Abstract: To evaluate the activated sludge PHA synthesis yield prediction under the condition of mixed carbon sources, genetic algorithms were proposed to optimize the weights and thresholds of the BP artificial neural network, and the research established the prediction model based on the GA-BP network. Food waste fermentation liquid was applied as the substrate and activated sludge was used to synthesize PHA under the ADD mode. Based on experimental data, the comparison between the measured data and the model predictions was carried out, to verify the accuracy of the network prediction model and the prediction of the long-term PHA synthesis ability. Results showed that the GA-BP network model improved based on the genetic algorithm had high prediction accuracy than the traditional BP neural network model, and the model explored a feasible method to evaluate the long-term variation of the maximum PHA production yield in mixed microbial cultures and determined the reasonable enrichment time.