Tactical sugarcane harvest scheduling

Date
2010-12
Authors
Stray, Bjorn Jonas
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : University of Stellenbosch
Abstract
ENGLISH ABSTRACT: Computerised sugarcane harvest scheduling decision support is an active fi eld of research which ties in closely with the broader problem of automating and streamlining the various activities in the sugar supply chain. In this dissertation, the problem of providing decision support with respect to sugarcane harvesting decisions is defined within a number of contexts, each representing a typical kind of organisation of sugarcane farmers into a cohesive decision making unit with its speci fic requirements and limitations that exist in practice. A number of variations relevant to these contexts of an overarching tactical sugarcane harvest scheduling problem (THSP) are considered and solved in this dissertation. The THSP is the problem of providing objective, responsible decision support to persons charged with the task of determining optimal harvesting dates for a set of sugarcane fields across an entire season. Sugarcane fields typically diff er in terms of the age, variety, life-cycle stage and in many other properties of the cane grown on them. The growth of sugarcane crops may also be a ffected by environmental conditions such as accidental fires, frosts or storms which have a detrimental e ffect on crop-value. Since sugarcane is a living organism, its properties change over time, an so does the potential pro t associated with it. The practicalities of farming cause further complication of the problem (for example, seasonal changes alter the conditions under which the crop is harvested and transported). The rainy season carries with it the added cost of disallowing long-range vehicles to drive into the fields, forcing the unloading and reloading of cane at so-called loading zones. Other considerations, such as the early ploughing out of fields to allow them to fallow before being replanted, compounds the THSP into a multi-faceted difficult problem requiring efficient data management, mathematical modelling expertise and efficient computational work. In the literature the THSP has been viewed from many different standpoints and within many contexts, and a variety of operations research methodologies have been employed in solving the problem in part. There is, however, no description in the literature of a solution to the THSP that takes the negative e ffects of extreme environmental conditions on the quality of a harvesting schedule into account in a scienti fically justifi able manner; most models in the literature are based on optimising sucrose yield alone under normal conditions, rendering weak schedules in practice. The scope of the modelling and solution methodologies employed in this dissertation towards solving the THSP is restricted to integer programming formulations and approximate solution methods. The parameters associated with these models were determined empirically using historical data, as well as previous work on deterioration of sugarcane following environmental and other events. The THSP is solved in this dissertation by designing a generic architecture for a conceptual decision support system (DSS) for the THSP in the various contexts referred to above, which is capable of accommodating the e ects of extra-ordinary environmental conditions, as well as the introduction of a computer-implemented version of a real DSS for the THSP conforming to the framework of this generic architecture. The DSS building blocks include prediction models for sugarcane yield, sugarcane recoverable value under normal circumstances, the costs associated with a harvesting schedule and the negative e ects on sugarcane recoverable value of extraordinary environmental conditions. The working of the DSS is based on a combinatorial optimisation model resembling the well-known asymmetric traveling salesman problem with time-dependent costs which is solved approximately by means of an attribute-based tabu search in which both local and global moves have been incorporated. The DSS is also validated by experienced sugarcane industry experts in terms of the practicality and quality of the schedules that it produces.
