By Jim Sutton, Senior Manager, GE Power Boiler Services
Peter Spinney, Product Manager, GE Power Digital
First published in Cornerstone, Volume 4, Issue 2
Countries around the world face a tremendous challenge in providing ample clean water, sustainable food supplies, and jobs to their citizens, while protecting the environment. Central to this challenge is managing and improving the power production infrastructure. Today, and for the foreseeable future, coal-fired power plants play a pivotal role by providing low-cost electricity to much of the world. Natural gas and renewables are growing in importance and are changing the ways in which traditional power plants operate. The rate of change in the electricity production business is unprecedented and is creating new opportunities for digital, interconnected, more intelligent power plants that are better able to meet these new requirements.
GE believes digital solutions will provide this intelligence, transform the industry, establish new business models, and create unprecedented opportunities to address global energy challenges. Over the next decade, the International Data Corporation has projected that approximately ˜US$1.3 trillion of value will be captured as part of this transformation.With software and data analytics, combined with advanced hardware, new digitally enhanced power generation will deliver greater reliability, affordability, and sustainability. This can help lower costs, improve efficiencies, create growth opportunities, and reduce CO output.
Advanced analytics can dramatically improve operations in coal-fired power plants by reducing fuel consumption, improving reliability, and reducing emissions. If deployed at every existing coal-fired power plant globally, this new equipment-agnostic technology, “Digital Power Plant for Steam” software, could eliminate 500 million metric tons of greenhouse gas emissions—the equivalent of removing 120 million cars from the road.
TECHNOLOGY AS IT STANDS TODAY
Modern coal-fired power plants rely on a complex network of sensors, actuators, digital controllers, and supervisory computers to operate and coordinate each of the plant subsystems. Hundreds of feedback control systems serve to monitor plant processes and perform appropriate control actions, aiming to maintain optimum operating conditions regardless of system disturbances, such as changes in coal quality or electricity demand. However, the highly interactive nature of power plant parameters—where one parameter can affect many others—means that control is highly challenging, and plants are often not operated to their potential capabilities.
Power plants utilize a distributed control system (DCS)—an automated system that monitors and coordinates different parts of a power plant—to start and keep the plant running amid these changes. DCS is the most commonly used method of controlling the components in a modern plant, having replaced pneumatic, analog, and discrete controls.
DCS-enabled power plant controls perform quite well. Although advanced control is possible, most power plant DCS implementations use a basic scheme known as proportional-integral-derivative (PID) controller. A PID controller continuously calculates an error value, defined as the difference between a measured process variable and a desired set point. The controller attempts to minimize the error over time by adjusting a control variable—such as the position of a control valve, a damper, or the power supplied to a heating element—to a new value based on a mathematical algorithm. This PID control algorithm does not require information about the power plant operational process; it simply reacts to errors and adjusts the controlled elements to minimize errors over time.
The main disadvantage of this controls approach is that it is difficult to implement for the process of optimizing multiple variables. For example, a power plant operator may hope to reduce NOand CO while improving heat rate and superheated steam temperature balance. To achieve this, DCS suppliers have included the provision for operators to “bias” the controls. In this way, basic controls continue to operate well, but power plant operators are able to use their knowledge of the process to fine-tune the controls to meet their operational goals.
THE PLANT OF THE FUTURE
GE envisions a more comprehensive analytic solution that builds on historic plant DCS and data historians to deliver improved outcomes for plant efficiency, low emissions, and reliable generation. In a recent GE survey of over 100 power generation executives, 94% of those surveyed believe that the internet and improved analytics will transform their industry in the coming years.GE fully agrees and is developing both an overall web-based computer analysis environment (or enterprise tool), called Predix , and the individual applications that will provide the improvements. Predix is GE’s operating system for digital analytics for large machines that will manage data and supply tools to allow developers to easily create beneficial software applications. GE is now delivering many Predix applications. For the power industry, and coal-fired power plants in particular, some Predix applications are described briefly below.
