You are currently viewing: Articles



Jun-2017

Building a refinery monitoring system

Developing a process network and unit monitoring system that identifies significant changes in process parameters

SOUMEN MANDAL, P SRIDHAR, SOWMYA CHALLA, A N JADHAV, Deepak KEWLANI and Arnab BANERJEE
Indian Oil Corporation
Viewed : 153
Article Summary
Process unit monitoring and network monitoring in an oil refining/petrochemicals complex requires an understanding of the significance of any changes that occur in process parameters. One can understand this by maintaining a continuous effort to detect changes. However, the volume of effort required to monitor every change in the process parameters of a complex plant, and the need to dedicate an individual’s time to the task, are impractical options. A system which can incorporate expert knowledge to infer the significance of changes that have happened and also to propose remedial action is a more effective approach.

The intrinsic nature of a parameter
Every process parameter has its own characteristics. For instance, a parameter such as a pressure change by 0.01 kg/cm2 g might be significant, whereas for a parameter like steam flow a change of 
10 kg/h might be a relatively insignificant amount.

A processing network and unit monitoring system needs to incorporate the parameters which affect the quality and quantity of production streams. Typically, a network monitoring system must continually monitor 100 or so parameters, each parameter being different in nature (pressure, temperature, flow, and so on) and require logarithmic expression to plot in a single graph.

However, in steady operation barely 5% of these parameters is disturbed. Identifying changes or disturbances in the system requires the elimination of the steady state parameters.

A process parameter represents a process characteristic, hence a change in a process parameter represents a change in the process. Every process parameter has its own sensitivity (standard deviation). This standard deviation accounts for all of the inherent characteristics of the process. Here, inherent characteristics means disturbance residence time, quality of control action, and so on.

Changes in the flow rates of process streams which are to used for network monitoring can be inferred as follows:
Flow deviation = one week’s average value – current value

A significance process change can be inferred by the ratio of a change that has occurred to its standard deviation and this can be used to measure how many times a parameter has differed from normal process operation. This is illustrated in Figure 1.
Thus we can define a process deviation factor (PDF) as follows:

Flow deviation and process deviation factor are required in network monitoring to capture the quality and quantity of changes to parameters.

Effectiveness of process deviation factor
Trivial changes should be ignored and changes of actual significance are to be processed. This can be understood by means of Table 1 in which trivial Point C can be ignored and Points A and B can be focused upon.

Developing effective monitoring systems
Zeroing down to the disturbance parameters quickly is essential to understand the real picture. Here a dynamic system is required based on real-time data as a monitoring aid. Online system data can be imported directly from a real-time database in the form of a spreadsheet. Process unit/network 
dashboards can be developed as follows:
1. Current data for process parameters can be retrieved on a 15 minutes average basis to make it dynamic.
2. Standard deviations of each parameter are to be obtained/calculated (based on data obtained during the system’s steady condition).
3. Process deviation factor to be calculated for each parameter.
4. Process stream flows data can be retrieved on a previous hour average basis to make it more reliable. This data can be sorted on the basis of any deviation from the previous one week average. Based on this flow deviation, system flow disturbances can be identified.
5. Flow deviation and process deviation factor are to be sorted on the basis of order of magnitude.
6. The reliability of real-time data is the basic need of this system. Data fetching quality should be incorporated in the monitoring sheet
Continual updating and automatic sorting of these data in a spreadsheet facilitates easy and one click understanding of network changes in a process or process unit. Sample monitoring dashboards for automatic data sorting are shown in Figures 2 to 4.

Conclusion
1. At present, there is no reliable and effective method to determine changes that have happened in a process network.
2. Some process units need to be adjusted to allow for changes occurring in the upstream units, to avoid disturbance to the process.
3. Typically, enough time is available to make adjustments to a downstream unit to avoid disturbance, but lack of communication prevents actions which can be taken.
4. The method proposed here helps to monitor a vast process network of LPG/naphtha/kerosene/diesel in a refinery. Typically, these streams are produced in different process units.
5. With the proposed method, a member of staff responsible for process network monitoring can understand what changes have occurred in different units by a single click action as all process parameters are normalised to a process deviation factor.
6. It is possible to monitor a system effectively without having a complete knowledge of the system.
Current Rating :  1

Add your rating:



Your rate: 1 2 3 4 5