UniSoma – Votorantim Celulose e Papel – PLANFLOR 2
PLANFLOR – Long Term Forest Planning
PLANFLOR is a decision support system developed for the Forest Planning Department at Votorantim Celulose e Papel (VCP). Developed in AIMMS language, the system is based on a Mixed Integer Programming large scale model, for optimized forest planning.
VCP has pulp production units in Jacareí-SP and Três Lagoas-MS. It’s total forest area is 314 thousand hectares of planted eucalyptus, distributed among three states: SP, MS and RS.
The production unit in Jacareí, acquired by VCP from the former Papel Simão in 1992, was the first factory in the world to produce 1 million tons of bleached pulp with a only digester, in 1995. The production is based on the Kraft process, through which a mixture of sodium hydroxide and sodium sulfide are used to dissolve lignin inside wood chips.
The Jacareí unit is supplied with wood from Forest Units Capão Bonito and Vale do Paraíba, both located in the São Paulo state. Currently, in Brazil, due to favorable climate and genetic advances in the industry, the eucalyptus tree, in less than 7 years, hits its maturity state(in terms of cost / benefit) for the production of pulp.
Once the wood is cut down, the stand can be subjected to different types of intervention. The most common is regeneration, which leads to a new growth cycle corresponding to the cloned plants. Another option is to coppice the forest. The advantage of coppicing instead of regeneration is the reduced cost (silviculture), and its disadvantage is a decrease in productivity. Such decrease generally limits in up to 3 the number times that coppicing is applied to a stand, before regeneration is necessary. A third option is a delay the generation.
The biggest goal in Long Term Forest Planning is to determine the state sequencing of various stands to guarantee wood supply according to (a) factories strategic production plans and (b) costs involved in several stand management operations. The plan should also address the need for new stands to meet growing wood demand.
Besides this, it should consider the initial state of the several stands (standing wood stock classified by age). Such boundary conditions can present unadjusted forest conditions and consequently alternative cutting ages, different from the optimal 7 years, must be considered, to reestablish a system balance. The various intervention possibilities and cutting ages can be represented as a state machine.
There are still other conditions that have to be taken in to account, such as operational and budget constraints during several of the work stages (silviculture, transport and harvesting).
In the beginning of 2006, UniSoma was hired by VCP to develop a new version of PLANFLOR. It should substitute all tools used for long term planning at VCP, based on spreadsheets, systematizing good planning practices already developed. The first project challenge was the short 3 month deadline.
Besides this, for the project, UniSoma established an internal goal of profoundly revising the PLANFLOR mathematical model, in light of the computational improvements that occurred throughout the last decade, in terms of machine processing capacity and the evolution of algebraic modelling tools and commercial mathematical solvers.
The PLANFLOR version developed in AIMMS for VCP is based on a flexible and innovative mixed integer programming model, through which the evolution of stands according to the state machine is implicitly represented in the form of binary variables. By breaking this paradigm, no possibility is discarded a priori, which used to occur in the older version of PLANFLOR, due to the computational limitations and predefined sequences of states for stands were generated through the use of heuristics.
Such formulation is similar to commercial long term forest planning solutions, with the difference that in the new version of PLANFLOR, stands are treated individually (or ad-hoc), while other control solutions generally use forest areas (a set of stands) requiring an afterwards follow up for each stand, according to approximation methods. Literature (Johnson e Scheurman, 1977) classifies models such as the old version and the new version of PLANFLOR as Type I and Type II, respectively.
The new PLANFLOR considers, amongst others, the following input data:
- Current and new forest areas – the current forest areas are imported from SAP through an ODBC interface;
- Stand inventory data to build productivity curves;
- Several costs, such as standard cost, harvest costs, transport (according to an internal VCP procedure) and costs differentiating various types of third-party land contracts such as partnerships, leases and others;
- Accounting balance and asset base;
- Forest status: control over stand states, through the use of constraints controlling the main parameters such as minimal and maximal cutting years, type, cycle, and others;
- Constraint registration, which allow fine control over the level of activities for a group of stands selected by users. Included constraints are fixed and/or relative limits as well as smoothing constraints.
The main results achieved are the following:
- Stand management plan, detailed with main operations, volumes and planned costs;
- Management supply plan;
- Graphics and tables with information regarding transported wood volumes and costs, cut wood, standing wood, silviculure;
- Graphics and tables of constraint activities included by users;
- GIS visualization, allowing the identification of special constraints previously non specified, but that are operationally relevant.
PLANFLOR allows the representation of the evolution of different forest management activities, in all stands controlled by VCP, enabling integrated plan generation for factory pulp supply throughout the long term planning horizon. Besides this, the quantification of the various types of costs has allowed for “compromised” optimized solutions, for different planning scenarios, which has been translated in to significant economical gains.
“The deployment of PLANFLOR at VCP is providing an integrated forest planning vision, their economical impacts and great time economy in the generation of scenarios, all according to the company’s directives.”
Davi de Souza Rodrigues – Forest Planning Manager – 28th of May, 2009.
|AIMMS Service Partner UniSoma