Over recent years numbers of electric vehicles (EVs) in the Netherlands have shown a strong growth, from 1.100 in January 2012 91.000 in April 2016 [1], and sales continue to grow [2]. With a total number of 8 million cars in the Netherlands in January 2015 [3], and governmental policy aimed at 1 million EVs in the Netherlands in 2025[1], the growth potential of EVs in the Netherlands is high. For facilitating charging possibilities for these vehicles, numbers of public EV CPs in the Netherlands have grown simultaneously from 1.250 in January 2012 to 7.844 in April 2016 [1], and are also expected to continue to grow. Not only will the demand for public EV charging infrastructure increase with the growth in EVs, around the world governmental policy is also aimed at a continued large scale rollout of public charging infrastructure [4–7].
The public EV charging point rollout by the national government was started in 2009 and was aimed at overcoming the chicken-egg problem between EV sales and EV charging infrastructure. The national governmental policy until 2013 has been that local and regional governments could apply for a charging point EVnetNL (former Stichting E-laad) which managed the application, selection and installation procedure CPs. The national government subsidized these procedures [8].
Two rollout-strategies were used. In the first strategy, the applications were based upon a request by an electrical driver for a charging point near to home, a so called ‘demand-driven’ charging point. In the second strategy, the applications were based upon a decision by a local or regional government to place a charging point near public facilities (e.g. governmental buildings, shopping malls) or on strategic locations where (occasional) use was expected (e.g. sporting grounds and leisure locations), a so called ‘strategic’ charging point.
A demand-driven rollout suggests that the resulting CPs have at least one dedicated user and have a higher probability to be in a residential area. A strategic rollout suggests that the resulting CPs are used by a wider variety of users, that the use of certain CPs could be related to the opening times of nearby facilities, and that they might be in low-populated areas with a low demand for charging. This raises the question whether demand-driven CPs perform better than strategic CPs.
Research on public charging infrastructure planning and rollout has been performed since the uptake of EVs started [9,10]. Most studies focus on the location of charging infrastructure in relation to expected users and their charging patterns [6,11–14]. The assessment of the different charging infrastructure rollout strategies in relation to their performance has become an important aspect to urban planning [15–17]. So far, most studies tend to focus on only the early phase of rollout (100-500 CPs) [18], use a limited database in both charging transaction quantity and time-scale [19–21] or simulate EV charging data based on secondary data sources [22–24].
This study focuses on determining the differences in performance metrics of the two rollout strategies. Moreover, to advance understanding of charging infrastructure performance the root causes of performance differences between the two rollout strategies are studied. The paper aims to contribute to understanding whether, when and under which circumstances municipalities may be more successful in charging infrastructure development by using a demand-driven of strategies rollout strategy. It uses a real-world database of more than 1 million charging transactions, unequal in size compared to previous literature [20,25,26].
2. Performance measurement of charging infrastructure
Since the start of EV adoption literature has been written on the potential impact and risks of electric vehicles on the electricity grid under circumstances of large EV adoption [21–23]. There is little literature on charging infrastructure performance assessment of charging infrastructure. Few examples of research connecting stakeholders directly concerned with decision making for charging infrastructure rollout to performance are [15,27,28].
In [28] an analysis of the economic feasibility of international charging infrastructure business models is provided from different roles in the EV-ecosystem. From [17,28] the yearly transaction size (in kwh) was indicated as driver for economic feasible business models, next to additional pricing options. Therefore, this research uses the weekly energy transaction as one the performance indicators. The weekly energy transaction refers to the amount of energy transferred during the transaction, and is therefore both an indicator for the effectiveness of use of the charging point.
In [15] performance measures performance measures were drafted in terms of (1) effectiveness; does charging infrastructure facilitate EV adoption and (2) efficiency are CPs used in an efficient manner. To compare the performance of demand-driven and strategic CPs this research considers the performance indicators of policy makers at municipalities and charging point operators.