Asset Performance Management (APM): A power plant consists of many assets, such as a boiler, a generator, a turbine, a boiler feed pump, or a coal pulverizer. Each asset has condition data that is already being measured and recorded. The goal of this application is to transform machine sensor data into actionable intelligence by combining robust analytics and domain expertise. This predictive information drives toward the ultimate goal of zero unplanned downtime and an optimal maintenance schedule.
Operations Optimization (OO): In any power plant, each of the plant assets must work together to accomplish the overall goal of efficient production at the system level. The goal of OO is overall improvement in client operations with performance visibility across power plant and fleetwide footprints, providing a holistic understanding of the operational decisions that can improve efficiencies, reduce emissions, expand capabilities, and lower production costs. Some of these optimizations can be performed immediately by local interfaces with the plant DCS.
Business Optimization: With the increase in complexity of maintaining a stable generating grid, many regions are requiring power producers to correctly forecast and price the power being produced. This is a challenge for the operations team, who may have limited tools. This application provides intelligent forecasting and portfolio optimization to enable trading and operations teams to make smart business decisions that reduce financial risk and maximize the profitability of the fleet.
Cyber Security: GE’s advanced defense system is designed to assess system gaps, detect vulnerabilities, and protect the customer’s critical infrastructure and controls in compliance with the various national-level cybersecurity regulations.
Advanced Controls/Edge Computing: This application allows plant operators to leverage data and analytics to manage grid stability, fuel variability, emissions, compliance, and other challenges that affect machine performance, as well as to execute fast starts and efficient cooldowns to meet dispatch and market demands. The ability to perform in this manner will be critical in a world that maximizes renewable power sources with their inherent variability.
Predix: As previously described, Predix as an overall platform allows application developers to safely and securely access plant information and build apps to improve any aspect of system performance. Predix is an open architecture that allows both GE and independent developers to use built-in analytic tools to quickly build the apps.
Predix Operational Optimization for Boilersis one of the many smart technologies being developed and offered by GE that could be useful for many coal-fired power plants and is explored in greater depth below.
Today, boiler modernization is focused on more than NOreduction or heat rate. Goals are varied and the systems are asked to address a more diverse problem set. For example, boilers are challenged to control emissions, but also to deliver improved fuel efficiency and integrate complex air-staged combustion systems with different types of air quality control systems, such as complex combustion systems, selective catalytic reduction (SCR) systems, or selective non-catalytic reduction (SNCR) systems. Operating envelopes have expanded, intermittent renewables are increasing, and coal-fired power plants are being asked to ramp faster and also to ramp down to lower electricity output than ever before.
Predix Operational Optimization for Boilersis an analytic system that models how power plants respond to various inputs. Understanding how the interrelated systems interact allows a software solution to provide control biases—or set-points—to the DCS that improves performance. The software runs on a server at the power plant and communicates directly with the DCS. A coal power plant has multiple objectives: limiting plant emissions, achieving certain steam temperatures, while ensuring that the power plant operates as efficiently as possible. Predix Operational Optimization for Boilers understands multiple objectives and reacts much faster and more accurately than a human operator can because of the complex interactions and volume of data that must be assessed.
The basic configuration of the application is shown in Figure 1; the system builds models of power plant performance that predict what the plant’s state will be for the given set of inputs. As one example, the application is able to predict the likely values of the gaseous emission CO based upon current operating conditions. Predix Operational Optimization for Boilerscan also model the impact of the various parameters on heat rate. With this understanding, the software issues optimal bias-to-set-point signals to the DCS that both improve plant heat and ensure that CO emissions do not exceed plant requirements. This is a considerable improvement over traditional plant controls that typically do not take CO into account or include heat rate as an explicit optimization target.
Fundamental to the operational improvements offered by the software is the ability to build mathematical relationships that model the process behavior. Fast, optimal adjustments can be made using an accurate predictive model of the boiler processes. Several types of models are used in the Predix Operational Optimization for Boilerssystem, as summarized in Table 1.
