Differential Evolution Algorithm Based On-line Adaptive Control And Its Application In High Density Yeast Cultivation Processes

In fed-batch culture of yeasts, excessive substrate addition leads to severe ethanol accumulation which would destroy cells structures and functions and decrease substrate utilization efficiency, while insufficient substrate addition limits cell growth. To solve this problem, a self-adaptive control strategy based on differential evolution algorithm(DE) was proposed. On the basis of traditional proportional-integral-differential(PID) control strategy, the proposed strategy could adaptively identify the kinetic characteristic of yeast cultivations and predict the state parameter subject to control with the aid of autoregressive moving average model(ARMA). PID controller’s parameters were then automatically regulated by minimizing the objective function, the deviation of the predicted state parameter and its set-point via DE algorithm. On-line adaptive substrate feeding control system was thus developed. The major results and conclusions were summarized as follow:(1) Based on the existing dynamic model of S. cerevisiae culture system, performance comparisons in S. cerevisiae fed-batch cultures using traditional PID feedback control and the modified DE-PID control were studied by computer simulation, with ethanol concentration as the on-line feedback variable. The results indicated that ethanol concentration could be maintained at low level of 1 g·L-1 steadily with the DE-PID strategy, while biomass concentration could reach to the level of 34.45 g·L-1 which was 29% higher than that obtained with the traditional PID strategy.(2) In experimental S. cerevisiae fed-batch cultivations, DE-PID feeding control strategy was applied by controlling ethanol concentration or respiration quotient RQ at constant levels. In the former case, ethanol concentration could be controlled at the lower ranges(0.02 ~ 2.35 g·L-1) throughout the feeding stage(up to 30 h) but biomass concentration could only reach 23.25 g-DCW·L-1. On the other hand, biomass concentration could reach a higher level of 47.75 g-DCW·L-1 in the latter case which was 85.44% higher than that obtained by DO-Stat based feeding control strategy.(3) The proposed DE-PID control strategy was also applied for constant controls of ethanol concentration or/and dissolved oxygen concentration(DO), in high cells density culturing of recombinant P. pastoris. With the strategy, biomass concentration could reach 112.25 g-DCW·L-1 and 113.25 g-DCW·L-1 after 34 h cultivation, and ethanol concentrations in both cases could be maintained at low levels(0.09 ~ 1.75 g·L-1) during the fed-batch stage. Compared with the previously developed “improved” DO-Stat glycerol feeding strategy, the DE-PID strategy could achieve high cells density and repress ethanol accumulation simultaneously. In addition, the proposed control strategy stabilized DO control level and enhance the performance of both process control and cells cultivation.