The number of unique users is an indicator for the diversity in users by which a charging point is used and can be related to the facilitative role of municipalities, as decision making unit (DMU), of charging infrastructure for EV users. While this performance metric is valid as a metric, a limitation is that the total number of sessions per user on a CP is not taken into account. From this perspective a strategic charging point with 8 unique users and a total of 12 sessions outperforms a demand-driven charging point with 2 users and 20 sessions.
The connection duration refers to the duration in which a vehicle is connected to the charging point and can be regarded as a performance measure of the intensity of use. Depending on the local business model, the connection duration could also be part of the revenue performance for charging point operator [17,29].
The charging time ratio is the ratio between weekly total charging time and connection time and refers to the efficient use of a CP. This ratio mentioned as load flexibility is used in the aforementioned optimization studies as basis for the optimization algorithms [30], such as time dependent rescheduling of energy transfer from peak demand to lower demand hours. This implies that a low charging time ratio is may be inefficient from an infrastructure usage point of view, while it is effective for smart charging. The complexity of load flexibility in combination with time of day is not considered in this research for the purpose of this paper is to compare the performance of two rollout strategies rather than defining the optimization potential. Table 1 displays an overview of performance indicators.
Table 1: Charging point data structure
Performance indicator Measured achievement
Weekly number of unique users Facilitation of charging infrastructure
Weekly connection duration charged per charging point
Intensity in use of charging infrastructure
Weekly amount of kWh charged per charging point
Effective use of charging infrastructure
Weekly charging time ration
Efficient use of charging infrastructure
2.1. Hypothesis
The following hypothesis were setup and tested against the available data. (H1) Based on the definition and locations, strategic CPs are expected to outperform demand-driven CPs on the user facilitation [31]. Demand-driven CPs are installed due to one or more requests and are expected to start with one user weekly user from installation date with a potential increase over time, whereas strategic CPs could start a many levels and develop in any direction over time depending on the total EV population.
Regarding the total weekly connection time the following hypothesis was developed; (H2) Demand-driven CPs are expected to display longer weekly connection times than the strategic CPs. This hypothesis is based on the idea that different charging strategies occur at different locations, such as home charging strategies with overnight charging and random charging for long and short parking times[22].
Next hypothesis; (H3) Demand-driven CPs are expected to perform better on weekly energy transfer than strategic CPs. This hypothesis is based on the suggestion that demand-driven CPs will have at least one regular user charging with its use pattern, while strategic CPs are assumed to display a combination of irregular use patterns [16,31]. Moreover, the demand-driven locations, particularly home locations are expected to be at an end of trip location or end of day location, while strategic CPs could be at any position in the daily travel activities of an EV[32,33].
The last hypothesis is on the efficiency of both types of CPs; (H4) Strategic CPs are expected to outperform demand-driven CPs better on efficiency [25]. This hypothesis is based on research on the V2X potential in residential and non-residential areas [25,31] that mentions load flexibility for non-residential areas.
3. Method
At the start of this research several datasets were gathered from Charging Point Operator (CPO) EVnetNL [34]. The data was cleaned, transposed and aggregated in line with [HELMUS HOED ]to compare the performance metrics and gain insight in the underlying use patterns. Next, the public CPs in the Netherlands were placed in historical context in order to reveal potential relations between the context and root causes over time.
The next step was to test hypotheses against the data and interpret the results in the context of charging point rollout over time. The varying number of installed CPs over time causes the mean age of CPs to vary over time a well. For example, at any given time the distribution of ages of CPs contains of just installed CPs as well as old CPs. The age of a charging station may be related to the local maturity of EV adoption in its vicinity and may therewith influence the performance metrics as well. Next, uptake period after installation may be influencing the total performance of both populations over time. Therefore, two types of tests were performed (A) time series based on date that compare the performance metrics over time and (B) longitudinal series on the age of a CP that compares the performance metrics based on the performance in the nth week after installation.
