The invention discloses a reciprocating compressor fault diagnosis system and method based on a neural network algorithm. The reciprocating compressor fault diagnosis system and method based on the neural network algorithm comprises a field communication module, a data collecting module, a simulation model based on neural networks, an expert system module based on standard fault characteristics, afault diagnosis main machine and an upper computer. According to the reciprocating compressor fault diagnosis system and method based on the neural network algorithm, the neural network algorithm isutilized, measured compressor system status parameters are combined, and thus simulation models in all working conditions are acquired; compressor faults are simulated by changing system parameters tocalculate and obtain the standard fault characteristics, after the abnormal operation of the compressor occurs, through the comparison with the standard fault characteristics, fault parts can be accurately located, and elimination methods can be brought; and the diagnosis success rate is high, and real-time monitoring analysis of high precision equipment is not needed, and the diagnosis cost is lowered.