Electric vehicle charging dataset with 35,000 charging sessions from 12 residential locations in Norway

This data article refers to the paper “A method for generating complete EV charging datasets and analysis of residential charging behaviour in a large Norwegian case study” [1]. The Electric Vehicle (EV) charging dataset includes detailed information on plug-in times, plug-out times, and energy charged for over 35,000 residential charging sessions, covering 267 user IDs across 12 locations within a mature EV market in Norway. Utilising methodologies outlined in [1], realistic predictions have been integrated into the datasets, encompassing EV battery capacities, charging power, and plug-in State-of-Charge (SoC) for each EV-user and charging session. In addition, hourly data is provided, such as energy charged and connected energy capacity for each charging session. The comprehensive dataset provides the basis for assessing current and future EV charging behaviour, analysing and modelling EV charging loads and energy flexibility, and studying the integration of EVs into power grids.


Value of the Data
• The global rise in Electric Vehicle (EV) adoption, anticipated to reach a projected 35 % sales share by 2030 [ 3 ], underscores the critical need for comprehensive data and research on EV charging.• A significant gap exists in the availability of EV charging data necessary for detailed analyses and modelling of EV charging patterns and energy flexibility [ 4 ].• Our dataset addresses this gap by providing detailed information on plug-in/plug-out times and energy charged from over 35,0 0 0 charging sessions across 12 residential locations in Norway, representing 267 diverse users within a mature EV market.• By employing methodologies outlined in [ 1 , 5 ], we have integrated realistic predictions into the dataset, resulting in user-friendly EV charging datasets essential for a wide range of EV research.• The predictions encompass battery capacities and charging power for each EV user, charging time, idle time, and plug-in State of Charge (SoC) for each charging session, as well as hourly data such as energy charged and connected energy capacity for each charging session.• The comprehensive dataset provides the basis for a variety of energy studies, including analyses of EV charging behaviour, load forecasting, energy flexibility, and the integration of EV charging into power grids.

Background
The use of EVs is central in meeting emissions reduction goals outlined in the Paris Climate Change Agreement [ 6 ].Globally, EVs held a 14 % market share in 2022, with Norway standing at 88 % [ 3 ].The predominant use of home and workplace charging influences the energy and power loads of buildings [ 7 ].However, EV charging can often be shifted in time without affecting user habits, making it a promising solution for utilizing energy flexibility.Consequently, smart charging is becoming an increasingly important topic [ 8 ], particularly in buildings and locations with limited grid capacity or excess solar photovoltaic (PV) electricity.
This data article refers to the paper "A method for generating complete EV charging datasets and analysis of residential charging behaviour in a large Norwegian case study" [ 1 ].The study revealed a general lack of EV charging data essential for data-driven analyses and modelling of EV charging and flexibility.The case study in [ 1 ] included a dataset with > 35,0 0 0 residential charging sessions, which is made openly accessible in this data article.

Data Description
This article describe the dataset of the case study in [ 1 ], available in the linked repository.It consists of four datasets (csv-files) as described in this section.In the data provided with this article, Central European Time (CET) zone is used, which is GMT + 1. Daylight saving time (DST) applies.
Dataset 1: Residential charging reports Dataset 1 consists of original charging report, as outlined in Table 1 .The dataset describes plug-in/plug-out times and energy charged from 35.377 charging sessions across 12 residential locations.The data for location TRO_R is further described in [ 5 , 9 ].
Dataset 2: Predictions per user Dataset 2 consists of predictions per user, as outlined in Table 2 .Dataset 3: Predictions per charging session Dataset 3 consists of predictions per charging session, as outlined in Table 3 , for all User IDs with predictions in Dataset 2. Dataset 3 can be merged with Dataset 1 (by session_id), to get the complete charging session data.Dataset 4 consists of hourly predictions per charging session, as outlined in Table 4 .

Experimental Design, Materials and Methods
This section offers a detailed account of the methods used for data acquisition, including data collection and cleaning for Dataset 1, user predictions for Dataset 2, and EV charging predictions for Dataset 3 and Dataset 4.