A linear regression was performed on B from week 13 to 104 (steady state period after installation). The combination of both hypothesis testing is expected to file out potential contextual biases. Relevant use patterns as root causes were found by matching existing use patterns from literature against the calculations of the performance metrics in table 1.
3.1. Data
The data used for this paper are provided by Charging Point Operator (CPO) EVnetNL, and consists of two datasets. The first dataset consists of data regarding details on EVnetNL CPs. This dataset was extracted from the EVnetNL installation management system, in which details on charging point applications, the installation, the maintenance and related financials are stored and monitored. Table 2 shows the headings of this dataset. The second dataset consists of data regarding details of charging transactions performed on these CPs between January 2012 and March 2016. Table 2 shows the headings of this dataset.
Both technically invalid transactions (incomplete data, negative energy transfers, connection durations below fifteen seconds) in line with previous EV data analysis research [35] and practically erroneous transactions (connection shorter than five minutes: for demand-driven CPs these faulty transactions are 1,8% of all transactions, whilst for strategic CPs this amounts to 2,6%) were filtered from the initial dataset. After filtering these datasets provide data on 1.007.137 transactions on 1.742 different CPs by use of 55.3850 unique charging cards. Of these CPs, the application types, 1.011 (58%) were strategic, and 731 (42%) were demand-driven.
Table 2: Charging point data structure
Data Definition
Charging point code ID code for charging point
Charging point address Street, number, postal code, municipality
Rollout-strategy Applied for by EV driver (demand-driven), or by government (Strategic)
Charging point model Manufacturer and model
Connectors Number of connectors on charging point
EAN grid connection code Identification code for grid connection
Connection installation date Date of first time charging availability
Table 3: Charging transaction data structure
Data Definition
Transaction code ID code for transaction
Charging point code ID code for charging point
Connector code ID code for connector
Charging card code (RFID-code) ID code for charging card
Started Timestamp at the start of transaction (EV connected)
Ended Timestamp at the end of transaction (EV disconnected)
Connection duration Duration of connection
Charging duration Total time wherein energy transfer took place
Energy-transfer (kWh) Energy transferred
The number of users that have used CPs was derived from the number of charging cards that have been used for identification on the charging point, as no details are registered on the identity of the user itself. Two potential situations might affect the deviation of charging cards from the unique number of users; one person using more charging cards and more users using the same charging card. The former could occur in case of accidently lost cards. the deliberate change of charging card provider or the use of more charging cards at the same time, e.g. due to tariff differentiation.
In case of lost cards or a change in service provider, the effect on the performance metrics was expected to be minimal since the usage of the old card stops as soon as the new cards is being used. The impact of users having more charging cards at the same time is expected to be minimal in this research for this is only beneficial in case of using specific cards for specific CPOs. In this research a dataset of one CPO was used and therefore number of users using more charging cards at the same time was expected to be minimal. Despite all possible situations, more unique cards being used was seen as an indicator for more unique users at a charging point.
4. The case of charging infrastructure rollout in the Netherlands
The analysis of the historical developments of charging infrastructure rollout by EVnetNL enables a better understanding of the context of both rollout strategies. The number of public EVnetNL CPs installed in the Netherlands between January 2009 and March 2015 and the adoption of EVs are shown in figure 1. Since the public charging point rollout started in 2009 [16], and the first EVs came on the Dutch market in 2012, most CPs up until 2012 were strategic, and very little demand-driven applications were received before that. Before January 1st 2012, 480 strategic CPs were installed against 58 demand-driven CPs. Due to budgetary reasons the Dutch national government stopped the application process for subsidised public CPs in January 2013, and all requests for demand-driven CPs were put on hold for five months [36].
