The project has been carried out through a detailed review of the science and technology of fast pyrolysis and related processes and a contribution to the development of a Biomass Information System. The basic scientific principles of fast pyrolysis for liquids are described and reviewed in Chapter 2 followed by a explanation in Chapter 3 of how these are applied to working processes and the main problems that have been encountered. Chapter 4 summarises the current status of fast pyrolysis and Chapter 5 describes in detail the more commercially and technically advanced of these processes. Opportunities for chemicals recovery and upgrading of vapours and liquids to more valuable chemicals and transport fuels is reviewed in Chapter 6. Chapter 7 summarises the contribution made to the BIS project and includes copies of the data suplied as an annex to that chapter. Finally Chapter 8 provides some conclusions on the current status and opportunities for each of the areas described.

CHAPTER 2

Polymerisation or deterioration of the liquid can be caused by temperatures above around 100°C which adversely affect physical properties such as viscosity, phase separation, and deposition of a bitumen-like substance. Heating the liquid to reduce viscosity for pumping or atomisation needs to be considered carefully and thoroughly tested, although in-line steam heating to 90°C has been used successfully in combustion trials (42). Exposure to air also causes deterioration, but at a slower rate than temperature increase. Maintenance in a sealed enclosure has been claimed to cause substantial pressure increases, so some minimal venting is necessary to avoid pressure build-up, but minimise exposure to oxygen. Pyrolysis liquid has been stored in this way in a useable form for up to two years without problems. Liquids produced from refuse/MSW appear to be much more unstable (119,131,132).

The low pH arises from the organic acid content (e. g. acetic and formic acids), and is therefore corrosive. Mild steel is not suitable for handling or storage. Polypropylene piping has been used to overcome this problem.

5.18.1 Summary

Both fixed and fluid bed reactors have been operated for fast pyrolysis and disposal of wastes, with catalytic upgrading of the products in the case of the fluid bed work. The objective has been to carry out screening studies on feedstocks, products and particularly examine the production of noxious products that may mitigate against implementation.

5.18.2 Description

A 200 ml stainless steel fixed bed reactor has been used to pyrolyse tyres. The reactor is externally heated and nitrogen is used as a carrier gas. Heating rates up to 80°C/min and temperatures from 300-720°C (82). More recently, work using polystyrene as the waste has been carried out with secondary thermal cracking of the product vapours at 500, 600 and 700°C (83). Experiments were also carried out with a third catalytic cracker using ZSM-5 catalyst maintained at 400°C which lowered the proportion of styrene oligomers but increased the proportion of PAH.

Work has also been carried out on fixed bed catalytic upgrading of model compounds as a prelude to the pyrolysis of biomass. Model compounds such as furfural, cyclopentanone, anisole, ethylacetate and methanol were upgraded over 10 g ZSM-5 in a fixed bed using nitrogen as a carrier gas and a range of temperatures ranging from 300°C to 500°C, WHSV 4+0.5 (84). The results showed that the EHI [Effective Hydrogen Index] was important in determining the conversion to other products. As the EHI increases for each model compound, the degree of mass conversion also increases. The importance of this is that it was estimated that the products of wood pyrolysis would have a lower degree of mass conversion compared to liquids produced from RDF or rice husks. For biomass liquids with EHI number less than 1, a catalyst at 500°C would produce a liquid with the lowest oxygen content. For biomass liquids with a EHI greater than 1, a catalyst at 400°C would give optimal conversion.

Subsequently, a dual fluid bed was used to pyrolyse biomass. Biomass was pyrolysed in an externally heated 0.075 m diameter bed, 1 m high with nitrogen as the fluidising gas. This is shown in Figure 5,16. Part of the reactor freeboard was packed with ZSM-5 catalyst. Liquids were collected before and after pyrolysis for comparison. Selected results are given below for yield of oxygenated compounds from biomass pyrolysis.

Products

For both reactors, the liquids were recovered and analysed by GC/MS. Analyses for biomass derived liquids are shown in Table 5.18. Quantification of PAH was performed by GC/FID. The recovery of styrene form the polymer was 53.0 wt% ± 1% at cracking temperatures of 500 and 600°C decreasing to 34.0 wt% at 700°C. The effects of metal salts on the pyrolysis of cellulose has been studied in a fixed bed reactor and by TGA (85).

