Computing and optimizing total energy input to the boiler

Metso supplied the boiler for the 100 MW wood-fired Nacogdoches Power project in Texas. Photo courtesy Metso.

By Roger Leimbach, Metso Automation

It is estimated that more than 37 percent of the energy used in the U.S., excluding the generation of electricity, is for industrial applications. Further, about 50 percent of all large industrial boilers (>250,000 MMBtu/hr) use solid fuel, which include different combinations of biomass, coal and other waste products.

There are about 1,400 industrial boilers in the U.S. whose input is over 250,000 MMBtu per hour. Of these, there are over 600 grate fired boilers burning solid fuel. Optimizing solid fuel industrial boilers is critical to improving efficiency and reducing emissions.

In addition to industrial boilers, there are more than 1,200 utility-sized pulverized coal fired boilers whose ability to respond to load changes is often inhibited by age and design. They also account for the majority of all emissions for NO_X and heavy metals such as mercury. Thus it is critical that we focus on these boilers to reduce emissions.

Optimizing combustion control is critical to reducing emissions and increasing efficiency. But what defines the combustion control in a large solid fuel boiler? Any energy reduction initiative that ignores the combustion control is not going to be relevant and will most likely result in marginal improvements. Further, the use of high level advanced applications such as neural networks will be too expensive for most industrial applications.

It is important to differentiate firing applications such as fluidized bed combustion, grate firing and pulverized coal. It is also important to develop control strategies for each that are simple to implement and update. These strategies should be easy to understand and not require a high degree of expertise to maintain. The strategies should be compatible with the overall boiler control system, not be platform-oriented or solely dependent upon mass fuel flow. Instead, they should be based on actual energy flow.

Further, the boiler control strategies should be capable of changing load at maximum rate without causing upsets and instability in the process. They should be able to run in Automatic Generation Control or be base loaded. AGC is important to all utilities and it must be inherently built in to the boiler control and turbine control. As one utility engineer recently stated, "We won't build it if we can't dispatch it."

Boiler Turbine Coordination in an Industrial Environment

In many industrial installations, there are multiple boilers supplying steam to a process and/or steam turbines driving electric generators. The turbines can be connected to a steam network and operate in backpressure mode or condensing mode. There are as many different combinations as there are plants. However, what is typically missing is the coordination of the turbines and steam hosts with the boilers. This must be addressed.

It goes without saying that the boilers should be controlling the header pressure to which they are supplying steam. In too many cases the turbines and pressure reducing stations are operated in pressure mode, sometimes called turbine follow mode. In this case the boilers are base loaded and the header pressure is maintained by the turbines governor valves.

The boiler demand should be based upon header pressure and a feedforward which is divided among the various boilers according to their relative size and efficiency. The control system should allow for operator bias of individual boiler demands.

The feedforward should be the primary control element for the boiler. The pressure control should provide minimal integral action. Integral control should be used sparingly in the boiler demand. The feedforward should not be steam flow since this is a regenerative feedforward. Regenerative feedforwards tend to drive the demand in the wrong direction when an upset occurs or if the fuel quality should change. It will cause the pressure control to oppose the change and cause further upsets and destabilization.

Figure 1 is a depiction of an advanced boiler control in an industrial configuration with the energy flow computation and boiler optimizer. Further definition of the energy flow calculation and optimizer are explained later. Note that each boiler on the header has a "participation algorithm" which allows for apportioning the demand according to its individual size, efficiency and response. Also, the operator has the ability to bias each boiler accordingly. The various airflow signals and fuel feeders have a participation algorithm to perform the same function.

Fluidized Bed Combustion (FBC) Boilers

FBC boilers are used for a wide variety of applications, mostly in industrial uses which typically have several boilers working together on sophisticated steam networks where demand can change rapidly. The boilers acting in parallel can also be uncontrolled. That is, their output is completely dependent upon the process to provide waste fuel which may not be measurable. Also, FBC boilers are likely to use relatively hard-to-burn fuels such as combinations of biomass and waste coal, all of which can and will change.

Bio-fuel is not necessarily a homogeneous fuel. It is often a mixture of different fuels such as bark, forest cuttings, agriwaste and waste building materials. Also, a wide range of fuel types can be processed to keep the overall cost at minimum.

FBC boilers set a different standard for controls. The fluidized bed of sand and ash within the furnace has a very large inertia. This will limit the dynamics of the boiler. On the other hand this enables the FBC boiler to burn high moisture content fuels – as much as 60 percent. While the fuel will change greatly the fuel will burn relatively fast, there is still quite a range in the combustibility of the fuels. The evaporation, pyrolysis and combustion times will vary greatly.