Figure 1A: Public charging point rollout in the Netherlands between 2012 and 2015
Figure 1B: Monthly unique users on CPs and EV adoption in the Netherlands
Between Jan-Jun 2013, only CPs that were already accepted in 2012 were installed at the start of 2013, which results in the smaller increase in CPs in this period. From June 2013 the CPs from the applications that were put on hold were resumed, and the installation of these CPs was finished in October 2014 resulting in a large growth of CPs [37]. From November 2014 to of January 2015 no CPs were installed which were subsidised by the government.
In the period between 2012 and 2015, the CPs from two different rollout-strategies grew with comparable numbers: 531 strategic and 673 demand-driven CPs were installed. Therefore, to ensure fair comparison in hypothesis testing on longitudinal data the analysis was conducted from CPs installed between January 2012 and January 2015. This resulted in a comparable distribution of start weeks and number of CPs per week after installation for the two CPs types.
In figure 1B the EV demand per population and the FEV and PHEV adoption both are plotted over time. It can seen that at the end of 2013 the EV sales, particularly for PHEV, display a sudden rise due to an at that time expected decrease in fiscal benefits for EV ownership starting 2014 [38]. The sudden increase of EVs and subsequent demand-driven CPs may invoke a temporary imbalance between demand and supply of CPs, due to the installation delivery throughput time. The growth of particularly PHEV is expected to be reflected in the performance analysis.
Figure 2 shows that the CPs are located all throughout the Netherlands, and the map shows CPs in the Netherlands on January 1st 2015, and differentiated by rollout-strategy, where several aspects stand out.
First, the map shows that most CPs in general and especially demand-driven CPs are not uniformly distributed across the country. Most demand-driven CPs are located in the highly populated area of the ‘Randstad’ (the triangle between Amsterdam, Rotterdam and Utrecht), and in larger cities in rural areas such as Eindhoven, Tilburg and Arnhem. This is to be expected, as a higher population density brings a higher probability for the use of EVs, (and thus a higher demand for CPs), and people in higher-density areas might be dependent on public EV charging as they do not own ground on which to install a private charging point.
Figure 2: Public charging point locations in the Netherlands on January 1st 2015
The second, aspect that stands out in the map is that strategic CPs are more evenly dispersed geographically over the country. This means that both in high and low populated areas strategic CPs can be found.
In conclusion by observing the case of e-mobility and charging infrastructure in the Netherlands we conclude that (1) Longitudinal performance analysis needs to be limited to data of CPs installed between 2012 and 2015 in order to do fair hypothesis testing (2) A mismatch in supply and demand could occur in time series data due to the rise of EV sales and (3) The distributions underlying the metrics need to be compared as part of the root causes for performance differences.
5. Results
5.1. Weekly number of unique users
The results for hypothesis that strategic charging point outperform in facilitating EV users more than demand-driven CPs (H1) is displayed in and figure 3A and B. As mentioned in section 3.1, the number of unique charging cards are expected to reflect the number of unique users. In figure 3A (time series) and figure 3B (longitudinal) a comparison of the average number of unique charging cards per week per charging point for populations. As a fraction of users at the total population
Figure 3A: Average number of unique charging cards per week per charging point, 2012 – Mar 2016
Figure 3B: Average number of unique charging cards per week per charging point after installation date
Both figures show an increasing amount of weekly users for both types of CPs, which is expected given the growth of EV sales at the same time. In support of hypothesis (H1), an increasing difference in between both types of CPs is seen in favor of strategic types: strategic CPs thus seem to facilitate more users than demand-driven CPs.
While before 2013 strategic CPs were used by less unique cards than their demand-driven counterparts, during 2014, strategic CPs showed on average 20%-25% more unique CPs per charging point in comparison with demand-driven CPs, and in 2015 this difference increased to on average 1.3 more weekly users for strategic CPs.
Particularly interesting is the moment of takeover in figure 3A around Oct-Nov 2013. This corresponds to the sudden peak in sales of EVs in the Netherlands. At the start of the takeover for populations a temporary increase is shown in weekly unique users, which suggests that figure 3A reflects the total number of EVs on the market as well, rather than only an increased use of CPs. The root cause analysis may reveal the relation of EVs on the market affect the growth of use on strategic CPs more than at demand-driven CPs, which could underline the importance of rolling out strategic CPs at times that a large numbers of EV sales take place.