For tyres, the maximum yield of liquid was obtained at 80°C/min and 720°C with a yield of 58.8 wt % liquids, 14.8 wt% gas and 26.4 wt% char. Detailed analysis of the liquids by SEC, FTIR have been performed.

In zeolite cracking, the conversion of oxygenates in the pyrolysis vapours occurs at lower catalyst temperatures to H2O and to CO2 and CO at higher catalyst temperatures. Coke formation on the catalyst was typically 11.4-13.1 wt%, although the run times over which this occurred at not quoted.

A variety of pyrolysis models have been derived therefore to account for the pyrolysis of particles, taking into account the process parameters noted in Table

2.3. Some of these models are summarised in Table 2.4 with their predictions and their shortcomings. A more detailed version which includes the formulation and assumptions has been complied by Bridge (85). Each model has its own particular feature or characteristic. These are detailed in Table 2.5.

Table 2.4 Summary of Single Particle Pyrolysis Models
Authors Bamford et al. (84)
Predictions • Predicts temperature and weight loss profiles • Predicts volatiles evolution rate	Shortcomings • Cannot be used to predict product yields or composition • No sensitivity analyses were carried out • No convection term • Assumed constant physical properties
• Predicts temperature and reaction rate profiles • Predicts gas generation rate • Suggests that heat of reaction and activation energy are important parameters for gas generation • Suggests that competing reactions are sensitive to the heat flux	Cannot be used to predict product yields or composition No convection term Breaks down at higher heating flux when investigating the effect of heating rate on the competing reactions
Panton and Rittmann (86)
• Shows that convection term is needed in heat balance equation • Suggests that the burning rate depends on particle size	Can only be used to investigate the importance of convective heat transfer
Kanuary et ai. (87)
Kung (88,89)
Maa and Bailie (64, 90)
• Introduces a new parameter, Lewis No., the ratio of thermal diffusivity to mass d’rffusivity • Predicts concentration and temperature profiles • Suggests that the higher the Lewis No., the greater the conversion of the solid and the smaller the temperature gradient within particle • Suggests that heat of reaction affects pyrolysis rate	Cannot be used to predict product yields or composition Does not predict volatiles evolution rate
Fanetal. (91,92,93)

• Introduced two pyrolysis numbers, • Py (ratio: reaction time to heat pene­tration time) and Р/ (Biot No x Py) •

• Evaluated the importance of external and internal heat transfer •

* Derived four simple models

* Predicts conversion and temperature profiles Predicts conversion times Particle size affects conversion Carried out sensitivity study

Predicts char yields Predicts cracking activation energies

Predicts temperature histories Predicts weight loss Calculates reaction times Derived simple expressions to calculate heat up time and devolatilisation time Pyrolysis is complete at 500°C

Predicts product yields and composition

Predicts volatiles release rate Predicts temperature profile Predicts effects of moisture Carried out sensitivity studies

Predicts product yields and composition

Predicts volatiles release rate Predicts temperature profile Predicts effects of moisture Carried out sensitivity studies

Predicts volatiles and gas yields Simple model gives good agree­ment with experimental work

Predicts high temperature drying profiles at >150°C applicable to wet particles up to the free-water continuity point (-0.45)

Simple kinetic scheme used internal flow convection effects on thermal degradation analysed on dependence of wood and char • properties

effects of grain orientation included • variation of transport phenomena, reacting medium properties and primary and secondary reactions included

Table 2.5 Features of the Models

Bamford et al. (84), Panton et al. (86), Wichman and Meleaan (113), Alves et al. (107,108).

Kanuary and Blackshear (87), Pyle and Zaror (101,102), Di В Iasi (109,111), Hastaoglu et al. (112), Kung (88,89).

Maa and Bailie (64,90)

Desrosiers and Lin (93), Saastamoinen (99),

Models including Mass Transfer Effects Fan et al. (91-93), Kansa et al. (94) Di Blasi

(109.111) , Kothari and Antal (78,79), Stiles (103), Villermaux et al. (50,51,100),

Uncoupled Heat and Kinetic Approach Philips et al. (98),

Models which predict product Yields Capart et al. (75,97), van Ginneken (95),

Wichman and Meleaan (113), Di Blasi et a!.

(109.110.111) .

Heat transfer rates are considered by some researchers to become problematical at large biomass capacities, particularly for ablative pyrolysis which is a surface area controlled process rather than a volumetric process.

