Applied Control Systems, Inc


Control Implementation & Design
FOR
AUTOMATIC GENERATION CONTROL
ON A
FLUIDIZED BED BOILER

by
Frank Ryan - Senior Application Consultant, A.C.S.
For
Herb Gery – Senior Application Consultant
Deceased August 1998

 

APPLIED CONTROL SYSTEMS
426 South Main Street
Pittsburgh, Pennsylvania 15220

ABSTRACT

This paper will discuss some of the features associated with the control approach used to manipulate the the circulating fluidized bed boiler (CFB) and turbine governing valves to facilitate the implemention of Automatic Generation Control (AGC) at the Scrubgrass Generating Plant located in Scrubgrass Township, Venago County, PA.

SECTION I

INTRODUCTION

The challenge was to deal with the very long time delays associated with the dynamics of the fluidized bed boiler and its resistance to change. The time constant of the system is in the order of five minutes or longer. The AGC requirements from Pennsylvania Electric Company, the purchasing utility, mandated that load changes of plus or minus 2 MW be implemented in no less than five minutes. While those of you who live in the pulverized coal world may feel that that is not much of a challenge, it was difficult with the fluidized bed. Successful implementation carried with it economic rewards from the utility so the pressure to perform was significant.

CFBs commonly burn trash coal also known as ‘gob’. This has a widely variable BTU and high ash content. The ‘gob’ burned at Scrubgrass varied in BTU content from as low as 5500 BTU/LB to as high as 8000 BTU/LB and was normally delivered very wet. An interesting aspect of this fuel, which comes from a variety of sources, is that its response to change varies significantly. Currently, this can be as little as 30 seconds or as long as 3 minutes for a change in feed rate to be felt in an initial change in header pressure. Suffice to say that this data is quite casual and no meaningful effort has been made to investigate this curious phenomenon. Speculation abounds.

SECTION II

PLANT DETAILS

BOILERS             Two --405,000 LB/hr Circulating Fluidized Bed boilers manufactured by Tampella Power with the following control systems on a per boiler basis:

  • Four graveometric coal feeders
  • One Induced Draft Fan
  • On Primary Air Fan
  • Flue Gas Oxygen Control
  • Steam Header Steam Temperature control System
  • Single/Three element drum level control
  • One Fluoseal Blower
TURBINE             ONE--95 Mw Steam turbine manufactured by GEC Alsthom.
CONTROL SYSTEM             Max 1 Distributed Control System manufactured by Leeds & Northrup with the following features:

  • 4 CRT Screens
  • 4 Operator Interface Keyboards
  • 2 Operator/Engineer Consoles
  • 495 Thermocouple Inputs
  • 70 RTD Input
  • 630 Current Type Analog Inputs
  • 1856 Digital Inputs
  • 304 Analog Outputs to control drives
  • 4 Pulse Type Digital Outputs to Governor Control
  • 1232 Digital outputs
  • 14 Pairs of redundantly configured control processors
  • 6 System Cabinets Physically Distributed
Emissions Control         Limestone injection, thermal deNox, fabric filter baghouse.

Commercial Operation         June 1993
Owner         Scrubgrass Generating Co., L.P.

SECTION III

CONTROL OVERVIEW

TURBINE CONTROL

When on AGC, the turbine is under positive megawatt control. This is done with a cascade loop as shown in Figure 1. The outer loop compares an operator set megawatt demand to the actual value of gross megawatts. The outer loop controller then develops a demand for a valve position that will produce the required megawatts. When Penelec calls for a change, and all changes are limited to a maximum of two megawatts, the AGC interface will insert a corresponding bias into the set point of this controller.

Valve position is measured by calculating pressure ratio, which is the ratio of turbine first stage pressure to throttle pressure. This is a proven technique for obtaining this value and has some significant advantages over the mechanical position signals that were available. It is also used as a feedforward in the boiler control system as will be described later on. In addition to being more reliable than a mechanical measurement of valve position, pressure ratio has the advantage of being a linear representation of the effective valve opening. This is much simpler than attempting to calibrate the four position signals from the four valves (that proved to be unreliable) that were operated in pairs sequentially with overlap at 75%.