The uptake graphs shows a linear growth for both types of CPs (DD/Str: 0.15 vs 0.28 users/week, R2=0.96 vs 0.95). The value of 1.5 for both types of CPs at the first week of installation is higher than the expectation of a single user at the start.
From figure 3A and B it is concluded that the data support the hypothesis that over time strategic CPs outperform demand-driven CPs on the number of unique users (H1). We expect that the difference will increase over time as adoption of EV increases.
5.2. Weekly connection duration per charging point
The hypothesis (H2) stated that the demand-driven CPs were expected to outperform the strategic CPs on weekly connection times. Figure 4A and B display the weekly total connection duration overaged on all the CPs within the data. Regarding average total connection duration per week, shown in figure 4A, it can be observed that the demand-driven CPs have on average structurally longer than at strategic charging point. This supports the acceptance of (H2).
In figure 4A it can be seen that both types of CPs display a structural growth over time, where the relative growth of strategic CPs is higher than the demand-driven CPs. Next, it can be seen that while both types of CPs display a seasonal pattern with dips during holiday seasons, the impact of seasonality is much higher at demand-driven CPs.
In figure 4B it can be seen that the uptake of connection time increases linearly over time for both types of CPs. Next, the typical adoption pattern in the first weeks after installation are seen for demand-driven CPs. A reason for this could be that EV users may require a period of time to familiarize themselves with the EV and establish their charging habits [39]. An interesting finding is that the difference in the slopes of both types of CPs is insignificant (DD/Str: 0. 22 vs 0.24 kWh/week, R2 : 0.96 vs 0.95), which is contrary to figure 3B.
Refera at age distribution ove time and development …. This suggests that H2 may remain accepted for a long term.
Figure 4A: Average total connection duration per week for two rollout-strategies, 2012 – Mar 2016
Figure 4B: Weekly average total connection duration per week after installation date
5.3. Weekly energy transfer per charging point
The third hypothesis (H3) stated that demand-driven were expected to outperform on weekly energy transfer. In figure 5A (time series) and B(longitudinal) the weekly energy transfer per charging point are shown.
From in figure 5A it can be seen that demand-driven CPs display a structurally higher weekly energy transfer than at the strategic CPs, which confirms H2. From a business case perspective this suggests that for the current data demand-driven CPs are more attractive to charging point operators. Next, it can be seen that both types of CPs display a seasonal pattern with a typical dip during holidays. Furthermore, like in figure 3 it can be seen that the sudden EV increase at the end of 2013 results in an increase of weekly transaction size as well. Finally, the difference between the weekly transaction sizes tends to be decreasing in favour of strategic CPs (min 2.65, max 14.4, µ: 39.2).
The uptake graph figure 5B displays a steady growth of the weekly energy amount for both types of CPs (DD vs Str: 0.20 vs 0.31 kWh/week; R2 0.94 vs 0.96). Note that the seasonal pattern is averaged out due to the different installation dates resulting in a linear growth for both populations. An interesting detail in figure 5B is the curve of demand-driven energy take up found in the first 10 weeks for demand-driven CPs and 5 weeks for strategic CPs.
Figure 5A: Average total energy transfer per week for the two rollout-strategies, 2012 – Mar 2016
Figure 5B: Uptake of average energy transfer (kWh) per week after installation date
While both graphs support the validity of H3 for at least the time between 2012 to 2016, it is questionable how long this difference remains for figure 5B suggests an takeover of strategic CPs. Root cause analysis of use patterns may reveal more insight in the development of this performance metric.
