Control and compensation algorithm for the temperature measurement lag problem of thermal resistors in distillation towers
Thread Content
Does anyone have any sharing on temperature measurement lag compensation algorithms? As well as case studies of actual solutions. In the temperature control project using a sensitive plate in the distillation tower, the temperature measurement had a lag of ~180 seconds, resulting in significant temperature overshoot and poor product quality. After preliminary analysis, the possibility of the thermometer being inserted too deep has been ruled out. This temperature control is achieved through a flow-temperature cascade for cooling the produced fluid, with the flow sub-circuit providing good control over the reaction rate.
Even after attempting to increase the differential, the effect remained poor. Although the valve responds quickly, the temperature still changes slowly. A temperature regulation cycle lasts about 7 minutes (from peak to trough temperature), with an overshoot of around 13 degrees in both directions. The valve opening can be quickly adjusted from 15% to 85%, in about 20 seconds. With the addition of differentiation, there are fluctuations at the zero-crossing point of adjustment (relative to the temperature setpoint), but this does not affect the final outcome.
I searched for online resources, and most of them covered the PSO control algorithm or other prediction algorithms, but are there any shared practical experiences? The time constant for temperature measurement can be determined experimentally; are there any simpler compensation methods (fitting formulas)?
When it is difficult to solve problems relying solely on the differential function of PID, what compensation algorithms from control theory can be used? Push on your own.
I’ll hold on a bit longer myself. I’m considering the performance of a fuzzy + PID approach; it’s relatively easy to implement. Please share some suggestions.
Fuzzy algorithms are probably the easiest to implement and also convenient to adjust. I would also appreciate guidance from experts.
What causes temperature changes? Can’t we use the change amount as a feedforward?
After introducing the feedforward-fuzzy PID, the situation improved, but the temperature overshoot was still at ±15 degrees. I’m thinking about considering compensation for the time constant on the sensor, that is, creating a historical value curve based on the time constant. When there is a slight change in temperature, it provides a predicted change for compensation. Please give me some suggestions.
If feedforward still doesn’t work well, then consider the impact of large lag. Try using the Smith predictor
Could you share some specific examples of how to do this? Want to learn about the implementation process of *fuzzy PID