Disadvantages of the PID control algorithm
Thread Content
This post was last edited by Renzhengyun on 2022-3-2 at 12:45. The PID algorithm has many advantages and is suitable for most industrial processes. However, for processes with high capacitance, integration, and large pure lag, no matter which tuning method is used (self-tuning, internal model control, fuzzy control, or other artificial intelligence-based tuning methods), even if the PID parameters are theoretically optimal, the PID control algorithm has significant drawbacks: 1. Slow resistance to disturbances and large fluctuations ； 2. The rising speed should not be too fast, otherwise overshoot will occur ； 3. Long tuning time ； 4. After the controller output saturates, it takes a very long time to return to normal control ； 5. The control accuracy is not high at steady state; the controlled variable oscillates at a certain frequency, with the amplitude varying sometimes high and sometimes low ； 6. Poor adaptability: control performance deteriorates, or even worsens, when operating conditions and set values change. The chemical industry features many processes with high capacitance (such as sensor plate temperature) and integral processes (such as tower liquid level and tank pressure). Properly addressing the control issues of these processes greatly contributes to improving the degree of automation in chemical plants. Sometimes, in the face of severe disturbances, PID automatic control is not as fast as the manual adjustments made by experienced operators. I hope everyone can discuss how such control problems are addressed in chemical processing processes. In terms of tuning methods, most PID controllers rely on empirical formulas, without comprehensive consideration of performance. So far, the use of differentiation has been limited, and there is no unified or reasonable approach; some people prefer to use differentiation while others do not. Based on my recent work, the role of differentiation is extremely significant; it is currently underestimated. Appropriate differentiation structures (standard differentiation, differential precedence, incomplete differentiation, etc.), along with parameters, have a great impact on the performance of systems.
This post was last edited by Renzhengyun on 2022-3-2 at 12:44. For such systems, we generally use the Universal Control System; this system is suitable for various complex chemical processes. It combines the advantages of PID and predictive control while overcoming their disadvantages. It is easy to implement, features a self-tuning mechanism, and offers advantages such as plug-and-play operation, fast tracking and interference resistance, good robustness, and high precision.
This post was last edited by Renzhengyun on 2022-3-2 at 12:44. Universal Control System: https://www.bilibili.com/video/BV1cy4y1G7J2
New things. “For such systems, we generally use a universal control system... Cloud servers, public internet – how important is security? For cloud servers, how does the system respond in the event of an external network failure or a network attack? The petrochemical industry must consider safety, real-time performance, and the confidentiality of user data. How is it charged?
You are absolutely right! The cloud platform is merely a means for debugging. As we all know, due to the pandemic over the past couple of years, it has been very inconvenient to travel to sites for on-site project debugging. By establishing this system, we can first check in the short term whether the system is controllable, what the level of control precision is, and if remote control works well, then on-site debugging will yield even better results (with potential delays reduced), thereby **reducing the time required for on-site debugging. With this system, on average, on-site commissioning can be completed in 1-2 days; sometimes it even takes less time (1-2 hours to get it operational). We can also adopt various measures (dual networks, output limiting, automatic removal of abnormal conditions) to ensure real-time performance and security. Currently, a high-temperature furnace system is controlled via remote operation; it has been running continuously for over 200 days, and no abnormalities have been detected in the system, with the temperature control accuracy remaining within plus or minus 0.2 degrees. Of course, this approach still poses security risks; it’s best to carry out debugging and operation on-site if possible
What’s most crucial for us is still the control algorithm itself; let’s communicate more in the future. Thank you very much for your attention!
I don’t really believe in a “universal control system”; if such a system existed, automation would be far too simple! ! !
Yes, process objects are highly complex (there are issues such as detection problems, delays in actuators, process defects, and equipment abnormalities), requiring ongoing inspection, improvement, and enhancement. We have also successfully implemented several projects using this algorithm, so we have a certain foundation. If anyone has a system that is difficult to control, they can remotely connect to our system (as long as the host computer of the system to be controlled has internet access) in order to carry out control and check the performance; practice is the only criterion for testing effectiveness.
Theoretically, if large disturbances can be measured, good control results can be achieved by using feedforward (dynamic and static) PID and model predictive control. In practice, however, many large disturbances are unmeasurable, and in some cases it is even unknown where they originate, which poses significant challenges to control systems. Our system has the ability to rapidly suppress unpredictable interference and exhibits good dynamic performance.
We also need to continuously improve the algorithm, with the aim of replacing the PID loops and conventional complex loops used in current PLCs and DCS systems. This will help reduce tuning time, improve tracking speed and steady-state control accuracy, as well as enhance the ability to suppress disturbances and improve robustness.
In summary, control methods can be divided into three categories: feedback control based on black-box theory (PID), model-based control (MPC, DMC, etc.), and data-based control (MFC, MFAC, etc.). Each type of control has its advantages and disadvantages, and is suitable for different situations; no single control method can solve all control problems.