1. Basic Concepts and Control Principles of APC

Advanced Process Control (APC) is a general term for a class of control strategies that differ from traditional PID control, mainly used to handle complex industrial control problems where conventional control methods are ineffective or even inapplicable. The main types of advanced process control include decoupling control, model predictive control, statistical quality control, adaptive quality control, internal model control, fuzzy control, and neural network control. Among these, model predictive control (MPC) is the most commonly used in the renovation of sulfuric acid production. Model predictive control is a control strategy based on the dynamic model of the process. It predicts the future trend of the controlled variable through the predictive model and corrects the prediction for future moments based on the deviation between the current measured value of the controlled variable and the previous prediction for the current moment, thereby making control decisions in advance and optimizing the control input at the current moment. Through the three core steps of model prediction, feedback correction, and rolling optimization, model predictive control can promptly overcome uncertainties caused by model mismatch and external disturbances, significantly improving the dynamic performance of the control system. Regardless of how the algorithm is designed, model predictive control always has the three basic features of a predictive model, rolling optimization, and feedback correction. For instance, if the process requires a setpoint of 300°C and the feedback received is 298°C, the rolling optimization takes the setpoint and feedback temperature, and through an internal optimization algorithm, generates a series of plans such as opening the valve by 1% now and another 2% in the next step. These plans are then sent to the predictive model to predict trajectory data, which is returned to the rolling optimization. The rolling optimization repeatedly compares with the setpoint and selects the optimal valve adjustment action. As the controlled object changes, the actual feedback value and predicted feedback enter the feedback correction module, where there will be a prediction error. Through the correction algorithm, the results of the predictive model are corrected, and the error is rolled over and repeated until the target temperature of 300°C is reached.

2. APC Multivariable Decoupling Process

There are many examples of strong multivariable coupling in industrial fields. Taking the boiler system as a specific instance:

Suppose the opening degree of the fuel valve is MV1, the rotational speed of the induced draft fan is MV2, the furnace temperature is PV1, and the furnace negative pressure is PV2.

The coupling relationship between them is that when the fuel valve opening (MV1) is increased, the furnace temperature will rise, the combustion will become intense, and the furnace negative pressure will also increase. As the induced draft fan speed increases, the temperature and negative pressure will decrease. If traditional PID control is used for the two loops, they will inevitably conflict, resulting in continuous oscillation of the system. The way APC solves this kind of problem is to include the coupling relationship between variables in the modeling process, and consider all PVs and MVs in the rolling optimization. When the temperature needs to be increased, the MPC optimizer will calculate that when the fuel valve opening is increased and the temperature reaches the required level, the negative pressure will also increase and may exceed the upper limit. The operation will be to slightly increase the induced draft fan speed to offset the increase in negative pressure, coordinating the relationship between the two to meet the process requirements. The traditional decoupling method is to decouple the two relationships and control them separately. When controlling variable A, variable B needs to be determined through calculation to achieve compensation. In contrast, MPC only needs to know the desired result of the process, and variables A and B will collaborate through the MPC optimizer to achieve the control target. The decoupling of MPC is not achieved through an additional "decoupler". The coupling relationship is naturally included in the optimization through the dynamic matrix of the prediction model. The optimal solution necessarily coordinates the mutual influence between all variables.

3. Misunderstandings of APC

Nowadays, some enterprises blindly believe in advanced control, thinking that after upgrading to APC, they can achieve unmanned control. However, in reality, if we compare the entire control system to a human body, advanced control methods are like our brain, achieving prediction based on data and modeling. Traditional PID is like our nerves, which have no thinking ability but have a very fast adjustment speed. And our field instruments and actuators are like eyes and limbs. Accurate instruments, good actuators, and excellent PID tuning are all crucial for the implementation of APC and are indispensable. In the wave of automation upgrades, the inability to achieve automation is often due to high instrument failure rates and incorrect instrument selection. The basic PID is not properly tuned. Even if advanced control is added, it cannot achieve the desired control effect. Modern industry still mainly relies on PID regulation. Even for multi-variable systems that can be solved by PID, APC is not considered. In production, for problems with strong coupling, large lag, and high value, advanced control can be considered. The emergence of APC does not replace PID. PID control is the foundation of APC, and APC is an optimization of PID. "Taking multiple measures simultaneously" is an inevitable step to achieve the ideal control effect. We need to demystify advanced control technology and face the wave of automation upgrades with a pragmatic attitude, rather than blindly worshiping it. Otherwise, we will end up in vain.