Abstract
This paper covers the fully implemented solution to forecast facilities’ power demand based on production flow rate.
Objectives:
Share how a systematic approach will forecast the demand and optimize resource allocation.
Highlight how this approach enhanced reliability, planning, and, most important, the environmental impact through reducing the carbon footprint.
Power consumption planning plays a crucial role in determining the resources required. An efficient and accurate forecast will be key to optimizing facility resources and support decision- making processes. The tool was developed utilizing a machine learning algorithm to predict the power consumption forecast for any given production flowrate. Capitalizing on the big data that we have stored over the years, the model was built to anticipate the total power consumption for the facilities, taking into consideration shutdown periods for some the equipment, turnaround periods and many factors. The design of the system is on historical data and multiple scenarios that we were run through over five years. The approach to come up with the model started with data exploratory phase, where we utilized unsupervised learning to study our data. Before that, the data went through a lengthy process of data cleaning, to explore the correlation between the variables. After understanding the correlation and the factors that play major roles in the model, the right algorithm was picked and power consumption modeling finalized. Currently, the model is providing the power forecast for the business plan, as well as to capture optimization opportunities within process. Capitalizing on big data and smart infrastructure leveraged us to build a smart forecasting tool that will have a great impact in systemizing planning demand.
As the power demand for any industrial facility usually takes the biggest share of the budget and sharply varies, based on several dynamic variables. Calculating the estimate for the power requirement is a challenging task. This paper provides clear insights on how to accurately forecast the power requirements, by using a machine learning model that conducts deep analysis for the facility historical power utilization trend. The implementation of this smart forecast tool will result in realizing several advantages:
Accurate estimate for the industrial facility power budget, which will support decision making to allocate the required budget efficiently and optimize facility resources.
Identify energy saving opportunities by comparing the actual power consumption with the optimum one generated by the forecast tool at the same given operations circumstances.
Save man-hours as the tool analysis doesn't require human intervention.
This paper describes an innovative systematic approach to forecasting the power demand, using a supervised learning technique to accurately predict the power consumption requirements. The basic concept of this approach can be applied not only in the petroleum industry but extended to other industries.