Data collection and cleaning for Dataset 1
The residential charging reports were collected from charge point operators (CPOs) and apartment buildings in Norway: For locations ASK, BAR, BER, BOD, OSL_1, OSL_2, OSL_S, TRO: • Data was provided by Current Eco AS [ 10 ], which develops Charge Point Management Systems.
For location BAR_2: • Data was downloaded by a housing cooperation, using the EV charging portal from EV charging manufacturer Zaptec Charger AS [ 11 ].
For location KRO: • Data was provided by the CPO Kople AS [ 12 ].
For location OSL_T: • Data was provided by the CPO Mer Norway AS [ 13 ].
For location TRO_R: • Data was provided by the CPO NTE Marked AS [ 14 ].  1 ).The data cleaning procedure is described in Table 5 .Finally, time zone and daylight-saving time (DST) corrections were made before adding calendar data, such as weekdays.
User predictions for Dataset 2 For predicting charging power and net battery capacity per user in Dataset 2 ( Table 2 ), the following steps were taken [ 1 ].

EV charging power (charging_power)
1. Assumption : Each user ID has at least one session where the charger is unplugged while still charging.2. Calculation : If unplugged during charging, the connection time equals the charging time.The average charging power ( P charging ) is calculated using Eqs.( 1) to (3) , with data from Dataset 1: plugout_time ( t plug-out ), plugin_time ( t plug-in ), and energy_session ( E charged ).CP connection time for an EV session: When plug-out during charging: Average EV charging power: Identifying key sessions: The highest P charging value for each user ID is selected as the preliminary prediction ( P preliminary ).

Filtering Errors and Outliers:
To improve accuracy, P preliminary is compared to typical EV charger capacities, categorized into three levels: • Level 1: < 4 kW (PHEVs and earlier BEVs) • Level 2: 4-8 kW (Standard BEVs) • Level 3: 8-11.5 kW (Newer/larger BEVs) 5. Validation: If a user ID has at least two sessions within the same category, P preliminary is accepted as the final charging_power prediction ( P user ).Otherwise, the session is considered an outlier, and a new P preliminary is calculated.This process repeats until all user IDs have a final P user .
• 6 IDs with P user < 2 kW were removed due to market inconsistency (too low P user ).

Net battery capacity (battery_capacity)
1. Assumption : Each user ID has at least one session where the EV battery is charged from a defined minimum to a defined maximum state of charge (SoC) level.2. Calculation : The session with the highest energy_session ( E charged ) for each user is selected.
This maximum energy value is multiplied by an efficiency factor ( ɳ = 88 %) to calculate the approximate energy stored in the battery ( E battery ), as shown in Eq. ( 4) .Maximum energy stored in battery: Battery capacity prediction: The calculated maximum energy values ( Eq. ( 4) ) are divided by an assumed SoC range for the charging session ( Eq. ( 5) ) to predict battery capacities.Two SoC ranges are used, depending on battery size: • Small/Medium (EV-SM): 10 % minimum SoC, 90 % SoC range.

Energy charged per hour (energy_charged_i)
Energy charged during sessions is distributed hourly using the methodology from [ 5 ].
Calculation : The hourly charging loads ( E load (i) ) are calculated by multiplying the EV charging power prediction per user ID (charging_power, P user ) with the hourly charging time, as shown in Eq. (6) .It is assumed that charging starts immediately after plug-in and the charging power remains constant throughout the charging time.
Charging load hour i: E load(i ) = P user × t charging (i ) where

Duration of the EV charging time (charging_time) and non-charging idle time (idle_time)
The duration of EV charging time and non-charging idle time is calculated using Eqs.( 7) and (8) .
EV charging time: t charging = E charged / P charging (7) Idle time per session: t idle = t connection − t charging (8) If the initial assumption of EV charging power (charging_power) is filtered during the charging power assumption, the EV charging time may exceed the connection time for some EV sessions.In such cases, the charging_time is designated as NA.

Idle energy capacity (idle_session, energy_idle_i)
Hourly values for idle energy capacities are predicted by using the methodology presented in [ 5 ].
Calculation : Hourly idle energy capacities for each session are calculated by multiplying the idle times by the assumed charging power for the user ID, as shown in the following equation.

Table 5
Data cleaning procedure for original charging reports.