Article pubs.acs.org/est
Solid Recovered Fuel: Materials Flow Analysis and Fuel Property Development during the Mechanical Processing of Biodried Waste Costas A. Velis,†,‡ Stuart Wagland,† Phil Longhurst,† Bryce Robson,§ Keith Sinfield,§ Stephen Wise,§ and Simon Pollard*,† †
Cranfield University, Centre for Energy and Resource Technology, Department of Environmental Science and Technology, School of Applied Sciences, Cranfield, Bedfordshire MK43 0AL, United Kingdom ‡ University of Leeds, School of Civil Engineering, Leeds LS2 9JT, United Kingdom § Shanks Waste Management Ltd., Dunedin House, Mount Farm, Milton Keynes, Buckinghamshire MK1 1BU, United Kingdom S Supporting Information *
ABSTRACT: Material flows and their contributions to fuel properties are balanced for the mechanical section of a mechanical−biological treatment (MBT) plant producing solid recovered fuel (SRF) for the UK market. Insights for this and similar plants were secured through a program of sampling, manual sorting, statistics, analytical property determination, and material flow analysis (MFA) with error propagation and data reconciliation. Approximately three-quarters of the net calorific value (Qnet,p,ar) present in the combustible fraction of the biodried flow is incorporated into the SRF (73.2 ± 8.6%), with the important contributors being plastic film (30.7 MJ kgar−1), other packaging plastic (26.1 MJ kgar−1), and paper/card (13.0 MJ kgar−1). Nearly 80% w/w of the chlorine load in the biodried flow is incorporated into SRF (78.9 ± 26.2%), determined by the operation of the trommel and air classifier. Through the use of a novel mass balancing procedure, SRF quality is understood, thus improving on the understanding of quality assurance in SRF. Quantification of flows, transfer coefficients, and fuel properties allows recommendations to be made for process optimization and the production of a reliable and therefore marketable SRF product. flowsheet as a whole, is poorly characterized. Uncertainty about operations and the capacity to produce marketable SRF adversely impacts the financing and regulatory approval of new plants.16 The research interest here is in applying process science to reveal the links among materials flow, plant configuration, and fuel quality. Through the use of a novel mass balancing procedure, SRF quality is understood, thus improving on the understanding of quality assurance in SRF. Here, a validated analysis of material and fuel property flows through a full-scale SRF-producing MBT plant in the UK is provided. The performance of unit operations in the mechanical processing section is evaluated considering the following:1,9 (i) the yield of SRF output; (ii) the development of net calorific value (NCV) through the plant; and (iii) the dilution/concentration of fuel properties that raise processing and/or environmental concerns, specifically chlorine and ash content.17,18 Material flow analysis (MFA)19 was employed to map waste flows in a novel combination of (i) waste stream sampling informed by the theory of sampling (ToS);20 (ii) manual sorting for wastes
1. INTRODUCTION Solid recovered fuel (SRF) is becoming a significant contributor to the international resource and energy efficiency agenda. By utilizing SRF, renewable energy is recovered from the biogenic fraction of nonhazardous waste, replacing fossil fuels in energy-intensive industries.1 Around 15 Mt of SRF was available in Europe in 2004,2 half of it generated by mechanical−biological treatment (MBT) plants processing municipal solid waste (MSW).3 Market security for the SRF product depends on producing a fuel of suitable and consistent quality.1,4−7 A previous generation of refuse-derived fuel (RDF) production plants in Europe8 struggled due to insufficient or inconsistent quality6 of the fuel output.9 The United States has a long history of using RDF,10,11 where similar challenges led to a gradual decline. China12 and Korea13 are, for example, among the fast developing economies considering widespread cocombustion of SRF.14 Operators manufacturing SRF to a specified quality need to understand the materials flow of waste components (paper, plastics, wood) through their facilities to optimize plant configurations and produce SRF with predictable properties (calorific value, ash content, and chlorine content15,16) for end users. To date, the design and operation of MBT plants remains semiempirical, because the performance of unit operations, such as trommels and air classifiers, and of the mechanical © 2013 American Chemical Society
Received: Revised: Accepted: Published: 2957
July 6, 2012 October 10, 2012 February 11, 2013 February 11, 2013 dx.doi.org/10.1021/es3021815 | Environ. Sci. Technol. 2013, 47, 2957−2965
Environmental Science & Technology
Article
Figure 1. Sankey diagram for the mechanical processing section of the UK MBT plant, showing flows for the sum of combustible waste components. The width of the flows illustrates their relative magnitude. The ratio of any output flow to the sum of input flows provides the transfer coefficient (TC) for this output. TC values (Table 1) are estimated as average specific (per component) mass load (ar) for ca. 10 000 overall components mass input (combustible and noncombustible), at 95% confidence.
