In an operator training simulator, an actual plant or a plant yet to be constructed is reproduced on a computer by use of a dynamic simulator, and operator training is performed by means of this virtual plant. In that sense, this consists of making a mirror image of the plant. Mirror Plant is a product that strives to take advantage of this by transferring a virtual plant from an offline to an online environment. By doing so, it becomes possible to enable optimum operation by use of new indicators, while also aiming for a stable and safe operation of the plant. Although there are prior examples of online application of steady state simulators, we do not know of any precedents elsewhere in the world of online application of dynamic simulators.
The structure of Mirror Plant is achieved by taking full advantage of the superior features of Visual Modeler: a high fidelity, high performance, and compact plant simulator that has both steady state and dynamic calculation modes.
Mirror Plant offers several possible applications. Namely, it enables to visualize internal plant conditions, to predict future conditions, to search for optimal operation conditions, and to preventively detect potential problems.
Operation significantly changes when this technology is applied. Mirror Plant simultaneously enables achieving stable and safe operation while also improving the quality of operation.
Mirror Plant is constructed with three types of simulation models that perform different calculations: a mirror model that is always tracking the conditions of the actual plant to reproduce its behavior, an identification model that performs data reconciliation and adjusts parameters, and one or several analytical models that perform predictive and analytical calculations by use of the obtained plant data.
These three types of simulation models are built in the Visual Modeler plant simulator. These models are originally based on a common single model, and for a medium scale plant they can all be executed on the same PC.
Visual Modeler is the core application, and other modules of OmegaLand are also used and assembled to the Mirror Plant system.
A model that performs dynamic simulation is used and adjusted to have exactly the same transient conditions as the actual plant. These movements are tracked in the actual plant and then reproduced on the computer. The values of primary measurements and set variables for operation are given to the model every second by way of an online connection to a DCS (distributed control system). Set variables for operation are supplied to virtual control instruments that are included in the mirror model, and the measured values of certain state variables of importance are also followed. This method of following the actual plant by means of a model is called "tracking".
Discrepancies will occur between the actual plant and the mirror model when it is executed for a long interval in real time. These are due to the variations over a long time span of performance parameters such as heat transfer coefficients and catalytic activity. In addition, sensor measurement values can also be a source of errors. These types of discrepancies and errors are minimized by the identification model: a steady state model that performs adjustments by way of data reconciliation. One particularity of the identification model is that it performs dynamic compensation for material and heat imbalances, meaning that even non-stationary conditions can be handled to some extent. This is an original feature developed by Omega Simulation, which is referred to as "data reconciliation with dynamic compensation". The values of the performance parameters obtained in the identification model are applied in the mirror model.
A variety of applications are made possible by capturing the current state of a plant from the mirror model and performing dynamic and steady-state calculations. For example, by performing dynamic simulation starting from the current plant conditions, it is possible to predict the future behavior of plant (this is called a transient state prediction). In other words, dynamic simulation makes it possible to gain knowledge of the effect of operations before they are actually performed. Also, plant conditions can be obtained by taking the current conditions as start values and performing steady state simulation to attain certain SPEC conditions (this is referred to as a steady state prediction).
One or several of these models can be executed simultaneously on a PC. The mirror model must always be running since it has to track the movements of the actual plant. In order to predict the future behavior of the plant, an analytical model for transient state prediction should always be operating at a calculation rate several tens time faster than real time. In addition, the identification model is executed at fixed intervals, and another analytical model may be used periodically to perform steady state predictions, meaning that three or more plant models may be executed at the same time. Also, the same plant model can use both steady state models and dynamic models. This takes advantage of one of the features of Visual Modeler: the simulator can switch seamlessly back and forth between steady state and dynamic calculations. Therefore the calculation results of the mirror model can be easily used in the analytical model in order to perform steady state prediction calculations, just as the performance parameters and internal conditions estimated in the identification model can be easily brought back to the mirror model. In addition, calculation results can be sent to other PCs to present the information to the operators in graphical displays.
Using the Mirror Plant structure enables several applications to be realized.
The mirror model is a mirror image of the actual plant. Because it calculates all the state variables in the model, it is of course possible to know the flow rate, temperature, pressure and composition conditions in all the piping and equipment. This gives a clearer picture of the internal plant conditions since in the case of the actual plant there are only a few measurement points. One of the applications of internal visualization is to provide new alarms for parts of the plant where measurements are not performed.
