SIMUENGINE
4.7 M ODEL UPDATE STRATEGY AND APPLICATIONS
In additional to controlling the mill operation in a better way by providing the coal reserve and feed rate information to the power generation process control, the model accuracy can also benefit from the access of on-line real-time data. One particular situation is that the mill performance may be changed by various reasons, for example the refill of steel balls to the mill drum in a normal maintenance, the change of coal type or the biomass injection to the mill. Despite these changes of the operational condition, some other problems that are unpredictable may also affect the mill dynamics. These unexpected reasons are
supposed to be avoided and actions are supposed to take when necessary as these unexpected changes can cause potential incidents, in which the most typical case is the coal choking that blocks the coal to the furnace.
The change of the mill performance, no matter caused by normal maintenance or other unknown reasons, means that the model cannot represent the mill and the simulation result will drift away from the measured data. So the model should be updated when the simulation result shows a significant difference with the real-time measurement in the power plant. As the basic principle of the milling process remains, it is not necessary to change the mathematical equations in the model therefore the idea of the model update is actually to update its key parameters based on the measurements.
Monitoring the intermediate variables of the coal mill is supposed to be a continuous process and the unexpected problem takes place randomly. Therefore the update of model parameters must be capable to start automatically whenever necessary. For this purpose the on-line implementation program is composed of two main functions: the modelling and the parameter update. With the real-time input data collected from OPC network the modelling function estimates the mill outputs and the intermediate variables which are then broadcasted back to the OPC server. These variables can be displayed either on the program or in the Process Information system of the power plant control room. With another function in the program that accumulates the absolute error of the calculation in a defined period, which can be from minutes to hours, the periodic error in the past period can be obtained. When the result is smaller than the pre-set threshold,
which means that current model is accurate enough to represent the mill, the mill model will continue running without further actions. On the other side when the result exceeds the threshold, which means the mill performance has changed, all the input and output data from the past period will be saved and the re- identification will be carried out in parallel with the modelling based on these data. A set of new parameters can be generated after the direct search by the genetic algorithms, which is then shard to the model so the new model can represent the mill performance in time. The block diagram of the whole mill update system is shown as Figure 4.18.
Based on the on-line mill mathematical model, further applications and algorithms are developed to monitor the mill condition and to enhance the mill operation. As introduced in this chapter that the intermediate variables of the mill, including the mass flow rate of raw coal into the mill Wc, the mass of coal in mill Mc, the mass of pulverized coal in mill Mc, the mass flow rate of
pulverized coal out of mill Wpf and the mill level can be estimated by the mathematical model. By broadcasting these intermediate variables to the OPC
Coal mill Parameter identification On-line Model Display both measured and estimated mill output and the intermediate
variables
Inputs Measured outputs
Estimated outputs and intermediate variables
Periodic Error > threshold value
server the software can be very helpful to the mill condition analysis and can improve the mill control and performance to control the mill level at an optimised range.
The on-line implementation with the automatically update of the signature parameters can also greatly strengthen the condition monitoring. From the further analysis of the mathematical model and the change of the re-identified model parameters, it has been observed that some particular parameters are much more sensitive to irregular performances. For example the model parameter k1
can potentially reflect the change of fuel type fed into the mill; k10 refers to the
heat accumulation in the mill; and k15 _DE and k15 _NDE is related to the mill
choking problems. As the heat accumulation and the mill choking are abnormal events in the normal milling process that can potentially cause mill incidents, additional actions should be carried out as soon as a significant change of k10,
15 _DE
k or k15 _NDE can be observed for safety reason. However further studies on
these key parameters and the mill failure are not studied in this project as there is not enough data for the investigation.
4.8
S
UMMARYThe coal mill plays an important role in the fuel preparation process which is highly relevant to the post combustion carbon capture due to the contents of flue gases. So it is essential to understand the coal milling process. The coal mill modelling is reported in the chapter. A mathematical model based on the
working principles of the coal mill has been developed which is used to provide extra information to enhance the dynamic control. The unknown parameters of the model are identified by adopting the evolutionary computation technique with the data obtained from a real power plant in the UK.
Model validation shows that the model is able to represent the coal mill in normal grinding operation conditions. Extra information of the mill can be obtained via the model prediction, for example the coal flow rate and the mill level, which can be provided to the mill operators to enhance the mill operation. This chapter also introduced different approaches for on-line implementation of the mill model at a power plant. Self-adaptive parameters has been introduced in this on- line implementation to update the model when there is a maintenance work, or when an unexpected problem takes place. Monitoring some of the key parameters can be helpful to predict potential incidents but further investigation is needed before any conclusions can be drawn.
M
ODEL
D
EVELOPMENT OF
P
OST COMBUSTION
C
ARBON
C
APTURE
Different post combustion carbon capture technologies have been reviewed in Chapter 2. The process of the amine based chemical absorption has been studied in detail as it is currently the most established capture approach and is selected for study in this PhD project. In this chapter, further analysis of MEA based post combustion carbon capture is reported. The mathematical description of the capture system will be derived and the plant model including the carbon capture process is extended.