PID tuning with Ant Colony Optimization (ACO) : A framework for a step response based tuning algorithm

Abstract: The building automation industry lacks an affordable, simple, solution for autonomous PID controller tuning when overhead variables fluctuate. In this project, requested by Jitea AB, a solution was developed, utilising step response process modelling, numerical integration of first order differential equations, and Ant Colony Optimization (ACO). The solution was applied to two control schemes; simulated outlet flow from a virtual water tank, and the physical air pressure in the ventilation system of a preschool in Sweden. An open-loop step response provided the transfer function in each case, which, after some manipulation, could be employed to predict the performance of any given set of PID parameters, based on a weighted cost function. This prediction model was used in ACO to find optimal settings. The program was constructed in both Structured Control Language and Structured Text and documented in an approachable way. The results showed that the program was, in both cases, able to eliminate overshoot and retain the settling time (with a slightly raised rise time) achieved with settings tuned per the current methods of Jitea AB. Noise and oscillations present in the physical system did not appear to have any major negative influence on the tuning process. The program performed above Jitea AB’s expectation, and will be tested in more scenarios, as it showed promise. Autonomous implementation could be of societal benefit through increased efficiency and sustainability in a range of processes. In future studies, focus should be on improving the prediction model, and further optimising the ACO variables.