Prediction of Storage Quality and Multi-Objective Optimization of Storage Conditions for Fresh L. Based on Optimized Latin Hypercube Sampling.

Quality control of fresh during storage presents significant challenges, particularly regarding the unclear relationship between quality characteristics and storage conditions. This study analyzes the changes in qualitative and structural characteristics, including fruit hardness, soluble solid content (SSC), titratable acidity (TA), and vitamin C (Vc), under various storage conditions (temperature, duration, and initial maturity). We employed optimized Latin hypercubic sampling to develop radial basis function neural networks (RBFNNs) and Elman neural networks to establish predictive models for the quality characteristics of fresh wolfberry. Additionally, we applied the Particle Swarm Optimization (PSO) algorithm to determine the optimal solution for the constructed models. The results indicate a significant variation in how different storage conditions affect the quality characteristics. The established RBFNN predictive model exhibited the highest accuracy for TA and Vc during the storage of fresh wolfberry (R = 0.99, RMSE = 0.21 for TA; R = 0.99, RMSE = 0.19 for Vc), while the predictive performance for hardness and SSC was slightly lower (R = 0.98, RMSE = 385.78 for hardness; R = 0.94, RMSE = 2.611 for SSC). Multi-objective optimization led to the conclusion that the optimal storage conditions involve harvesting fruits at an initial maturity of 60% or greater and storing them for approximately 10 days at a temperature of 10 °C. Under these conditions, the fruit hardness was observed to be 15 N, with SSC at 17.5%, TA at 1.22%, and Vc at 18.5 mg/100 g. The validity of the prediction model was confirmed through multi-batch experimental verification. This study provides theoretical insights for predicting nutritional quality and informing storage condition decisions for other fresh fruits, including wolfberries.