1.       Introduction

Advanced control is centered on soft sensors and multivariable model predictive control, and contributes to generating profits by continuously realizing stable product quality, improved product yield, and energy saving.

Yokogawa Electric Corporation (Hereinafter referred to as our company) offers an advanced control solution "Platform for Advanced Control and Estimation" (Hereinafter referred to as PACE) that realizes engineering efficiency and improved maintainability after its introduction.

In this paper, we focus on the improvement of maintainability after the introduction of advanced control; and introduce the functions of PACE that can effectively and efficiently maintain controllability. We will also show that how PACE could contribute to operation improvement and innovation of the plant in future.

2.       Effect of advanced control

The mechanism by which the effect is obtained by the introduction of the soft sensor and the multivariable model prediction control will be described with reference to FIG. 1. Generally, the product quality in the process is controlled only in the laboratory analysis, and often the operation with the margin of the quality allowance High limit is carried out. The soft sensor estimates properties such as flash point, freezing point, and amount of impurities that are difficult to measure, namely quality, from measurable process values such as temperature, flow rate, and pressure, and realizes visualization of quality in real time ((1) Visualization).

Although PID control is generally used for controlling the process, it is difficult to control the process near the operation target with PID control for the process having a large dead time or mutual interference. The multivariable model prediction control has a model expressing dynamic characteristics of a process, inside a controller, and performs multivariable control based on the prediction by the model, and can stabilize a process which is even difficult with PID control. By using the soft sensor, quality can be treated as a continuous value like temperature, flow rate, and pressure, so that the quality conventionally obtained as a result of operation can be controlled ((2) Quality Control) and stabilized ((2) Quality stabilization).

Furthermore, when the control deviation becomes small, there is a margin for the product quality to reach allowable Upper limit, and the quality operation target value can be operated close to the product quality allowable Upper limit. In multivariable model predictive control, it is possible to obtain an operation target by linear programming while achieving constraint conditions according to priorities, and to realize optimal control to push to the target ((3) Optimum Operation).

The stabilization of the product quality, the improvement of the product yield and the energy saving can be realized by (3). In addition to these economical effects, the stable control provides advantages such as reduced alarms during steady operation, reduced wear of control valves, and reduced manual operation of operators.

3.       Challenges of advanced control

Advanced control has been applied in many process industries, mainly in the oil, petrochemical, chemical, and gas industries, and has been demonstrated to stabilize product quality, improve product yield, and save energy.

On the other hand, one of the problems of advanced control is the difficulty of maintenance after the introduction of advanced control. At the beginning of the introduction of the advanced control system, the effect is as expected. However, when the system is operated for a long period of time, the model error for the process increases due to aging of the plant equipment, facility changes, or changes in operating conditions, and the controllability of the advanced control system deteriorates, and the desired effect may not be achieved. Therefore, the model needs to be updated to reduce the model error.

However, in order to update the model, it is necessary to perform a step test to obtain the DATA of the dynamic response of the process, to identify the model using the DATA obtained in the step test, and to reflect the model in the advanced control. Conventionally, since the control is stopped during the step test, the effect of the advanced control cannot be enjoyed. In addition, since the process is brought to a steady state and the manipulated variable to the process is manually changed to the step, it can be said to be an inefficient work which consumes much time and labor. Therefore, when the step test cannot be performed, it becomes difficult to update the model. If the error of the model is further increased, the controllability deteriorates due to the advanced control, therefore advanced control may have to be stopped.

In view of the above, it is necessary to update the model without stopping the advanced control, and to maintain the controllability effectively and efficiently.

4.       PACE features

In order to solve the above problems, PACE has a function to update the model without stopping the advanced control and without spending time and labor, and to maintain the control effectively and efficiently. Here are five features of PACE:.

(1)     KPIs and Health Status Message Features

The KPI quantifies the operating state of the advanced control from 0 to 100%, and it is possible to grasp the operating rate quantitatively. In addition, the Health Status Message indicates the health status of the advanced control (Healthy or Unhealthy status) and the reason during unhealthy state, and hence it is possible to determine the cause of the low utilization rate.

(2)     Past forecast Trend function

PACE incorporates historians, which automatically store the actual DATA of advanced control and can be used to analyze and improve controllability. In the past, it was not easy to evaluate the error of a model, but in PACE, the Trend of predicted value at any time in the past can be displayed, and the error of a model can be evaluated by comparing the Trend of a measured value with that of a predicted value by superimposing them. In other words, the larger the difference between the predicted value and the actual value of Trend, the greater the error of the model with respect to the process.