While multi-fuel combustion will vary greatly, the fuel feed control must play a critical role. Energy flow must be maintained at a consistent rate to make sure pressure is maintained at setpoint and energy flow out of the boiler is at demand. Changes in energy flow to the boiler can cause fluctuations in combustion and steam production. If not precisely controlled, these fluctuations can risk plant availability.

The design and nature of an FBC boiler is such that its turndown is limited and response is relatively slow in comparison to gas fired boilers or pulverized coal fired boilers. Further, the number of FBC boilers used in electric production only applications is growing, not only in numbers, but in size. Some that will fire 100 percent biomass and produce up to 100 MW are now under construction in the U.S.

FBC boilers are not susceptible to high levels of emissions such as SO_X and NO_X. However, those that fire coal require limestone as an additive to capture SO_X . Limestone is used in large amounts for this purpose and is considered an expense to overall operations that can amount to millions of dollars depending upon the size of the boiler. Because FBC boilers operate at lower temperatures, NO_X is not considered a problem, but if the bed temperature rises above the NO_X threshold, more ammonia is required as a mediating additive. These emissions are all bed temperature sensitive and require extreme optimization to maintain low levels of emissions. Thus, the fuel control, without any compensation for changes in moisture, composition or heating value is problematic.

The answer to these issues is a computation referred to as the Fuel Power Compensator (FPC). The FPC was developed for the world's largest biomass -fired boiler, Alholmens, a 240 MW unit operating in Finland since 2002. The unit can burn any combination of biomass and coal. Biomass is a combination of peat, bark and wood residue. Coal is the backup fuel.

The continuous use of oxygen consumption and energy balance calculations are the backbone of the FPC. In the case of oxygen consumption we mean the oxygen consumed by the actual fuel being burned. This can be estimated by the excess air ratio and combustion air flow. Oxygen consumption translates well with energy flow rates. The unique Fuel Power Compensator combines the features inherent in the energy balance estimator and oxygen consumption calculation.

The reason for the combination of the two computations is that the energy balance calculation is based upon averages and is relatively slow – but very accurate. The oxygen consumption calculation is relatively fast – milliseconds – but it has errors. To be precise the oxygen consumption calculation has errors due to the transport delay from combustion to the oxygen measurement. So combining the two provides an accurate view of what is happening in furnace in real time.

The Fuel Power Compensator has been used on many multi-fuel boilers with excellent results. The results at Alholmens are shown in Figure 2. In this case the operator is manually adding coal in 10 percent to 15 percent steps while the fuel control is responding to the FPC. Note how the steam flow and pressure are responding.

FBC Boiler Optimizer

There are many different types of optimization programs available that are based upon model predictive control, fuzzy logic and neural networks. One should be selected that meets the objectives of the plant and can be maintained by the human resources already at the plant. In addition it should be able to use standard already existing process instrumentation such as oxygen analyzers, flow and pressure transmitters and on-line emission analyzers.

It should also be able to utilize operator know-how and experience. It should be able to be added to any existing control system and provide biases to setpoints in the control. It should be able to run in a PC or the control system controllers.

First, let's look at the basic controls for a typical FBC boiler. Figure 3 shows the basic controls. The biggest difference is the bed temperature and fuel feed controls.

The load is controlled by the steam demand in the form of a non-regenerative feedforward based upon steam flow, steam header pressure and steam header pressure setpoint. These are incorporated in the boiler demand calculation which includes dynamic compensation (1). This is needed to allow AGC. While the boiler integrates the difference between the energy input and energy output of the boiler, the pressure error is primarily a proportional only process. The long boiler time constant attributable to all boilers, and especially FBC boilers, does not permit the pressure control to operate as an integral control. The change in boiler demand is due primarily to the change in the feedforward. The other changes to boiler demand are from fuel quality changes.

The fuel demand goes to the fuel feeder control (11) and to the air flow control (3) where excess air is used to trim the air flow demand to the secondary air control (6) and the primary control (7). The ratio of secondary to primary air is set by the operator or the optimizer control (5). The ID fan controls furnace pressure according to a feedforward from the air flow demand (8). The bed temperature control is executed by recirculating flue gas into the bed. The flue gas tends to slow combustion and reduce the bed temperature.

The most unique control loop associated with a FBC boiler is the bed temperature. The bed is composed of many tons of hot sand and ash which is fluidized by primary air. The fluidization process is very important to the control of emissions and is important to minimizing limestone consumption in a coal fired FBC. Further, bed temperature is a function of fuel quality and changing boiler load because it is a function of the thermal balance of the bed. This is perfect for the application of a fuzzy controller.

