For the urgent demand of real-time data compression of large turbine-generator units online monitoring and remote diagnosis， a novel real-time data compression and reconstruction method combining lifting wavelet transform and hybrid entropy coding is proposed. The signal to be compressed is firstly decomposed with biorthogonali filters based on lifting wavelet transform and the hard-threshold method is then applied to the wavelet coefficients. In order to implement higher efficient compression， a hybrid entropy coding scheme is employed to code the reserved coefficients. This method has been applied to the real-time data acquired from a large turbine-generator unit. Simulation results indicate that the proposed scheme which combines both lossy and lossless compression techniques can achieve satisfactory compression performance in applications.