The inner loop controller simply adjusts the turbine valves to get the required pressure ratio. Of special interest is the interface between this control and that of the turbine. This is in the form of another special algorithm called a DIAT. This stands for Duration Impulse Adjustment Type. The algorithm turns on either an increase or decrease digital output, the duration of which is a function of the control error. It integrates the on times to establish its stabilizing feedback. As a result, no analog signals are required from the governor control in order to effect the interface.

There is also a turbine following mode available, which is also shown in Figure #1. This is used during emergency conditions and is very effective in attaining a level of generation that can be supported by a boiler that is limited for some reason.

BOILER CONTROL

As shown in Figure #2 the basic boiler control is a conventional boiler following system. Steam header pressure establishes the total demand for boiler inputs. Each boiler is placed on automatic individually and a means of biasing loads is provided.

Of particular interest is the Energy Demand algorithm that is used to aid the header pressure controller in its regulation of the boiler inputs. The alogorithm calculates pressure ratio, the same calculation that is used in the control of the turbine as described previously. Because this quantity is a linear representation of the effective turbine valve opening, it is an excellent measure of the need for boiler inputs. As a feedforward signal, pressure ratio has the distinct advantage of reflecting demand changes only, and is not influenced by upsets within the boiler as is the case with steam flow when that measurement is used as a feedforward. For instance, a reduction in steam flow can mean either a load decrease, in which case a reduction in boiler inputs is required or it can mean a reduction in fuel quality or other upset, in which case an increase in inputs is required. This ability of pressure ratio to distinguish between load change and boiler upset stems from the fact that a change in valve position will effect the numerator and denominator of the ratio in opposite directions causing a change in the calculation. Boiler upsets, however, cause the numerator and the denominator to change in the same direction and by the same percentages leaving the ratio unchanged.

Pressure ratio is calibrated by multiplying it by the pressure set point. This is necessary because a given value position will require a supporting value of input in direct proportion to the desired value of header pressure.

A major feature of the Energy Demand algorithm is the fact that it calculates a rate action for the purpose of over or under firing the boiler during load change. Before this is done it is desirable to have filtering available, particularly if the signal has a process noise component as is sometimes the case with turbine first stage pressure when the turbine valves are in a sensitive range. This filtering action permits the alogorithm to either ignore or provide a minimal response to signal changes within a preselected band. This means that the rate computation will be made only on meaningful signal changes.

The next computation multiplies the rate of change of the calibrated and filtered signal by the signal value at which the rate was calculated. It then adds this to its current value. The net effect is to provide a correction that increases exponentially as load increases. This is done to over and underfire the boiler during load changes as the boiler's stored energy changes in an exponential fashion. This matching of the rate action, or dynamic compensation as it is sometimes called, to the needs of the boiler enables the maximum use of this function since the need for a compromise tuning is much diminished.

SECTION IV

RESULTS

Figure #3 shows the results of load change tests run to determine compliance. This data was taken from the video trending display of the Distributed Control System. It shows a series of four-two megawatt changes imposed on the system as step changes between times 15:11 and 15:54. The upper and lower lines are the overall megawatt limits. Generation demand is shown as a series of two-megawatt instantaneous ramps. The megawatt response is reasonably crisp with no signs of instability after the change. This is important because load changes are like wars. Easy to start but hard to stop. A period of approximately ten minutes was set aside between each change to look for any signs of instability, which did not appear. The plant continues to operate on AGC and has done so for four years without problem.

Of interest is the fact that there is no attempt to lead or lag the airflow during load changes. This would seem to be advantageous because it would burn off inventory during a load increase to get initial response thereby buying time for the increased coal input to have its effect and the opposite effect would assist load reductions. This was tried but found to be ineffective. The most significant aid to the system in implementing load changes was the use of Energy Demand as a feedforward to the header pressure controller. The closed loop megawatt control was deemed as being helpful in avoiding the effect of pressure recovery on megawatt output.