EV charging station utilisation: time-based vs energy-based measures

A charge point operator's decision

Operators in NCR decide three things from these numbers: where to add connectors, where to add a charger of a different power rating, and whether to introduce idle fees or time limits. Each decision reads a different measure. A site with high time utilisation and low energy utilisation does not need more power; it needs cars to move. A site with high energy utilisation is a candidate for another charger. A single "utilisation %" on a dashboard hides which case you are in.

The data a CPMS gives you

A charge point management system usually exports sessions built from the chargers' transaction messages, plus a status log:

Timestamps must share one time zone. Chargers left on UTC while the dashboard runs in IST shift every session by five and a half hours.

Four formulas

• Available minutes = window minutes − downtime minutes inside the window
• Time utilisation % = Σ clipped plugged minutes ÷ available minutes × 100
• Energy utilisation % = Σ kWh inside window ÷ (rated kW × available minutes ÷ 60) × 100
• Idle-while-plugged % = (plugged minutes − charging minutes) ÷ plugged minutes × 100

Clipping means each session contributes only the part between the window's start and end: MAX(start, window_start) to MIN(end, window_end).

The worked query

Synthetic data for 10 September 2026: three connectors at two hypothetical sites. S01 began the evening before, S09 runs past midnight, S04 was exported twice, and C2 was down until 8 AM.

CREATE TABLE connector (connector_id TEXT, site TEXT, rated_kw REAL);
CREATE TABLE session (session_id TEXT, connector_id TEXT, plug_in TEXT, plug_out TEXT,
                      charge_end TEXT, energy_kwh REAL);
CREATE TABLE downtime (connector_id TEXT, down_from TEXT, down_to TEXT);
INSERT INTO connector VALUES ('C1','Sector 18 mall',60),('C2','Sector 18 mall',60),('C3','Expressway plaza',30);
INSERT INTO session VALUES
('S01','C1','2026-09-09 23:10','2026-09-10 00:40','2026-09-10 00:30',38),
('S02','C1','2026-09-10 09:00','2026-09-10 10:00','2026-09-10 09:50',30),
('S03','C1','2026-09-10 13:00','2026-09-10 15:30','2026-09-10 13:45',22),
('S04','C1','2026-09-10 18:00','2026-09-10 19:00','2026-09-10 18:55',35),
('S04','C1','2026-09-10 18:00','2026-09-10 19:00','2026-09-10 18:55',35),
('S05','C2','2026-09-10 10:00','2026-09-10 11:00','2026-09-10 11:00',18),
('S06','C2','2026-09-10 20:00','2026-09-10 21:30','2026-09-10 21:00',24),
('S07','C3','2026-09-10 07:00','2026-09-10 09:00','2026-09-10 09:00',40),
('S08','C3','2026-09-10 17:00','2026-09-10 20:00','2026-09-10 19:00',48),
('S09','C3','2026-09-10 23:30','2026-09-11 01:00','2026-09-11 00:45',30);
INSERT INTO downtime VALUES ('C2','2026-09-10 00:00','2026-09-10 08:00');