counter-balanced. Transfer coefficients (TCs)1,9 were estimated using the materials flow analysis (MFA) software STAN2.22 Mass Balancing, Flows Reconciliation, and Transfer Coefficients. Plant operational data were used to overcome (i) an absence of automated flow measurements, typical for solid waste processing plants; (ii) a practical need to minimize downtime; and (iii) the fact that certain flows were not physically accessible. Operational data included the mass input to the mechanical processing section (SP1) from the crane feeding system, system outputs, and batches of various outputs as they left the plant (weighbridge data) for the sampling period (Feb−Sept 2010). These data complemented on-site flow measurements (mass over time) for non-Fe metals (SP14) and eddy current rejects (E_C rejects, SP15), along with literature data employed for the non-Fe metals (mass ratio of non-Fe/SRF from a UK MBT plant of similar technology23). An average reconciled mass flow map was constructed in two stages. Because input and output flows alone do not suffice for computing a full set of flows, a simplified model was created ignoring the smallest of the inner flows (A_CL 2 lights, F17; and flows leading to Fe mixing, T/F). With this simplification, the split between the mass leaving the trommel and air classifier (A_CL 1) was established. Next, the gross estimate for A_CL 1 lights, along with its uncertainty, was used to produce estimates for each flow in the mechanical section. Insignificant were (a) the air pollution control (APC) residue output (0.04% w/war of processing section input); (b) the recirculation of the E_C
suited to mechanical processing; (iii) wastes characterization; (iv) modeling, sensitivity analysis, and flow balancing; and (v) MFA with data reconciliation and error propagation. In concert, these elements provide a comprehensive evidence base for the production of SRF of reliable quality; an outcome that has proven elusive in UK MBT plants.
2. MATERIALS AND METHODS A glossary of all terms is provided in the Supporting Information (SI), Table SI1. Flowsheet, Sampling, and Materials Flow Analysis. The MBT plant studied receives residual MSW remaining after curbside recycling and some commercial and industrial waste. After initial bag-opening and light shredding (to 150−300 mm), material is bioconverted3 in windrows of ca. 5 m height for 2 weeks to a moisture content of 13.4 ± 5.5% w/wd. The biodried output is transported to the mechanical processing section of the plant to produce SRF and other materials. The flowsheet, including the role and the abbreviation of each unit operation, is detailed in the Supporting Information (SI; Figure SI1, Table SI2). Sixteen individual flows (F) were sampled (sampling point, SP) (Figure SI1) using the ToS.6 MFA9,19,21 was used to map and balance the flows and track fuel properties as they developed. Flows were depicted in Sankey-type diagrams that visualize the direction and magnitude of flows through the plant22 (Figure 1). Inner flows, for which no data were initially available, were estimated and all flows were 2958
dx.doi.org/10.1021/es3021815 | Environ. Sci. Technol. 2013, 47, 2957−2965
Environmental Science & Technology
Article
Figure 2. Shankey diagram for the combustible mass fraction expressed on a dry and ash-free basis (daf) (i.e., moisture and ash content are not counted). This diagram shows the flow of actual mass from which it is useful to recover energy.
the average plant performance at the sampling period (see SI Table SI3 and Figure SI2). Sums of related waste components were balanced along with those for single waste categories (denoted below in italics, with their associated symbols): items suitable for combustion, Σ(com); items not suitable for combustion, Σ(noncom); paper/card and similar, Σ(P/C+TIS+CAR); sum of plastics (resins), Σ(COM +D_P+O_P_P+P_F); and textile/fabric and similar: Σ(C/M +FL+S_P+SH+T/F). Next, fuel-related properties were determined for process stream samples using SRF standard methods for moisture content26 (% w/wd), ash content27 (% w/wd), total chlorine content28 (% w/wd), and net calorific value29 (MJ kg−1). Reconciliation for ash, chlorine, and net calorific value requires correct mass balances on a dried (w/wd) or as received (w/war) basis, whichever applies, to validate the total mass flows from subsampling and analysis. Reconciliation of the combustible dry ash-free matter (daf) validates the data sets for (i) the total matter “as received”; (ii) the moisture content determinations at each sampling point; and (iii) the ash content mass balance.
rejects to the input of the processing section; and (c) any nonsolid flows. Waste Component Characterization, Flow Balancing, and Fuel Properties. At each sampling point, flows were manually sorted into 25 waste component categories, including a fines fraction passing a 10-mm circular aperture sieve. Component mass percentages (the purity of that waste component in the flow1) were estimated using statistical analysis assuming normal distributions. Extended uncertainty for each property R (U95,ν () was calculated24 at the 95% confidence level, for ν effective degrees of freedom (d.f.). Where applicable, uncertainty was propagated throughout the computations using Taylor’s series approximation25 and used as input to STAN2 error propagation. Assuming normality for waste present in small amounts (
