Over the past few years, with the continuous maturity of cloud computing technology and the continuous expansion of application fields, more and more enterprises and institutions have begun to migrate their business and data to the container cloud platform. How to effectively predict the container workload has become the core issue to ensure the efficient use of cloud resources. Through accurate prediction, resource allocation can be optimized, thus improving service efficiency and cost-effectiveness. However, due to the large fluctuations in container workload, traditional prediction models often find it difficult to accurately capture the characteristics of its dynamic changes, thus accurately predicting container load in the cloud remains a significant challenge. In this paper, we introduced a patch-based convolutional neural network for predicting docker workload, which can capture both short-term and long-term dependencies in container resource time series. We used Alibaba cluster tracking to train our model and use MSE and MAE to evaluate our proposed method. The experiment results show that the PCNN model exhibits superior workload forecasting performance in comparison to commonly used workload forecasting models.