Due to the imbalance in container transportation, empty containers pile up at certain parts of the network, leading to shortages elsewhere. To overcome this issue, stock levels at the depots must be rebalanced, and container allocation operations must be conducted efficiently. In this PhD project, we develop and explore efficient Deep Reinforcement Learning (DRL) algorithms to dynamically solve container allocation and rebalancing problems. To achieve this, we determine candidate algorithms, develop their architectures, and evaluate their performances by measuring the optimality gaps and comparing them with selected state-of the-art DRL algorithms on related problems. We will work on a real-life use case, Den Hartogh Logistics, to develop applicable solutions and conduct realistic evaluations. The use case involves the transportation operations of approximately 7200 company-owned tank containers in Europe. Currently, the company addresses stock level imbalances using basic inventory approaches and allocations using intuitive greedy techniques. The use case will also provide a good opportunity to compare the solutions with a real-life baseline.