The dynamic nature of thermal processes means that predictive algorithms are an obvious choice for controlling processes of this type. Unfortunately, the product variability commonly found in real-world processes forces changes in the operating point. This, in turn, forces changes in the controller settings. Unfortunately, even for commonly used PI controllers, this is not usually done in practice. This approach results in a deterioration of the control loop efficiency indexes and, consequently, the efficiency of the process and even the quality of the product. Therefore, the implementation of a predictive controller in the control system that would not require personnel to reparameterize in the event of a change in the operating point seems to be a very attractive offer. The linear version of the DMC predictive controller meets these expectations, as it features a low computational complexity of the control law formula. This feature allows for its implementation in a PLC. It should be emphasized that the components of the control law for a specific operating point are the result of complex calculations that are difficult to perform in a PLC. A change in the operating point forces them to be redetermined. Therefore, only the combination of a PLC and an industrial computer (IPC) in an edge computing architecture allows the full use of the aforementioned advantages of the DMC predictive controller. This paper presents a predictive Edge Dynamic Matrix Control (EDMC) algorithm designed to control heat sources operate as a part of heat distribution systems. The EDMC algorithm is implemented partially in a PLC and partially in an edge device. This cooperation significantly increases the system’s available computational power and makes this solution possible to implementation in industry. In addition, the publication presents a comparison of the performance quality offered by the EDMC system described in relation to the commonly used PI controller.