J.R. Gibbins, F.C. Lockwood, C.K. Man and J. Williamson,     Imperial College, London

G.J. Hesselmann,     Babcock Energy Ltd, Renfrew

B.M. Downer and N.M. Skorupska,     National Power PLC, Swindon

A major factor affecting a coal’s performance in air-staged low-NOx burners is the amount of nitrogen remaining in the char after devolatilisation. Current standard proximate devolatilisation tests do not apply realistic heating conditions for PF combustion, but a recently-developed high-temperature wire-mesh reactor now allows relatively simple captive-sample measurements at heating rates of 104 K/s. Char nitrogen data is reported for devolatilisation temperatures from 400°C to 1800°C, including values for a range of UK and world-traded coals for which NOx measurements from three full-scale low-NOx utility plants and/or a pilot combustor are available. The most general correlation between char nitrogen and combustion NOx is observed for a peak preparation temperature of 1800°C with 0.15 s (or 2 s) hold time or 1600°C with 2 s hold, conditions which give the maximum release of nitrogen from the char.

Keywords

coal

combustion

NOx

devolatilisation

char

test

INTRODUCTION

Pulverised coal combustion in large utility boilers can result in the production of significant amounts of oxides of nitrogen (NOx - principally NO in this type of plant). Although some of this NOx is formed from nitrogen in the combustion air, the principal source of NOx is generally the nitrogen contained in the coal itself since the latter is available in more reactive form. NOx is not the only possible end product for coal nitrogen, however, which can also form molecular nitrogen, N2. In this respect nitrogen differs from most other pollutant sources in coals (e.g. sulphur, chlorine, heavy metals) in having the potential to yield a totally non-polluting and readily-disposable residue. A simplified scheme for the formation of NOx and N2 from coal nitrogen during combustion is shown in Figure 1 (based on Pohl and Sarofim(1)).

Because of the relatively low capital and operating costs involved, combustion modifications to maximise N2 formation and minimise NOx formation are attractive, either on their own or in conjunction with further NOx reduction techniques applied after the main fuel combustion has taken place (e.g. reburn, SNCR, SCR). The main principal in minimising NOx formation during combustion is to minimise the oxygen supply to the burning coal for as long as possible. In practice this means reducing oxygen availability during coal devolatilisation and the accompanying homogeneous combustion of the volatiles. Although this may result in slightly longer reaction times these near-burner processes are intrinsically fast enough (order 0.1 s) to permit some increase within current plant geometries (i.e. with acceptable flame lengths). In contrast heterogeneous combustion of the residual char is a relatively slow process (order 1 s) which requires the presence of excess oxygen to achieve acceptable rates and final burnout. As a result, NOx formation from the char N is unlikely to be reduced significantly with air-staged low-NOx combustion systems.

The intrinsic nature of the coal is the main factor affecting nitrogen distribution between volatiles and char during devolatilisation. As natural materials coals exhibit widely different combustion behaviour due principally to:

Origin was not a consideration when industrial activities relied primarily on indigenous coals from a limited geographic area and usually a single period of origin (e.g. Carboniferous in UK and Eastern US). For coals of similar origins a wide range of properties of industrial interest correlate with rank, as illustrated by the well-known Seyler chart for UK coals. Because of this broad cross-correlation of properties many characterisation tests (e.g. carbon content, vitrinite reflectance) can be used to indicate rank and hence, by analogy with the known performance of other coals of similar rank from the same origin, the combustion behaviour.

Although most rank-indicating tests are clearly not directly related to combustion processes, the proximate volatile matter content (or the fuel ratio Johnson(2), the proximate fixed carbon divided by the volatile matter) might be considered to represent a more direct measure of a coal’s devolatilisation behaviour during combustion. This probably is the case for slow combustion of lump coal on a grate, but it is well-known that volatile yields under the rapid heating conditions applied in PF combustion are significantly higher than proximate yields. A previous wire-mesh study Gibbins et al(3) has confirmed that this is due to a combination of the absence of retrograde char-forming reactions within the bed of coal in a proximate analysis crucible when dispersed samples of finely ground coal are heated, and a greater release of volatiles from the particles themselves at higher heating rates. The enhancement is not constant, however, nor does it appear to vary in a consistent manner with rank. It therefore appears that, in as far as proximate volatile matter or fuel ratio does give some indication of PF combustion behaviour, this is also primarily due to the proximate VM being another reasonable indicator of rank within a series of coals from a similar origin.

With an increase in the use of world-traded coals and international participation in power station construction and operation projects, the commercial demand for information on the combustion behaviour of a wide range of overseas coals has increased. In this case correlations based on the results of empirical ‘ranking’ tests on indigenous coals have been found to be inadequate, particularly when comparing Northern Hemisphere coals with Southern Hemisphere (Gondwanaland) coals Carpenter and Skorupska(4). In principle new sets of rank-based correlations could be derived for each geological ‘family’ of coals. In practice this approach is difficult to implement because of the number of different correlations that would have to be derived. Even then the origin of a coal may not be easy to determine and blends of coals from significantly different sources can be encountered. A more satisfactory approach is the development of new tests which subject coal particles to conditions comparable to those that will be encountered during PF combustion, In this case the results can be used to indicate performance directly.

‘Drop tube’ entrained flow reactors are probably the most widely-used small-scale test facilities for the devolatilisation behaviour of coals under PF combustion conditions (4). A small flow of powdered coal is injected into a stream of (usually pre-heated) inert or weakly oxidising gas and passed down an externally-heated tube. Typically peak temperatures are limited to 1450°C (if relatively low cost silicon carbide heating elements are to be used) and particle residence times to 500 ms (to keep the tube length around 2m) although more realistic conditions are possible (e.g. up to 1700°C with 3 s residence time in a recent design Hutchings and Williamson(5)). The char sample is typically collected in a water-cooled probe at the bottom of the reactor tube. Volatile yields can be deduced either from the difference between the weights of coal fed and char collected or from the relative ash (or specific elemental Fletcher(6)) contents of the coal and the char, assuming the ash behaves as an inert tracer. In both cases, tars condensing with the sample may cause difficulties and a small amount of oxygen can be added to the inert sweep gas (usually nitrogen or argon) to promote tar breakdown Thompson and Stainsby(7). The char/volatile nitrogen partition can be determined from the total weight loss and the measured nitrogen content of the residual char. NOx formation in air-staged combustion of coal blends has been observed to correlate with char nitrogen yields from drop tube measurements, with an apparent conversion of about 30% of the char N to NOx Nakamura et al(8).

While drop tube reactors have yielded many insights into coal combustion behaviour, and arguably provide the most realistic conditions for small-scale experiments, they have some limitations to their use as a routine test. In the simpler designs, temperatures and residence times are near the lower margin of what might be considered appropriate for PF combustion. Total volatile yields (and hence char/volatile N partition) can also be difficult to determine accurately for coals with a low ash content or volatile ash constituents, while total char collection is always difficult to guarantee. To overcome some of these limitations, a...