Condensation of bio-oil vapours is a known problem. Cooling is not generally difficult but rather the collection of the cooled aerosol. Stable tar aerosols are formed which pass readily through the condensers and require special handling. Direct quenching in hot product oil has been used by Bio-Alternative, and also by Alten in their washing system when cooled separated product water was recycled through a packed tower. NREL and Ensyn also use direct contact with product liquid to aid collection. Electrostatic precipitators are known to work well.

Pretreated feed is claimed to give a "sticky" char product which agglomerates, sometimes in the reactor and sometimes in the downstream equipment causing build up of char-oil cakes resulting in eventual blockage. Another phenomenon is the partially depolymerised lignin which will tend to have a much higher viscosity than cellulose and hemicellulose degradation products. If this is condensed or collected in preference to other constituents due to inadequacies in the design of the liquid cooling and collection system, then blockage may again result. As high temperature char residues encourages product cracking, any char must be removed as it is formed and not allowed to be recycled or remain in contact with the vapours.

Most reactor systems (see Table 4.2) are based on orthodox fluid bed, circulating fluid bed or entrained flow reaction systems and are thus, in principle, able to be designed and operated in conventional ways. It is only the ablative type of reactors that tend to be unique to biomass.

Hydrotreating aims to remove the oxygen through a family of de-oxygenation

о

This is a carbon limited system and gives a maximum stoichiometric yield of 58% by weight on liquid bio-oil or a maximum energetic yield of about 50% wt. ignoring the significant hydrogen generation requirement which is discussed below.

A summary of the activities in catalytic hydrotreating of pyrolysis and liquefaction products is given in Table 6.5.

Table 6.5 Organisations Involved in Catalytic Upgrading of Pyrolysis and Liquefaction Oils since 1980 by Hydrotreatment and Related Processes

Battelle Pacific Northwest Laboratory (PNL), USA

64,65,66,67,68,69,70,

Colorado School of Mines, USA

Institute Nationale Recerche Scientifique (INRS), Canada Imperial College, University of London, UK Institute of Wood Chemistry, Germany Lawrence Berkeley Laboratory, CA, USA National Renewable Energy Laboratory, USA Saskatchewan Research Council Technical University of Berlin, Germany Technical University of Compiegne, France

Texas A&M University, USA 33, 86,

University of Chalmers, Sweden

University of Louvain (UCL), Belgium 95, 96, 97, 98, 99,100,101,

University of Rouen

University of Sassari, Italy

University of Waterloo, Canada

University of Saskatchewan, Canada

University of Toronto, Canada

Veba Oel, Germany

VTT (Technical Research Centre of Finland), Finland

The essential processing features for pyrolysis bio-oil hydrotreating are:

• high pressures of 70 to 200 bars to provide a high hydrogen partial pressure,

• a two stage process: an initial stabilisation reactor or reaction zone operating

at about 250-275°C followed by a more conventional hydrotreating process operating at 350-400°C. The initial stabilisation consumes little hydrogen but is essential to avoid polymerisation at the higher temperatures of hydrotreating. Without this initial stabilisation, the pyrolysis liquid rapidly polymerises and cokes the catalyst (110, 112). A temperature stepped single upflow continuous fixed bed reaction has been successfully employed for this purpose in both the USA (61, 69) and Germany (110, 112).

In practice, for pyrolysis liquids, hydrotreating yields of about 35% wt on wet liquids, based on better than 98% de-oxygenation, have been achieved from pyrolysis oils in a continuously operated carbon limited system which is about 70% of the theoretical maximum (62, 63, 116).

Conventional sulphided CoMo hydrotreating catalysts that are utilised commercially for hydro-desulphurisation have been found to be quite effective, although there is much potential for catalyst development both with conventional catalysts and with novel catalysts. Liquefaction oils do not require the initial stabilisation step and may be conventionally processed.

Hydrotreating is based on a modification or extension of well-established refinery practice for hydro-desulphurisation and involves a family of reactions including hydrogenation, de-oxygenation and cracking. The pyrolysis or liquefaction material being processed is made up of a wide range of classes of organic compounds, which all behave differently under different reaction conditions. The hydrotreating process, therefore, has to be "customised" to a particular feed and a specified product. It has been successfully demonstrated at a continuous laboratory scale on pyrolysis products (64, 65), liquefaction products (78, 84), selective extracts from liquid products (82, 107), lignin (75, 80, 81, 82) and black liquor (114, 115). Optimisation is required to establish the best catalyst system for the highly oxygenated liquids and the optimum process parameters.

