Multi-stage production planning with special consideration of energy supply and demand

This cumulative dissertation consists of five papers published in different scientific journals. All five papers are concerned with multi-stage production planning. Due to differing foci of the papers, this dissertation is divided into two parts. Part A embraces Papers 1 to 4 and contributes to a research stream that investigates energy aspects in multi-stage production planning. Part B features Paper 5 and investigates the influence of worker learning and forgetting on multi-stage production systems. Aside from the differing foci, the five papers also vary in the methodologies employed. The first paper presents a systematic review of the state of the art of decision support models for energy-efficient production planning. The remaining four papers develop mathematical models for supporting production planning decisions considering different measures to foster energy efficiency (Papers 2 to 4) and considering human factors (Paper 5). Papers 2 to 4 analytically solve the developed mathematical models. In contrast, Paper 5 draws on discrete-event simulation to derive effective production control policies. The following paragraphs summarize the five papers. Paper 1 systematically reviews the literature on quantitative decision support models which integrate energy considerations into mid-term and short-term production planning of manufacturing companies. The sampled articles are then classified and synthesized with regard to the characteristics of the modeling approaches representing different energy aspects. Based on the discussion of the sampled articles, Paper 1 identifies future research opportunities in the area of energy-aware production planning and thereby sets the stage for Papers 2 to 4 of this dissertation. Paper 2 studies how waste heat rejected by manufacturing processes in a two-stage production system can be utilized to foster energy-efficient production planning. Among the different ways of recovering waste heat, Paper 2 focuses on the conversion of waste heat into electricity using an Organic Rankine Cycle (ORC). To this end, it first describes this thermodynamic conversion process mathematically and then integrates it into a lot sizing model such that the electricity from the recovered waste heat supports the energy supply of the production stages. In a next step, Paper 2 proposes a solution procedure which derives optimal values for the lot size, the production rates of the two production stages, and the number of shipments between the two production stages that minimize production- and energy-related costs. In a numerical analysis, Paper 2 investigates how considering waste heat recovery in production planning can effectively reduce energy consumption in manufacturing and how it impacts production planning decisions. Paper 3 extends the model developed in Paper 2 and studies the use of an ORC-based waste heat recovery system (WHRS) combined with an electrical energy storage system (EESS). With the help of an EESS, generation and consumption of electricity from the WHRS can be decoupled. Using mixed integer linear programming (MILP), Paper 3 proposes a mathematical model that integrates time-varying energy prices alongside the technological processes of the WHRS and the EESS into the production planning problem of a serial multi-stage production system. This MILP model determines when production stages should process and how the WHRS and the EESS should be operated to optimize production- and energy-related costs. In a numerical analysis, Paper 3 examines how attaching an EESS to a WHRS can enhance its relevance for energy-aware production planning, particularly through providing the opportunity to store energy generated from waste heat in times of low energy prices and to then use it in times of high energy prices. Similar to Papers 2 and 3, Paper 4 also contributes to the research stream on energy-aware production planning. However, in contrast to the preceding papers, it focuses on the integration of onsite wind power into production scheduling of a flow shop system. Coordinating production scheduling and the energy supply from an onsite wind turbine poses a major challenge to researchers and practitioners as the intermittent character of wind power due to the vagaries of wind speed adds a stochastic component to production scheduling. The approach suggested in Paper 4 overcomes this challenge by first generating a large number of wind power scenarios that characterize the variability and the time dependence of wind power over time. A systematically reduced subset of these wind power scenarios subsequently serves as an input to a two-stage stochastic optimization procedure. Based on the reduced wind power scenario set, this procedure first computes a production schedule and energy supply decisions that minimize the total weighted flow time and the expected energy cost. The energy supply decisions derive whether the electricity generated by the wind turbine during a given time slot should be used to support the energy supply of the machines or be fed into the grid and thus determine the amount of electricity that needs to be drawn from the grid to guarantee an uninterrupted energy supply of the machines. These energy supply decisions are adjusted in a second step in real time as the actual wind power data are gradually revealed. In a numerical example, the effectiveness of the procedure in incorporating energy supply from non-dispatchable renewable energy sources (RES) in production scheduling is shown under various conditions. Part B of this dissertation consists of Paper 5. As Papers 1 to 4, Paper 5 is also concerned with efficiently managing multi-stage production systems. Papers 1 to 4 concentrated on how to effectively tailor the operation of production stages to energy supply from WHRSs or RES and time-varying energy prices. In contrast to these works, Paper 5 focuses on how to attune the operation of production stages to human characteristics such as individual worker learning and forgetting. To this end, Paper 5 first develops a generic simulation model of a serial multi-stage production system subject to learning and forgetting effects. Subsequently, it carries out an extensive simulation experiment to identify parameters of the production stages and their interactions which exercise a significant influence on system performance. Paper 5 then proposes flexible buffer management rules to counteract the impact of adverse production parameter combinations detected in the preceding simulation experiment. In a second simulation experiment, the performance of these buffer management rules is evaluated under various input parameter combinations.