Fuzzy controls can be broken down into several decomposed control loops with several "multiple input single output" (MISO) problems. Each is similar in nature to a traditional PID control loop where:

∆u(K)=K_p∆e(K)+K₁e(K),…

K = time step

∆u=increment of the control step

e = control error

∆e=difference in the control error

In a fuzzy controller we utilize a similar equation. The inputs e are "fuzzyfied" with three triangular membership functions; negative (low), zero (normal) and positive (high).

If we use a "defuzzification" method with a weighted sum we get a control algorithm as such:

Where ∆u(K) is the output of rule i of the four basic rules 1 or 2 at step time K. The ∆Uj(k) is the output of rules j 3 or 4. K_p is proportional gain and k_i is the gain of the integration part.

An example of this control is an industrial FBC boiler in a pulp mill that burns bark and peat whose load changes greatly over a small period of time. The inhomogeneous nature of the fuel and load changing add to the problem. The moisture (up to 60 percent) and the changing fuel quality will upset the boiler greatly since the energy balance of the bed is critical to maintaining bed temperature. For instance lower bed temperature can be judged to mean less fuel quality or more water in the fuel.

The bed is a form of energy storage system. The fuzzy logic controller will change the ratio of primary air to re-circulated flue gas to change bed temperature.

The first five rules shown in Figure 4 are duplicated according to boiler load – high and low. Also, tuning the controller is different for high and low loads.

The objective of advanced bed temperature control is stable bed temperature.

Grate Fired and Pulverized Coal Boilers

The typical fuel control for a grate-fired boiler is approximate, and in some cases pressure is the only indicator used to control fuel. The computations used can lag by several minutes. They do not indicate changes in Btu content in real time, or deal with other changes in fuel quality such as moisture and density. A real time indicator of energy input is needed. In addition, the air flow control must be split between primary air (undergrate), secondary air and overfire air.

The need to reduce emissions has led to a reduction in primary (undergrate) air from as much as 85 percent to as little as 60 percent. Overfire air has been increased to reduce NO_X. Backend controls such as scrubbers and SNCRs have been added to the units which require booster fans, all of which reduce efficiency. Increasing efficiency, while at the same time reducing emissions is more important than ever. An improved method to control fuel is necessary.

Metso has developed a strategy to control fuel based upon energy flow to the boiler. It is referred to as Heat Release. The principle behind it is based upon the assumption that energy into the boiler is equal to energy out of the boiler. This presumes to control the output you must be able to control the input, and to control input, you must be able to precisely measure the input. Measuring the energy input to a solid fuel fired boiler is difficult enough, when the fuel is homogeneous in nature, but when it is multi-grade biofuel it can be next to impossible. Fortunately a boiler is a self regulating process.

It integrates the difference between energy in and energy out. Stored energy in the water, steam and metal makes it possible to change load.

During load changes energy in does not equal energy out. It is common knowledge that on a load increase we must overfire the boiler. Why? There is a certain amount of stored energy that must be established for each new load point. Stored energy is non-linear with load. We must be able to determine the exact amount of overfiring required and we must be able to measure the change in energy input for this is the amount of energy that we must be able to support at the new load point.

Stored energy is like a giant flywheel that makes it possible to change load and match steam output with demand. Stored energy is created by system resistance, the fuel delivery system and heat transfer characteristics. A boiler is a giant integrator. It integrates the difference between energy in and energy out! Drum pressure is an indicator of stored energy and the derivative of stored energy is an indicator of the change in stored energy. Therefore we can state that a measure of fuel input is the energy output ±dPD/dt where PD is drum pressure.

Where P₁ is a measure of energy flow from the boiler. Note that if this is a central station boiler with a single steam turbine first stage pressure is used instead of steam flow.

Heat Release is a real time indicator of true energy input to the boiler. It detects changes in fuel heating value. It computes total heat release from all sources of fuel.

Heat Release was developed for pulverized coal fired boilers and can be used on grate fired boilers. Heat Release can be used as the sole fuel feedback.

In addition we must employ a non-regenerative feedforward as the energy demand. Figure 5 shows the energy demand on a load change. Note the difference between boiler demand and unit demand. This difference is the precise amount of overfiring needed to support the new level of stored energy.

This is the basis for the D-E-B system which employs Heat Release as the energy feedback. D-E-B has been used on over 1,000 boilers up to 1,000MW in size.