Lets just install demand since it is more
5.4. Weekly charging time ratio per charging point
The last hypothesis (H4) stated that strategic CPs were expected to outperform demand-driven CPs in terms of charging efficiency, measured in charging time ratio. Figure 6A and B display the charging time ratio over time and after installation. In figure 6A it can be seen that the charging time ratio of strategic CPs is continuously higher than its demand-driven counterparts (min 0.07, max 0.30, µ: 0.16). Next, it can be seen that the charging time ratio of strategic CPs slightly decreases over time until the mass uptake of EV in 2013 and thereafter it stabilize. Demand-driven charging CPs to stabilize. Typically, neither the seasonal pattern displayed in figure 4A nor the effect of EV increase were found in figure 6A. This suggests the charging time ratio is an constant element of EV users’ charging behavior and at least not depending on seasonality.
Next, figure 6B displays a slight decrease for both types of CPs after installation having the same slope (DD vs Str: -1.00e-03 vs -1.14e-03 kWh/week; R2 0.93 vs 0.85).. This indicates that the efficiency of both type of CPs is slightly decreasing after installation and that the hypothesis is expected to be valid for a longer term. Both figures support the validity of (H4).
Figure 6A: Average charging time ratio per week for two rollout-strategies, 2012 – Mar 2016
Figure 6B: Charging time ratio per week after installation date
6. Root cause analysis of performance metrics use patterns
6.1. Weekly number of unique users
From section 5.1 it appeared that the strategic CPs facilitated more users than demand-driven CPs from 2014 onwards. There are three possible factors that are the root causes of this; (1) the general effect of supply versus demand (2) the short and long term effect of population size on non-incidental users and (3) the effect of the probability for incidental users on the unique amount of users.
Supply and demand are part of the chicken and egg problem of charging infrastructure [35]. In particular, a shortage in CPs is expected to lead to EV users searching for CPs, either strategic or demand-driven, in the vicinity of their preferred charging location [40]. The number of weekly unique users at both types of CPs is regarded as EV demand per population and expected to be a proportion of the total EV adoption in the whole country.
Figure 1B shows that for both demand-driven and strategic CPs the EV demand matches the EV adoption in shape up to 2014. From 2014 two aspects strike out (1) during the mass EV adoption from the end of 2013 the ratio of monthly unique users at demand-driven charging infrastructure and EV adoption drops from 1:10 to 1:15 (2) both type of CPs, particularly the strategic CPs, display non-monotonic growth from 2014 until summer 2015.
From the combination of figure 1A and B it can be seen that the total and population specific supply and demand ratio have decreased from 2014 onwards. The long term effect of population size on the ratio of monthly unique users of both populations can be explained by the fact that more typical demand-driven users use strategic CPs than vice versa. Given the marginally increase of supply side (see figure 1A) the ratio of unique users is expected to be shifting in favor of strategic CPs.
The effect of the population size has a short and long term effect on the increased number of non-incidental weekly users. A short term is seen in the non-monotonic growth of unique charging cards in figure 1B and figure 3A during high increased population growth. It can be related to a mechanism of new EV users temporarily using the public charging infrastructure in times between acquiring their EV and installation of a home or demand-driven charging point. This results in a temporary increase in weekly EV users and decrease after install of CPs. From of figure 1B it appears that this effect is larger for strategic CPs than demand-driven CPs, resulting in a larger addition of weekly new users for strategic CPs.
A long term effect is found in the fact that demand-driven CPs are installed for specific users and the strategic ones for general use nearby locations of expected demand. A deeper analysis on the recurrence ratio of EV users strategic CPs revealed that 15% of EV users with more than 10 transactions remain active 26 weeks after their first session at a strategic charging point. This can be explained by a shift from visitors to workers or daytime chargers.
Figure 7: frequency polygon of incidental uses of charging infrastructure
The incidental users, in some literature regarded as visitors [15,41] are part of the root causes for (H1) as well. Due to their location and purpose the strategic CPs the have a higher probability of attracting visitors. In figure 7 the frequency polygon of the number of sessions on the whole population per charging point type per EV user is shown, focusing on the domain of 1 to 30 charging sessions. This shows that in the incidental users occur more often at strategic CPs than demand-driven CPs.
