Eastman: Real-Time Optimization of Utility Systems

Background

Eastman’s site in Kingsport, Tennessee uses more than 40 trillion Btu of source energy each year to meet the energy demands of the site’s manufacturing plants. Most of the purchased coal and natural gas is used to produce steam and generate electricity within four powerhouses containing 17 boilers and 19 turbine generators. There is significant variation across this equipment in terms of capacity, age, efficiency profile, fuel source, operational parameters, and constraints. Identifying the ideal setpoints for the fleet to meet current energy demands under a set of ever-changing constraints is a complex problem. Without real-time tools in place, operators must rely on general strategies that are likely to lead to suboptimal efficiency. Due to the complexity of the site utility system and its constraints, third-party software solutions were not able to provide recommendations that were realistic and consistent.

To solve this problem, a model of the powerhouse equipment was internally developed. Access to real-time setpoint recommendations enabled operators to implement changes that reduced energy use by over 2% at the Kingsport site, saving approximately 960,000 MMBtu per year and avoiding 72,000 tons per year of CO₂ emissions.

Solutions

Eastman’s Power Department and Global Natural Resource Management (GNRM) group partnered together to develop and customize an optimization model for Kingsport’s powerhouse equipment. Using information from the site’s data historian, the model can simulate the real-time mass and energy balance for the system and evaluate current constraints. A mixed-integer optimization routine is applied to determine the optimal setpoints for all powerhouse equipment within the defined equipment, environmental, and demand constraints. Several scenarios are evaluated at 5-minute intervals and results are presented to powerhouse operators through an interactive interface. Access to these real-time setpoint recommendations enables operators to implement changes that improve site-wide efficiency and reduce energy costs at Kingsport.

Plans to develop the optimization model were initiated by the Strategy Manager for Power Department and an engineer within the GNRM group. Several months were needed for initial development and to populate the model with general system information. The software was internally developed using open-source tools to build a modeling user interface. This allowed users to create a digital twin of the Kingsport site’s combined heat and power system, including equipment constraints and operating characteristics. The software, database, and dashboards were all developed using tools that did not increase external costs. The only minor cost required for implementation was for a small server that could be used to complete the automated processes and host the development user interface. Over 1,400 units across the site have been modeled so far. This modeling interface also provides the capability to evaluate changes in what-if scenarios.

Once the initial model was functional, weekly review meetings were conducted to review progress and define changes that were needed before implementation. With the model in place, software was developed to automatically carry out the processes needed to produce setpoint recommendations. These steps involve retrieval of real-time information from the site’s data historian, simulation of the system to determine the current site energy and mass balance, and optimization of the system within current constraints. A mixed-integer optimization routine is used to consider potential changes in equipment status. Key outputs from each step in this process are stored in a database for use in the operator dashboard and other reports. Each optimization run produces over 5,000 data points that may be compared to current and simulated values.

The System Manager for the Power Department was included in the weekly meetings to review development progress, identify improvement opportunities, and lay out an implementation plan. Additionally, powerhouse shift managers were included in the discussion and provided access to the recommendations. More direct implementation of recommendations at the operator level continued to be phased in and an operator dashboard was developed using software commonly used within Eastman. This dashboard contains a live connection to the database to display the latest setpoint recommendations and updates at 5-minute intervals. The dashboard also provides real-time data for potential constraints and advanced energy analytics for utility use across the site. These tools help to identify inefficiencies throughout the system and provide guidance on specific actions that may be taken to run at peak efficiency within current system constraints. Mobile versions of each dashboard were also created.

The Power Department and GNRM group continue to communicate and identify improvement opportunities. For example, a special interface was recently developed to allow Power Department staff to easily enter temporary equipment or system constraints without modifying the core model.

Other Benefits

The modeling success of Kingsport’s Power Department is currently being replicated for secondary utility systems that supply compressed air, cooling water, refrigeration, and heat. Many of these systems are also supplied by a wide variety of equipment with complex constraints. The capability to include additional systems in the modeling alongside the powerhouse equipment ensures that the setpoint recommendations provide optimum performance for the site. Once complete, it is expected that coordinated implementation of setpoint recommendations across all site utility systems will lead to a larger reduction in source energy consumption. Future additions to the model may include setpoint recommendations for selecting manufacturing equipment.

Eastman is also evaluating the opportunity to replicate the optimization model at other sites. Many chemical plants have energy-intensive processes or utility systems where optimization may help to significantly improve efficiency. This is particularly true for large, complex systems where the optimum mode of operation is not obvious or when a standard operating strategy will not cover many of the operating scenarios that are encountered. Since the functionality of this model may be developed using open-source software and customized to fit the needs of the company, a similar approach could be employed at many plants across the industry.

ANNUAL ENERGY USE

ANNUAL ENERGY COST

ANNUAL WATER USE

ANNUAL WATER COST

ANNUAL PLANT ELECTRICITY USE

ANNUAL PLANT ELECTRICITY COST

IN BUILDING ENERGY USE

IN BUILDING ENERGY COST

ACROSS NETWORK USE

ACROSS NETWORK COST

ANNUAL EMISSIONS

USE

COST

ANNUAL WASTE COST

ANNUAL WASTE USE

Eastman: Real-Time Optimization of Utility Systems

Related Content