Conventional sulphided cobalt/molybdenum (Co/MoS) hydrotreating catalysts have proved successful. A variety of related catalysts including Ni, Co, Mo in oxide and sulphide forms on silica and alumina supports have also given quite good results in work at Battelle Pacific Northwest Laboratory, the University of Louvain (UCL) and the Institute of Wood Chemistry (see Table 5 for full citations), although catalyst activity of the oxide forms is significantly lower. However the absence of sulphur and sulphided catalysts may be an advantage to be exploited if longer term tests show that sulphur retention is a problem with the essentially sulphur-free pyrolysis oil feedstock.

There is uncertainty over the catalyst lifetime as few extended runs have been reported. At least one continuous run of 8 days has been completed but with substantial deterioration in activity (110, 112). There is no information on the cause of this loss in activity, although physical examination of the catalyst afterwards suggests that the water in the bio-oil may have attacked the support and drastically reduced the surface area through agglomeration. Other hypotheses include alkali attack, coking, and loss of sulphide from the catalyst. There is also uncertainty over catalyst stability and activity with regard to the sulphur and possible contamination of the product oil. One of the concerns is sulphur stripping from the catalyst which

would seriously impair its effectiveness. The water product may have an adverse effect on conventional catalysts, and they may require modification to improve their water tolerance. Other contaminants may also have an adverse effect but this has not yet been explored and long term testing and detailed catalyst examination is necessary if this route is to be further developed.

Other catalytic environments have also been tested including the use of hydrogen donor solvents (89, 90, 97) and less conventional catalyst systems including oxide forms of CoMo and NiMo and modified NiMo which have so far proved to be less active than sulphided catalysts (97). Soltes employed platinum and palladium catalysts in a hydrogen donor solvent, claiming superior results to other catalysts (88).

Basic work has been carried out on model compounds (75, 101) and constituents of wood, notably lignin to devise a more effective method for utilisation of wastes from the pulp and paper industry (75, 80, 81, 101). This has included hydrotreating black liquor (113 — 115). Use of model compounds has some value in establishing pathways and kinetics as well as establishing minimum and optimum reaction parameters, but the pyrolysis liquid is so complex that this approach may be of only limited use in the short term. There will, however, be potential benefits from this approach in developing partial hydrotreating to derive a stable but only partially de — oxygenated product and in the production of chemicals rather than fuels.

Integrated conversion and upgrading has been employed in a number of configurations and processes including a first stage of hydrolysis with hydrotreating (76) and solvolysis followed by hydrotreating (79).

Analysis of products is always an important contribution to improve understanding of the processes involved so that they can be optimised and scaled up. Most of the organisations listed in Table 5 have developed appropriate analytical techniques, and some publications have focused on these analytic methods (87, 95, 98)

One of the useful side-effects of full hydrotreating, i. e. less than 2% wt oxygen in the final product, is that the product water readily separates from the hydrocarbon product and is relatively clean. The water will also tend to strip out the alkali metals present, for example, in wood ash and resolve one of the problems in using biomass derived fuels in turbines.

A brief overview of the current position on fast pyrolysis processes for liquids is given below to show the extent of activities while more details are provided in Chapter 4. A recent review of processes that are advanced technically and/or commercially available has been published (12).

3.5.1 North America

Fast pyrolysis of biomass for liquids began in North America around 1980 and has seen significant RD&D effort since then with one successful commercial organisation offering plants with a performance guarantee and several demonstration and pilot scale processes.

3-37

Ensyn’s RTP technology evolved from the research on fast pyrolysis carried out at the University of Western Ontario. They are still the only commercial organisation selling fast pyrolysis plants with a performance guarantee. There are three operational plants: a 25 dry te/d unit at Red Arrow in Wisconsin and an 80 kg/h unit and 10 kg/h R&D unit in Ottawa. Six plants are currently at a design or construction stage: 60, 25, 25, 15, 7.2 and 1 dry te/d which includes the 15 t/d unit for ENEL in Italy for delivery in 1995. Further plants for fuel production are in advanced negotiation in the USA. Ongoing development work includes hot vapour filtration to reduce ash and char, liquid filtration to reduce char, supply of oil for engine and turbine testing in Canada and Europe, combustion testing, upgrading and product characterisation. .

