Accurate and reliable price forecasting of agricultural products is significant for promoting the production and distribution of agricultural products, optimizing resource allocation and improving market efficiency. Due to the non-stationary feature in agricultural price, based on decomposition integration and kernel density estimation(KDE), this paper proposes a hybrid model for agricultural price forecasting that can quantify the uncertainty of potential forecasts by converting traditional point forecasts into interval forecasts. Firstly, the price sequence is decomposed through variational modal decomposition (VMD) determined by energy entropy (EE); secondly, K-means clustering is used to reconstruct the intrinsic modal functions (IMFs) into low-frequency components and high-frequency components, forecasted by different methods. In addition, adaptive kernel density estimation (AKDE) is established through dynamic window and whale optimization algorithm (WOA), which is used to construct prediction interval with the residual signal obtained by VMD. Finally, to validate the superiority of the proposed model, comparative experiments with three different datasets are conducted. The results show that prediction performance of the proposed model is better than other models in both point forecasts and interval forecasts, and it can provide more accurate uncertainty information to agricultural participants.