AFRIKAANSE OPSOMMING: Gerekenariseerde besluitsteun vir die skedulering van suikerriet-oeste is 'n aktiewe navorsingsveld wat nou verwant is aan die bre ër probleem van die outomatisering en vaartbelyning van 'n verskeidenheid aktiwiteite in die suikervoorsieningsketting. Die probleem van die daarstelling van steun rakende suikkerriet oestingsbesluite word in hierdie proefskrif in 'n aantal kontekste oorweeg, elk met betrekking tot 'n tipiese soort organisasie van suikerrietboere in 'n samehorige besluitnemingseenheid met sy spesi eke vereistes en beperkings in die praktyk. Verskeie variasies van 'n oorkoepelende taktiese suikerriet-oesskeduleringsprobleem (TSOSP) wat in hierde kontekste relevant is, naamlik die probleem om objektiewe, verantwoordbare steun aan besluitnemers te bied wat verantwoordelik is vir die bepaling van optimale oesdatums vir 'n versameling suikerrietplantasies oor die bestek van 'n hele seisoen, word in hierdie proefskrif bestudeer en opgelos. Suikerrietplantasies verskil tipies in terme van ouderdom, gewastipe, posisie in die lewensiklus, en vele ander eienskappe van die suikerriet wat daar groei. Omgewingstoestande, soos onbeplande brande, ryp of storms, het verder ook 'n negatiewe impak op die waarde van suikerriet op sulke plantasies. Omdat suikerriet 'n lewende organisme is, verander die eienskappe daarvan oor tyd, en so ook die potensi ele wins wat daarmee geassosieer word. Boerderypraktyke bemoeilik verder die skeduleringsprobleem onder beskouing (seisoenale veranderings beïnvloed byvoorbeeld die wyse waarop suikerriet ge-oes en vervoer word). Addisionele koste gaan voorts met die re ënseisoen gepaard, omdat die plantasies dan nie toeganklik is vir langafstand transportvoertuie nie en suikerriet gevolglik na spesiale laaisones gekarwei moet word voordat dit op hierdie voertuie gelaai kan word. Ander oorwegings, soos die vroe ë uitploeg van plantasies sodat die grond kan rus voordat nuwe suikerriet aangeplant word, veroorsaak dat die TSOSP 'n moeilike multi-faset probleem is, wat goeie databestuur, wiskundige modelleringsvernuf en doeltreff ende rekenaarwerk vereis. Die TSOSP word in die literatuur vanuit verskillende standpunte en in verskeie kontekste oorweeg, en 'n aantal uiteenlopende operasionele navorsingsmetodologie ë is al ingespan om hierdie probleem ten dele op te los. Daar is egter geen poging in die literatuur om 'n oplossing vir die TSOSP daar te stel waarin daar op 'n wetenskaplik-verantwoordbare wyse voorsiening gemaak word vir die negatiewe e ffekte wat uitsonderlike omgewingstoestande op die kwaliteit van oesskedules het nie; die meeste modelle in die literatuure is op slegs sukrose-opbrengs onder normale omstandighede gebaseer, wat lei na swak skedules in die praktyk. Die bestek van die wiskundige modellerings- en gepaardgaande oplossings-metodologie ë word in hierdie proefskrif vir die TSOSP beperk tot onderskeidelik heeltallige programmeringsformulerings en die bepaling van benaderde oplossings deur lokale soekprosedures. Die parameters wat met hierdie modelle en soekmetodes geassosieer word, word empiries bepaal deur gebruikmaking van historiese data asook bestaande werk oor die degradering van suikerriet as gevolg van omgewings- en ander eksterne faktore. Die TSOSP word in hierdie proefskrif opgelos deur die ontwerp van 'n generiese argitektuur vir 'n konseptuele besluitsteunstelsel (BSS) vir die TSOSP in die onderskeie kontekste waarna hierbo verwys word en wat die e ekte van uitsonderlike omgewingsfaktore in ag neem, asook die daarstelling van 'n rekenaar-ge ïmplementeerde weergawe van 'n daadwerklike BSS vir die TSOSP wat in die raamwerk van hierdie generiese argitektuur pas. Die boustene van hierdie BSS sluit modelle in vir die voorspelling van suikerrietopbrengs, die herwinbare waarde van suikerriet onder normale omstandighede, die verwagte koste geassosieer met 'n oesskedule en die negatiewe e ekte van omgewingsfaktore op die herwinbare waarde van suikerriet. Die werking van die BSS is gebaseer op 'n kombinatoriese optimeringsprobleem wat aan die welbekende asimmetriese handelreisigersprobleem met tyd-afhanklike kostes herinner, en hierdie model word benaderd opgelos deur middel van 'n eienskap-gebaseerde tabu-soektog waarin beide lokale en globale skuiwe ge ïnkorporeer is. Die BSS word ook gevalideer in terme van die haalbaarheid en kwaliteit van die skedules wat dit oplewer, soos geassesseer deur ervare kundiges in die suikerrietbedryf.
Description
Thesis (PhD (Logistics))--University of Stellenbosch, 2010.
Keywords
Sugarcane harvest scheduling, Sugarcane harvesting -- Decision support systems, Combinatorial optimisation, Dissertations -- Logistics, Theses -- Logistics, Dissertations -- Operations research, Theses -- Operations research
Citation