This is a function for making predictions of what the behavior of the plant will be in the future. One possible usage of this function is to repeat predictive calculations of one hour or so in order to establish operation guidelines or to take actions in advance with upcoming alarms. Another possible usage is to perform case studies to investigate appropriate operations prior to performing them.
A steady state balance can be performed to study operating conditions in advance. For instance, determining the amount of feed and the reaction conditions required to reach a certain product rate.
It may be said that optimization is the ultimate purpose of operation. Examples that can be considered include minimizing the use of energy by steady state optimization, or performing brand switching in the shortest possible time by way of dynamic optimization. Case studies can be performed by transient state prediction and steady state prediction to find the operating methods and conditions that are more suitable, and globally more rigorous optimization will be achieved simply by virtue of the higher fidelity of the model.
An instance for which the calculated values of Mirror Plant and the measurement values of the actual plant deviate can be interpreted as an indication that an error has occurred in the plant. Mirror Plant therefore makes it possible to rapidly discover abnormal plant conditions and sensor failures.
To improve the controllability by adjusting the PID parameters, an investigation can be done in advance by using the plant model before the parameters adjustments are applied in the DCS.
Training and education can be carried out by making use of the current plant conditions, and instances in which trouble and abnormal conditions occurred can be saved for training purposes.
We have seen that Mirror Plant offers a variety of functionalities, but on the other hand it may be difficult to grasp precisely in what way it can be useful. Let's have a look at the benefits that have been reported from the user side.
This is the provision of operational information to the operators. They can know in real time the values of variables that cannot be obtained by the DCS, such as state variables, energy consumption, product quality and the conversion rates. In this respect, Mirror Plant closely resembles a soft sensor system, but since it uses a physical model, it is of superior fidelity and performs better extrapolations. In addition, it can target a wide range, namely the entire plant. Also, it is possible to present a case study for finding the optimum operating conditions by considering the current state of the plant. Furthermore, by predicting and displaying future conditions, operators can know beforehand what the behavior of the plant will be. In terms of operation indicators, this can become a useful source of information.
This is also another type of information, but for alerting the operators. There are two types of alarms: alarms for the state variables that do not appear in the DCS, and predictive alarms for state variables. The latter in particular tends to help in stabilizing the plant, since irregularities can be detected earlier. If a target value is specified in the event of an alarm, a required set variable and the transient state leading to the target will be suggested. In addition, the preventive diagnosis function is useful for equipment and sensor maintenance, as it provides direct guidance for detecting failures and false indications.
Operations can be investigated by use of a variety of case study functions. According to the engineering staff, this is a powerful tool for performing operation design. Case studies can be performed either for steady state operation to determine the steady state after current value setting changes, or for transient operation to examine the response of the plant with respect to the set variables.
Veteran operators are rich in experience and familiar with every corner of their plants, but we believe that only few of them can provide explanations with a physical basis. On the other hand, the engineering staff will typically have P&IDs and documents that do not match the reality or that do not clearly trace the history of plant retrofits after the time of construction; compounded with the fact that they are kept busy by other daily tasks, it can realistically be said that there is not much time available for fulfilling the duties of plant operation design.
The process of building a Mirror Plant system provides a detailed model and an understanding of the details of the plant. Compared to the construction of an operator training simulator, the collaboration between the operators and the engineering staff is indispensable. If the engineering staff is mainly involved the construction work, it can be of great help in human resource development. It is also essential to get the operators involved in the process of tuning and online operation. Once it starts to be applied in the actual plant, the model becomes a common means through which discussion can take place between the operators and the engineering staff.
Being able to forecast what will happen and to visualize the internal state of the plant will help reduce the psychological burden of the operators. Situations can be handled well in advance with the margin provided by predictive alarms. Abnormalities can be found in the parts of the plant where there are no measurements, and it is possible to take early measures to avoid dangers and irregularities. It is conceivable that visualization and preventive operations will allow stable and safe operations of unprecedented levels. Moreover, adjustments tend to be made with frequent operations when looking only at current plant conditions, whereas operations can be conducted in a more targeted manner once predictions can be made.
There is also the possibility of obtaining operational indicators such as product rates and quality, and energy consumption. Up until now, quality and product rates were decided as a result of having performed plant operation, but by way of predictive calculation, it becomes easier to search for targeted operating conditions, or to output a planned production.
In these ways, Mirror Plant aims to sophisticate plant operation. It is a new technology that allows operators to perform operations while being aware of composition and energy, and the office personnel to plan operations in advance and in accordance with the demands of the market. Mirror Plant will completely change the concept of plant operation.
Copyright © 2017 Omega Simulation Co., LTD. All Rights Reserved.