WITH s AS (
  SELECT DISTINCT * FROM session
),
w AS (
  SELECT s.*, c.rated_kw,
         MAX(julianday(plug_in), julianday('2026-09-10 00:00')) AS a,
         MIN(julianday(plug_out), julianday('2026-09-11 00:00')) AS b,
         MAX(julianday(plug_in), julianday('2026-09-10 00:00')) AS ca,
         MIN(julianday(charge_end), julianday('2026-09-11 00:00')) AS cb,
         (julianday(charge_end) - julianday(plug_in)) AS charge_span
  FROM s JOIN connector c USING (connector_id)
  WHERE plug_in < '2026-09-11 00:00' AND plug_out > '2026-09-10 00:00'
),
per AS (
  SELECT connector_id, rated_kw,
         SUM((b - a) * 1440) AS plugged_min,
         SUM(MAX(cb - ca, 0) * 1440) AS charging_min,
         SUM(energy_kwh * MAX(cb - ca, 0) / charge_span) AS kwh_in_day
  FROM w GROUP BY connector_id, rated_kw
),
avail AS (
  SELECT c.connector_id,
         1440 - COALESCE(SUM((MIN(julianday(down_to), julianday('2026-09-11 00:00'))
                  - MAX(julianday(down_from), julianday('2026-09-10 00:00'))) * 1440), 0) AS avail_min
  FROM connector c LEFT JOIN downtime d USING (connector_id)
  GROUP BY c.connector_id
)
SELECT p.connector_id,
       ROUND(a.avail_min, 0) AS avail_min,
       ROUND(p.plugged_min, 0) AS plugged_min,
       ROUND(100.0 * p.plugged_min / a.avail_min, 1) AS time_util_pct,
       ROUND(100.0 * p.plugged_min / 1440, 1) AS time_util_24h_pct,
       ROUND(p.kwh_in_day, 1) AS kwh,
       ROUND(100.0 * p.kwh_in_day / (p.rated_kw * a.avail_min / 60), 1) AS energy_util_pct,
       ROUND(100.0 * (p.plugged_min - p.charging_min) / p.plugged_min, 0) AS idle_plugged_pct
FROM per p JOIN avail a USING (connector_id)
ORDER BY p.connector_id;

Energy for a session that crosses a boundary is split in proportion to its charging minutes inside the window, which is an approximation. If your system stores interval meter values, sum those instead.

What the shortcut reports

The common shortcut filters sessions by start date and subtracts plug-in from plug-out, with no clipping and no deduplication:

SELECT connector_id,
       ROUND(SUM(julianday(plug_out) - julianday(plug_in)) * 1440, 0) AS naive_plugged_min,
       ROUND(100.0 * SUM(julianday(plug_out) - julianday(plug_in)) * 1440 / 1440, 1) AS naive_util_pct
FROM session
WHERE date(plug_in) = '2026-09-10'
GROUP BY connector_id ORDER BY connector_id;

C1 loses the 40 minutes after midnight from S01 and gains 60 minutes from the duplicate S04. C3 counts all 90 minutes of S09, though only 30 fell on 10 September. C2's figure ignores its eight hours of downtime, so it looks like the weakest connector at 10.4% when, on the time it could actually serve, it managed 15.6%.

Reading the three connectors

C1 and C3 have similar time utilisation, yet C3, at half the rated power, delivers nearly twice the energy utilisation. C1's problem is dwell: shoppers plug in, finish charging within the hour and leave the car for the rest of the visit. That argues for an idle fee or a parking-time rule at the mall, not another fast charger. C3's highway customers charge and leave, which is what a busy, low-idle connector looks like.

Other traps in charging data

• Rated kW is a ceiling, not an expectation. Many vehicles accept less than the connector's rating, and charging tapers as the battery fills. Energy utilisation will never approach 100% on a real site. Compare sites with each other and with their own history, not with the rating.
• Zero-energy sessions. Failed authorisation or a cable that never locked creates sessions with 0 kWh. Count them separately as a reliability measure; they should not add plugged minutes.
• Open sessions. A session with no plug_out, from a lost network message, can run for days. Cap or exclude it, and log how many.
• Shared power cabinets. Where two connectors share one power module, their combined energy cannot exceed the cabinet's rating. Measure energy utilisation at the cabinet level too.
• Downtime definition. Decide whether a charger that is online but faulted on one connector counts as down, and write the rule down.

For the operations or network-planning head

A two-by-two works well: time utilisation on one axis, energy utilisation on the other, one dot per connector, sized by sessions. High-time, low-energy connectors are an idle-management conversation; high-energy connectors are expansion candidates; low on both is a location or visibility question. Put each site's downtime minutes beside the chart, because a reliability problem looks exactly like a demand problem on a utilisation dashboard.

Where to learn it

Interval clipping, deduplication and denominators that exclude unavailable time are SQL skills the Data Analyst programme covers across 3–4 months, live online with optional in-person labs at our Noida centre. For another case where the physics and the reporting unit must agree, see solar plant performance ratio.