A process of upgrading the real-time dispatch at Rocky Reach and Rock island power plants was initiated by Chelan County PUD for increased efficiency and enhanced coordination between the two projects. The two plants are located on the Columbia River in Western United States and are part of a cascade of large projects operated by different owners, including Grand Coulee, Chief Joseph at the upstream end of the Mid-Columbia, and Wells further downstream. Rocky Reach has 11 units with a total capacity of 1250 MW with some upstream storage capability, and Rock Island has 19 units with a capacity of 650 MW and fairly limited storage. The implementation of an upgraded real-time software system is in response to a change in the Mid Columbia coordination process between the various plant owners and operators. One of the requirements for the upgraded real-time dispatch system was to meet a total system load demand imposed on both projects in real-time, while meeting the various operating license constraints. Hence, water management was an integral part of the solution as it was important to synchronize the operation between both plants, especially considering their close proximity, resulting in the overall operation being coupled both hydraulically and electrically. It was decided to adopt an optimization-based approach, where the entire dispatch by plant and by unit is re-calculated at every cycle. Such an approach can be described as a smart AGC (Automation Generation Control) function, where not only the real-time application looks after the redistribution of load changes to the various on-line units, but makes decision on water management, moves generation from one plant to another, and changes unit commitment. Given that load balancing is re-calculated on a 4-second cycle frequency, this was a challenging problem both in terms of numerical efficiency but also as related to data flow performance. It should be noted that most optimization technologies as applied to water management typically deal with longer time scales due to the computational time generally required to find near optimal solutions through well proven mathematical programming methods for non-linear systems with discrete decision variables. The solution integrated Hatch’s commercial model RT Vista to perform the optimization of the real-time dispatch, where further customization was applied to meet the specific requirements of the project. The application was tightly interfaced with Chelan County PUD’s other systems and models, including SCADA system in the control room, Energy Management System (EMS) web Accounting, and the short-term model that provides water management guidance over several days. To deal with the numerical challenges of the real-time environment, a two layer scheme was used, with one optimizer running continuously at a 5-minute time interval while considering the forecast in load, inflows as well as the various water management constraints, and another instance running continuously at a 4-second interval to respond to near instantaneous load changes. A large amount of data was required to communicate back and forth between the new real-time dispatch function and Chelan County PUD’s other systems. Therefore, the development of the interface was a significant activity during this deployment. This paper will discuss the various challenges faced during the system implementation, the experience gained since its deployment and its overall use in the control room environment.