PREDICTION TOOL FOR CONTROLABLE BOILER LOSSES
Design a tool predicting the improvement potential of coal power plant efficiency, based on historical operation data of boilers, to be used by GE Power sales teams in demonstrating BoilerOpt product benefits to their clients.
In a coal-energy market where emissions regulations are ever more stringent, the ability to improve heat rate on units is critical as utilities strive to update their generating capacity with renewables, from both cost and emissions viewpoints. GE’s BoilerOpt product has the ability to improve heat rate, which is the energy that must be expended in order to obtain a unit of useful work (Heat Rate is the inverse of efficiency), by dynamically modifying boiler parameters in real time, with more granularity and with a more data driven approach than plant operators tend to use. The problem in selling BoilerOpt to potential customers is that every plant is operated differently. For two identical plants, with two very different operators, BoilerOpt may achieve significantly different results at one than the other.
Similarly, for two very different plants but the same operator, BoilerOpt will also likely achieve significantly different results. In order to effectively advertise and sell the product, it is necessary to estimate the magnitude of any benefits which BoilerOpt may be able to provide.
The six controllable parameters which BoilerOpt can exert some degree of control over are: excess air, air heater exit gas temperature, superheat steam outlet temperature, reheat steam outlet temperature, superheat attemperator spray flows, and reheat attemperator spray flows. While the qualitative correlation between each of these individual parameters and plant heat rate are relatively well understood, there are also significant interconnections between the six individual parameters, making the task of estimating change to each parameter and the resulting impact on heat rate overall difficult. Presently, the sales team uses historical operating data of the plant in question to predict potential opportunity available in installing BoilerOpt. This is achieved by computing the loss in heat rate due to the difference between the actual values and values that the unit should be able to achieve. The individual impacts are then added together to provide an estimate of the total potential opportunity for the plant.
Data driven analysis of coal power plant heat rate factors is a key component of selling GE’s BoilerOpt product to potential customers. Given a more robust mechanism for predicting the opportunity for improving heat rate at a given coal power plant, sales pitches to potential customers can be tailored for the specific plant in question, and both the customer and GE can enter into contracts which are mutually beneficial and likely to yield significant improvements in heat rate.
The overall heat rate of a coal fired boiler is influenced by a variety of factors, many of which are not well understood. GE’s BoilerOpt product is able to effect changes to a small subset of these factors, ones which, individually, have a relatively straightforward relationship to plant heat rate.
The main challenge is estimating by how much BoilerOpt will be able to change these factors. For example, if excess air varies between 3% and 4% with the mean at 3.5% and one standard deviation being +/- 0.4%, what will the mean value of excess air be when BoilerOpt better manages the air-fuel mix. Given the significant variation in excess air, it seems clear that there is some room for improvement, but it also seems like there is almost no chance that operating the unit perfectly can drive the mean value of excess air down to 3%. The challenge is to figure out, given the nature of the distribution of excess air values along with other information about current operating conditions, what the most likely mean value is.
A second challenge is how to consider interrelationships between the controllable parameters.
As an example, the reheat steam outlet temperature and reheat attemperator spray flows have a strong relationship to one another. The more frequently the steam temperature exceeds a plants design value, the more spray flow will be seen in the attemperators to cool the reheat section. However, if it is cooled too much, then valuable steam energy is being lost, as unlike superheat attemperator sprays, the reheat sprays cannot be recouped. While the contribution of both of these to plant heat rate can be expressed qualitatively (increases in attemperator sprays increase the heat rate, increases in reheat steam temperature decrease the heat rate), the goal is to quantify to what extent BoilerOpt might be able to optimize the two individually, and how that would impact the overall plant heat rate.
Considering this, the desired outcome for this challenge is a web based application which would allow salespeople and engineers within GE to take data provided by customers without a BoilerOpt installation and instantaneously generate a figure for the potential opportunity from installing BoilerOpt at the site.
This application would be able to predict possible improvement for each of the six controllable parameters (excess air, air heater exit gas temperature, superheat steam outlet temperature, reheat steam outlet temperature, superheat attemperator spray flows, and reheat attemperator spray flows) and total improvement expected. Additionally, it should provide the probability to reach the predicted improvements.
In this case, Predix is readymade for you to manage and implement the most meaningful analysis and help identifying key insights and trends. Predix is built for remote monitoring, diagnostics, and control of industrial assets. Predix Asset, Predix Time Services and other analytics and services are available for you to deploy the best web based application to help salespeople and engineers
Due to the nature of BoilerOpt installations, any success indicators will only be available after installation and sometime of operating the system. The success of BoilerOpt at its goals has been measured by performing on-off tests. These generally lasted one to two weeks, where the system was online for half of the time. Ideally the operating conditions of the unit were held constant, though due the nature of coal power plants, this was rarely the case. Data was filtered to provide for similar operating conditions, and the key metrics were computed in the two different cases.
The intention is to continue measuring BoilerOpt effectiveness in this way, with perhaps more rigorous requirements for the plant operations during the tests. Hence, if we guarantee a 0.5% improvement in heat rate with a BoilerOpt installation, and the plant has been operating with an average heat rate of 10,000 BTU/kWh, the average heat rate value after installation of BoilerOpt (assuming all external factors remain constant) should be 9,950 BTU/kWh.
In order to assess the tool developed in the frame of this Challenge, we will compare the results from previous on-off tests done on previous projects with the improvement predictions provided by the developed tool.
This tool will be used by various GE’s engineering offices, especially sales leaders and tender engineers. Main objective of the tool is to accelerate and improve preparation of technical offers including BoilerOpt.
A simplified version of this tool might also be directly used by GE’s Customers who want to quickly estimate the improvement BoilerOpt can achieve.
Plant data for potential customers (without BoilerOpt installations) is available to some extent. We typically ask for a year of data at ten-minute resolution of all sensors which could contribute to our analysis. These typically include steam temperatures, steam pressures, attemperator sprays, excess air, air heater exit gas temperatures, ambient air temperatures, and other points. Furthermore, data from sites with existing BoilerOpt installations is also available, though it is pertinent to note that none of the existing BoilerOpt sites are choosing heat rate as the primary optimization metric. Most of the sites optimize for NOx levels, so the decisions made by BoilerOpt in the attempt to reduce NOx could potentially cause actions which degrade heat rate.
Samples of historical data will be provided later to support development of the tool.
This challenge is interesting from not only technical point of view, but also the potential of scaling the solution developed. The accelerated Start Up will develop a first pilot prediction tool for improving the controllable boiler losses and might have many other potential fields of application in the Energy Business and other businesses.
Please see here additional information on BoilerOpt Opportunity Assessment Methodology.