The addition of incidental users to the non-incidental weekly users explains that the strategic CPs to outperform the demand-driven CPs on this performance metric. Figure 1B shows that over time the difference between strategic and demand-driven is increasing. We expect this increase to continue until strategic CPs have a maximal incidental use. This will most likely happen when the incidental and regular demand around strategic CPs saturate the supply. In the dataset we don’t yet see this condition.
6.2. Weekly connection duration per charging point
In section 5.2 it was shown that demand-driven CPs outperform strategic CPs on total weekly connection duration (figure 4A) and the longitudinal graph (figure 4B). This suggests that the performance difference might remain stable. The explanation of the total connection duration can be found in the distribution pattern of the duration and timing of charging sessions. These both relate to use type characteristics of charging behaviour [41].
Figure 8: Distribution of charging transactions in thousands per 5 minutes’ connection duration
From the distribution of charging sessions duration (figure 8) it appears that longer connections occur more often on demand-driven CPs while shorter connections occur more often on strategic CPs. Next it shows that for both types of CPs the connection distributions are multimodal, suggesting that the total distribution is composed of several sub distributions of use patterns. Particularly interesting is that in figure 8 the tops and valleys of the sub distributions appear to occur at the same connection time while the intensity of the peaks differ. Figure 8 suggests three distributions (i) 0-6 hours (ii) between 6-12 hours and (iii) larger than 12 hours. The latter is less clear for strategic CPs than for demand-driven CPs. This indicates that the charging intentions are similar for both types of CPs while the local use type composition is different. For demand-driven CPs figure 8 also suggests another distribution of sessions shorter than 1 hours. From analysis of the top 20 locations with sessions shorter than 30 minutes (~6500 transactions) it appeared that all locations could be related to vicinity to Points of Interests evoking very short sessions for EV users needing a parking spot nearby the POI. The difference in connection distributions for both types of CPs suggest that the underlying dynamics of time based use patterns of start and stop of connections and energy demand is different as well.
The use patterns at the CPs with regard to the time of day is shown using the time at which connections are started and ended. Connection distributions of start (figure 9A) and stop times (figure 9B) are shown below. The connection stop distribution shows the opposite to figure 9A; the evening shows comparable activity, however connections are ended more in the morning (between 7 AM and 11 PM) on demand-driven CPs.
From the connection duration distribution, it was found that both types of CPs displayed several typical connection duration distributions. A deeper analysis of starting time distribution for short connection times (less than 6 hours) and long connection times (more than 6 hours) is shown in figure 10. This figure clearly reveals that short sessions tend to start charging during daytime. For short sessions, both types of CPs have the same distribution shape with small peaks around 9AM, 1PM and 18PM. As expected from figure 8, the short sessions of strategic CPs exceed the demand-driven CPs throughout the whole day.
Figure 9A: Connection starts distribution on working days, 2012 – Mar 2016
Figure 9B: Connection stops distribution on working days, 2012 – Mar 2016
Particularly interesting are the different peaks of both type of CPs for sessions longer than 6 hours. Demand-driven CPs have a much larger peak for overnight chargers, corresponding with typical home charging, whereas the peak for morning chargers, typically office charging, is higher at strategic CPs. The morning start-peak on strategic CPs is twice as high compared to demand-driven CPs, and the evening start-peak on demand-driven CPs is twice as high compared to strategic CPs. The connection stop distribution leads to the idea that charging sessions at strategic charging point are not end of daily trip sessions and could therefore have a transaction volume.
Combining the graph of figure 10, figure 9A and figure 9B suggests that demand-driven CPs are typically used by residents with overnight charging, whereas strategically placed CPs are typically used by daytime chargers. For the weekend connection distributions results show that activity is evenly spread (a normal distribution), throughout the day (between 7 AM and 12 PM) on both charging point categories.