More than 120 installations of this boiler optimization technology have been installed and continue to be supported by GE on coal-fired boilers. Most of these clients are top utilities in the U.S., although there is now significant growth internationally. One example of a successful installation occurred at Calaveras Power, JK Spruce Station Unit Number 1, located near San Antonio, Texas. The unit is a 600-MW tangentially fired coal unit originally manufactured by GE in 1990. The Ovationcontrol system was manufactured by Emerson.
The results from the optimization are summarized in Table 2, where several key performance indicators (KPIs) are listed at full load conditions. The first two KPIs are heat rate and boiler efficiency. Heat rate measures how much heat input from the fuel is required to produce a kWh of electricity. A lower heat rate is better as it means less heat input (less fuel) is required to produce the same amount of electricity. Examining this KPI in the table in more detail, the next column lists the average heat rate achieved by the plant during periods where the optimizer was not running. The next column lists the average efficiency achieved with the optimizer turned on. As can be seen, with the optimizer turned on, the plant was able to achieve a lower net plant heat rate of 104 Btu/kWh, or a 1.08% improvement in heat rate. The second KPI focuses on the overall boiler efficiency, which looks at boiler performance separate from the overall plant performance.
The next two lines in Table 2 describe KPIs associated with gaseous emissions. Significant improvement was made in reducing NO, a key objective of the optimizer. The average value of CO emissions increased somewhat but CO was reliably maintained below the target maximum value of 125 ppm. The next KPIs are related to the actual superheated and reheated steam produced by the boilers. With the neural net in operation the plant was able to get close to the ideal steam temperature for this plant, 1005°F, without exceeding it. Particularly important is the reduction in the need for reheat spray flows. Reheat spray flows deteriorate plant efficiency. This KPI was improved by 4.25%. Superheat spray flow was increased by 6.99 klb/hr, but this does not impact plant efficiency in the same way as reheat spray.
Figure 2 shows how operation heat rate varies with the optimizer on versus off at several different loads. With the optimization software off (green), the heat rate is the highest, meaning the boiler was the least efficient. With the software on (blue) the heat rate was improved. As can be seen in the figure, improvement was made at all loads but the biggest gains were at high loads and low loads.
These results were achieved by maintaining proper fuel and air control, which is critical to achieving efficient combustion and cost-effective compliance with environmental regulation. Figure 3 shows how the software can improve efficiency while reducing NOemissions by controlling the fuel/air ratio.
The availability of digital solutions that optimize performance now allows GE to make more comprehensive upgrade offers that give greater value than solely providing the equipment. One example of this is low-NOupgrades. Previously, GE had been limited to providing OEM low-NO burner, SCR, and SNCR hardware as stand-alone packages with manual commissioning services. For low-NO system upgrades, over 700 burner upgrade projects have been commissioned, and over 100 upgrade projects have been successfully implemented for post-combustion SCR and SNCR. With the robust closed-loop optimization system, it is now possible to offer full packages with SNCR and SCR that not only minimize NO , but allow optimization of boiler heat rate, while minimizing the cost of sorbent needed in the SCR or SNCR.
The robust nature of the neural network software also allows GE to identify wear in key components, such as coal pulverizers. This ability to understand the condition and operating circumstance allows GE to offer more comprehensive multi-year service agreements that provide specified performance and reliability assurances.
Coal-fired power plants are an important part of the global infrastructure to produce electricity and face increasingly challenging operating requirements. In high-growth areas such as China and India, new high-efficiency coal plants are coming online. There is a real need to ensure that these plants, along with the massive installed base, operate with maximum efficiency and minimum emissions throughout their lifetime. Additionally, coal-fired power plants face the increasingly challenging generation mix that includes renewables. More renewables on the grid mean that power plants will need to ramp up and turn down as never before. Tighter emissions regulations also mean that more variables must be tightly controlled. Tomorrow’s smarter power plants can simultaneously address these multifaceted challenges with digital controls.