The Centre for Renewable Energy Sources, CRES, began a project on flash pyrolysis of biomass in spring 1990 as a consequence of the considerable European interest generated in direct production of liquid fuels from the work of Alten and Bio-Alternative. This was one of six activities started at the same time in the EEC JOULE programme (22) — the others were Egemin, INETl, Union Fenosa, University of Aston and University of Twente, all of which are described in this paper.

5.6.1 Description

The process has a nominal capacity of 10 kg dry wood per hour and is based on the principle that the by-product char from flash pyrolysis liquid production has to be used to provide process heat to drive the process. This is derived by burning the char in a conventional bubbling bed to heat sand which is then carried through a recirculating bed with the hot combustion gases to carry out flash pyrolysis in the riser section of the recirculating fluid bed (23). Instead of having two separate reactors as usually practised in steam and pyrolytic gasification processes, these are integrated into one vessel as shown in Figure 5.5. The design is in line with the recommendations of the IEA Bioenergy Agreement liquefaction group of integrating char combustion into the flash pyrolysis reactor.

A cold model has been constructed to understand the hydrodynamics of the system and a hot unit has been built, commissioned and tested. Although some liquids have been produced, no reliable heat or mass balance data is yet available.

INTRODUCTION

This chapter describes the underlying mechanisms and pathways of fast pyrolysis for liquids production before showing how these principles have been applied in working processes in subsequent chapters.

PYROLYSIS MECHANISMS AND PATHWAYS

Introduction

As biomass is heated, its various components become chemically unstable and thermally degrade or vaporise. A number of studies have shown that the main components of most biomass types, i. e. cellulose, hemicellulose and lignin, are chemically active at temperatures as low as 150°C (1). This has recently been indicated by the kinetic parameters determined by Bilbao et al. (2, 3, 4, 5, 6). Wood, is claimed to begin pyrolysis at 250°C (7). A review of the possible reaction pathways and mechanisms which the pyrolysis of wood may follow depending upon the reaction conditions are presented below. It is common to divide the reactions of lignocellulosic materials simplistically into primary and secondary pyrolysis reactions.

The component of wood which has received the most attention is cellulose. Cellulose occurs in most biomass types up to 50 wt % and has a well defined structure which allows its easy purification and separation (1, 8, 9,10, 11, 12, 13, 14, 15, 16, 17, 18, 19). This has been carried out at two different temperature ranges: up to 300°C and above 300°C. A reaction pathway for the pyrolysis of cellulose has been proposed by Shafizadeh (8, 20), Antal (9) and Kilzer et al. (10) as shown in Figure 2.

It is generally considered that primary pyrolysis of pure cellulose occurs by two competing pathways: one involving dehydration and the formation of char, CO2

and water and the second involves fragmentation and depolymerisation resulting in the formation of tarry products consisting mainly of levoglucosan as shown in Figure 2.1 (8, 9,10,11,12). At temperatures greater than 300°C, fragmentation or transglycosylation predominates which involve the conversion of cellulose into predominantly a liquid product consisting of levoglucosan and other anhydrosugars.

Two hypotheses have been proposed for the formation of levoglucosan: one by Tang (13) and Golova (14) that the glycostdic bonds are broken homolytically and that depolymerisation proceeds by a free radical mechanism. The second assumes a heterolytic transglycosylation reaction with depolymerisation proceeding by a carbonium ion intermediate (8, 9, 10, 11,15,16). The chemistry of the reactions has been reviewed in the literature (9,18).

Water, char, СО, CO2

Tar (primarily levogiucosan)

Key:

10 — Primary Pyrolysis 2° — Secondary Pyrolysis CH4, ^ ^2^4

Figure 2.1 Pure Cellulose Pyrolysis Pathways 1965-1983 (8, 9, 10,

20)

Scott etal. (17) more recently have proposed the Waterloo model for the pyrolysis of cellulose taking into account two major competing pathways for the primary decomposition of cellulose by fast pyrolysis Each pathway is capable of minor rearrangement reactions to account for the variety of different products produced, due to the dependency on the cellulose morphology, degree of polymerisation, presence of alkali cations and the process parameters such as temperature, heating rate and pressure. These potential reaction pathways are shown in Figure

2.2

(17). Atypical analysis of liquids obtained from the flash pyrolysis of cellulose is given in Table 2.1.

Fructose

■ці..- Other compounds

Figure 2.2 University of Waterloo Reaction Pathways 1988 (17)

Health hazards associated with pyrolysis liquids are also poorly understood. It has been claimed that these are no worse than coal tar or crude oil (133).