(3)     Automatic Step Test Function

The PACE has an automatic step Test Function and can automatically perform a step test to obtain the dynamic response of the process when the operation variable is changed stepwise while maintaining the constraints of the control variable while operating the advanced control. This allows step testing to be performed with dramatically reduced user manual effort while minimizing process impact thus, maintaining the effectiveness of advanced control. The resulting DATA of the dynamic response of the process can be used to reidentify the model.

(4)     Online application switching function

In the past, it was required to stop advanced control application to change or replace the running model; In contrast to that,  PACE supports two Operation modes: Staged (Read Only) and Live (Read and Write), which can be swapped without stopping advanced control hence having a zero downtime.

Following steps are involved: -
　1.       Parameters which are running in Live mode can be copied to the application in Staged mode from Reconciliation view.
　2.       The Staged application is then monitored to observe the controller behavior.
　3.       If the control behavior is as expected, then application mode is changed from Staged to Live, in this way the application
runs without big impact as parameters are already copied.
　4.       The previous Live application will be stopped and changed to Archive mode.

(5)     Online model update function

Previously, only the gain of the model could be changed online, but PACE can also change the type of transfer function, time constant, and dead time. For example, even if the model is defined as the model of the first order delay + the dead time, if the model of the second order delay + the dead time is confirmed to be valid by an automatic step Test Function, then the model can be changed online to the model of the second order delay + the dead time.

5.       PACE Case Study

The introduction case of the automatic step Test Function of PACE mentioned above is introduced. A model to estimate the properties of LVGO and HVGO was reidentified in Process Test Function obtained using an automatic step DATA system over 8 days in a vacuum distillation unit of an overseas petroleum refining plant.

The Trend in Fig. 6 is a comparison of measured and estimated values of the PCT (temperature with pressure compensation) used to estimate the 95% distillate temperature of HVGO.

The Trend consists of four time zones, and the comparison results for each time zone are shown below.

・The time zone on the far left is the comparison between the estimated value by the model at the time of introduction of advanced control and the measured value. There is a discrepancy between the estimated value and the actual value.

・The second time zone from the left is the comparison between the measured value and the estimated value of the model reidentified using the DATA of the manual step test at the introduction of advanced control and the DATA of the first automatic step test. There is also a discrepancy between the estimated value and the actual value.

・The third and fourth time zones from the left are comparisons of measured values with models re-identified in DATA for all automated step tests. There is little discrepancy between the estimated and actual values.

From the above results, it can be said that the deviation between the estimated value and the measured value is smaller and the error of the model is smaller in the model re-identified by using more DATA of the automatic step test. Since the automatic step Test Function performs step tests automatically without stopping advanced control, it is possible to perform step tests for a longer period of time by reducing the impact on plant operation and dramatically reducing the user's labor. Thus, a model with a smaller error can be reidentified.

6.       Outlook for the Future

Future prospect of PACE for improvement and innovation of plant operation is introduced. In 2019, our company launched the "OpreX Dynamic Real Time Optimizer" (RT-OP), an operational optimization support solution designed to improve plant productivity and maximize profits, in collaboration with PACE, to contribute to the optimization of plant-wide operations. The features of RT-OP are: a) to reduce deviation from optimum operation due to frequent updating of operating conditions (5 to 10 minutes), and b) to simplify maintenance by the same platform as PACE. By further integrating the RT-OP and PACE platforms, we will contribute to the optimization of plant-wide operations.

7.       Conclusion

This paper focuses on the improvement of maintainability after the introduction of advanced control, which is one of the issues of advanced control, and introduces the functions and introduction examples of PACE, an advanced control solution in our company that can maintain effective and efficient controllability. Our company aims not only to improve maintainability after the introduction of advanced control, but also to further improve engineering efficiency up to the introduction of advanced control. By further expanding the functions of PACE and cooperating with RT-OP, our company will contribute to the improvement and innovation of plant operations required in response to changes in the economy and  plant site environment.

*The company name, organization name, product name, logo, etc. used in this text are registered trademarks or trademarks of Yokogawa Electric Corporation, each company or organization.

▶Source: Monthly Keiso, May 2020 (Kogyogijutsusha)