Technology forecasting of electric vehicles using data envelopment analysis

This research performed technology forecasting (TF) of electric vehicles (EV) using data envelopment analysis (DEA) with the purpose to determine to what extent TFDEA can be applied to predict the technological progress of electric vehicles. This study was commissioned by SKF, who is interested in having a useful forecasting tool to analyze EV technological advancements and identify whether one of the existing EV configurations has potential to become the dominant design in the future. SKF dedicates a major part of its resources to supplying the car industry; therefore changes in the automotive industry may impose technological changes on their current development projects and state of affairs. New market opportunities or threats brought about by the introduction of electric vehicles need to be signaled in due time to be able to adjust corporate and research activities to better serve the car industry and maintain a strong market position. Electric vehicles are available in several configurations: battery electric (BEVs), hybrid electric (HEVs), plug-in hybrid (PHEV), and extended-range electric vehicles (EREV). This indicates that electric vehicles represent a heterogeneous class of products with different technical and performance specifications. Therefore, two aspects need to be accommodated by the forecasting technique used to produce electric vehicle forecasts: ? EV technology is not homogeneous, therefore the forecasting method should allow for the simultaneous analysis of different EV configurations in order to measure and predict technological change over the whole class of EV technology. ? EVs are characterized by several performance attributes which may be differently valued by different user categories, therefore the forecasting method should allow for multi-criteria evaluation of the technology performance and technological progress. This research used TFDEA to forecast the technological progress of electric vehicles. The reason is that TFDEA appeared to have significant advantages over conventional trend extrapolation methods. Unlike conventional techniques, TFDEA can simultaneously evaluate multiple technologies using multiple variables. Furthermore, TFDEA is an extreme point method which means that it can calculate the individual performance of an observation instead of calculating the average performance over the data set. For this reason TFDEA is able to identify the state-of-of the art frontier (i.e. the best performing technologies at a given time). In addition, TFDEA can determine rates of technological change without assuming non-correlated attributes and can account for dynamic trade-offs between performance parameters. The description of TFDEA fits the requirements identified for the forecasting method needed for EV technology. Remaining concerns about the usefulness of the method were related to the amount of data needed for the model to produce reliable results and the inherent assumptions of TFDEA listed below: 1. Technology performance is a linear function of the technology inputs. 2. The inputs of technology remain constant over time. 3. The rate of technological change remains constant over time. The focus of this study was to identify the impact of these assumptions on the accuracy and validity of the EV forecasts. A technical system analysis of electric vehicles was performed to provide understanding of the basic operation mechanisms of EV systems and of the relations between different EV design variables. Such information was necessary in order to properly identify and select those design parameters that are responsible for the EV performance and which can pose limitations to further technological advancements. For both families of vehicles, the output power of the propulsion unit, the charge storage capacity of the battery and the vehicle weight were found to be the main determinants for EV performance. In this study, the performance of BEVs was expressed in terms of acceleration possibilities and driving range, while for HEVs fuel economy, CO2 emissions and acceleration were selected as key performance indicators. The technical and performance attributes of EVs were used as inputs and outputs respectively in three TFDEA models. Two of the models ware applied on BEVs or HEVs only and were used to evaluate the individual technological progress of BEVs and HEVs as homogeneous products, while the third model was used to determine the rate of technological change over the full class of EVs. Each model was used to produce a forecast for yearly EV performance levels until 2020. These forecasts were verified for accuracy against a set of existing products. Then, it was analyzed how the data availability and the assumptions of the TFDEA model impact the reliability and validity of the forecast. The results of the analysis are shown below: • For the first 11-12 time periods all vehicles in the data set were ranked as SOA, therefore no rate of change could be calculated, which reduced significantly the possibility to analyze whether there is a visible pattern of constant progress. This was caused by the large number of attributes included in the model, combined with a low number of products released over a relatively short time window. • TFDEA assumes linear relations between technology inputs and technology performance. For electric vehicles, this assumption is realistic to a limited extent. The results showed that in the case of battery electric vehicles there seems to be a linear relation between battery capacity and electric range. • It was shown that TFDEA models consistently underestimated the performance parameters subjected to regulation. This indicates that the method is very sensitive to exogenous drivers of technological change. The EV case study shows that the approach to evaluate the performance of a technology as a linear function of its inputs may be an oversimplification. • The results of the three models show that the TFDEA cannot anticipate the introduction of potentially disruptive technologies, such as the PHV and the EREV. This is due to the fact that the forecasts produced with TFDEA indentify what may be feasible in the future based only on what exists today. TFDEA assumes that inputs remain constant over time and has no mechanism to identify future re-configurations of inputs which could lead to better performance. The present study has concluded that TFDEA is not a suitable method for analyzing technological progress of electric vehicle technologies. This is due to the high sensitivity to exogenous drivers and its limited capability to anticipate the introduction of potentially disruptive design configurations. These limitations are mostly a result of the assumptions that inputs and the rate of change remain constant over time. As a general note on TFDEA, it was observed that TFDEA would not be a useful forecasting tool for emerging technologies with significant economic and socio-political implications. The model could be used for mature technologies which have shown constant progress over time, given that no exogenous forces are expected to influence the technological change. Furthermore, TFDEA could be used for forecasting emerging technologies whose performance can be expressed with very few attributes (at most three times less than the number of products available), and whose performance is not targeted by governmental regulation. With respect to EV forecasting, this study identified that a simple analysis of technological progress is not sufficient to determine the evolution of EV technology. Due to the economic, environmental and political consequences, it is expected that the adoption of electric vehicles will not depend solely on performance, but also on different technological and context factors, such as battery technologies, available infrastructures, standardization opportunities, consumer acceptance, national interests and governmental support. To better understand the development possibilities of EV technology, this study recommends the use of technology forecasting and market shift indicators analysis to identify possible innovations in EV-supporting technologies, such as battery charging stations and smart grit technologies. Furthermore, combined analysis of consumer research and market structure analysis can help identify the market forces expected to affect further advancements of electric vehicles. In addition, monitoring government and industry plans can provide information on potential standardization opportunities and strategies meant to accelerate the adoption of BEVs.