Data Analysis and Planning
- 1. DATA ANALAYSIS AS THE BASIS FOR SUSTAINABLE LP MODELS Although LP model building has been practiced for many years and great strides have been made in terms of model complexity and sophistication refiners are still faced with models that do not reflect the performance of their plants. When should models be updated and what improvement in accuracy will be achieved and which parameters should be in the in the model? How do you ensure that the model is reflective of the unit specific to the refinery being modeled and not mostly based on a generic approach to LP model building where standardized reactor models are being used as the basis? Starting with the operating data as the basis for the model development or updating ensures that the model is intentionally built for the specific unit in question. The first step - Exploratory Data Analysis - gains insight and understanding of the data. This is followed by a Process Characterization step which determines the key process parameters and evaluates their behavior over the expected operating ranges. The final step in the data analysis is the development of models and correlations that describes the performance of the unit. The model building quality can be further enhanced through the use of reconciled data. A reconciled mass and volume balance also improves Volume Expansion Index (VEI) of the refinery. The benefit for a typical 200,000 Bbl/day refinery that improves from an average VEI to a 1 st quartile VEI is $20 MM per year. The diagram below shows the inter connections between Production Planning and Operational Analysis and Improvement. Model Error Analysis is important to discriminate between random errors and errors that have occurred as a result of a change in the process, the Process Monitoring module indicates when a shift in process behavior has occurred. QBM is used both to improve the current operation but can also be used to validate the Plan for a specific unit, i.e. how well does the Plan compare to the most efficient (benchmarked) operations. QBM is can also be used to set the energy targets for the Planned operation. Quantitative Data Recon Benchmarking (MBM) Plan Plan vs. Model Error Model Updating or Plan Implementation Actual Analysis Development Validation Production Planning RTDB Data Recon Model Updating or Model Error Plan Value (MBM) Development Analysis Validation s Exploratory Data Process Models, Process Quantitative Analysis Characterization Correlations Monitoring Benchmarking (QBM) Operational Analysis & Improvement Process Frontier Inc., San Diego, California Emails: dave@processfrontier.com, lanny@processfrontier.com, greg@processfrontier.com