dB, which is under the bit error rate standard of wireless optical communication. The system can realize long-distance underwater transmission of optical signals with low bit error rate.