Caribbean AI Association
Guh bak a di akaiv
Trinidad an Tobago / Workin piepa

How fi spot unusual monthly gas use inna Trinidad an Tobago wid di closest-match method

A test weh use four gas-consuming sector an no incident label

Lied aata · Adrian Dunkley   |   2026-11-20   |   v1.0

Ina plien Patwa

Two month can use nearly di same amount a gas overall even if di factory dem an power station dem get different share. Wi look at four use together fi find which 2023 month come closest to each 2024 month.

June stand out whichever single sector wi leave out. April come closer to di cutoff. Dat help a analyst choose weh fi start checking, an see why di national total alone give a different list.

Distance from each 2024 month to di closest 2023 month

Standardised distance
Jan0.581
Feb0.775
Mar0.968
Apr1.758
May1.362
Jun4.719
Jul3.824
Aug1.211
Sep1.415
Oct1.391
Nov1.424
Dec1.208
Di higher di score, di less familiar di mix across sector. Di four-sector cutoff was 1.703. Di score nuh have a unit. Wi calculate it from [1, 2].

Step dem ina di analysis

  1. Scale di sector rate dem
  2. Find di closest month
  3. Compare wid di cutoff
  4. Check di source record dem
Riid di whole piepa

Wi compare monthly gas use inna 2024 wid twelve reference month from 2023. Di closest-match method flag April, June an July. June pass di reference cutoff even when wi leave out each sector one at a time. A test weh use only di national total flag August an October too, but miss April. Di result dem point to record weh need checking. Di bulletin dem nuh label actual incident, so wi cyaan measure how accurate di detection was [1, 2].

Weh di risorch a aks?

Di four-sector test pick di same month fi check as di test weh use only total gas use fi di whole country?

Data an metad

From Table 3B inna di revised energy bulletin dem fi 2023 an 2024, wi tek gas use fi electricity, ammonia, methanol an liquefied natural gas (LNG), plus di national total. Each reading give di average daily rate fi one month, inna million standard cubic feet per day [1, 2].

Fi each pair a month, wi divide di difference fi each sector by dat sector own 2023 sample standard deviation. Wi square di answer dem, add dem up an tek di square root. Each 2024 month get a score equal to di smallest distance from it to any 2023 month. Dat a di one-nearest-neighbour way fi score how unusual a reading is [3]. Fi set di cutoff, wi find each 2023 month closest odda month inna 2023. Wi tek di largest a dem twelve distance. Only a score weh higher than di cutoff trigger a check.

Rizolt dem

Di four-sector cutoff come to 1.703. April score 1.758, June 4.719 an July 3.824. Di closest match fi June a August 2023. Inna June 2024 gas use fi ammonia was 294 million standard cubic feet per day, an fi methanol it was 409. Both a dem lower than dem 2023 monthly minimum, 359 an 483 inna di same order [1, 2].

When wi run di same calculation wid only di national total, it flag June, July, August an October. When wi leave out LNG an run di test pon di odda three sector dem, only June pass di cutoff. June clear di new cutoff inna all four test weh leave out one sector.

Wa di rizolt mean

Somebody checking April woulda see a mix across sector weh nuh common, even though di national total a 2,364 fall inside di 2023 range a 2,331–2,665 [1, 2]. Di score tell yuh which record fi compare. Yuh woulda need plant log an reporting note fi explain why two month differ.

Limit a di study

One reference year cyaan cover every season an industrial condition weh coulda come up. Di revised bulletin dem look back pon what done happen. Di method nuh have information bout planned shutdown, reporting mistake or confirmed interruption. A flag cyaan tell yuh di chance a one outage or how much it woulda cost fi miss a incident.

Run di analysis agen

Read di 24 row dem inna date order. All five volume column show di average daily rate fi dat month inna million standard cubic feet per day.

Use di four sector column fi work out di 2023 sample standard deviation dem. Divide by 11 when yuh a calculate di variance. Keep dem same scale fi di reference distance an di test distance dem.

Fi each reference month, tek out di month itself before yuh look fi di closest match. Use di largest a di twelve distance fi di cutoff. Compare each 2024 month wid all twelve reference month.

Do it again wid only di national total. Fi check how much each sector matter, leave out one at a time an work out di cutoff an di 2024 score dem again.

Data tiebl
MonthElectricityAmmoniaMethanolLNGNational total
2023-0124746256912642665
2023-0223945456912652638
2023-0325142748812732537
2023-0425340948311902419
2023-0526441951712302529
2023-0627146453411182501
2023-0729341854810942461
2023-0827037752310582331
2023-0929043855911782527
2023-1026137457610822342
2023-1126035959211802488
2023-1224940859010742415
2024-0125641449611582431
2024-0225448055812842686
2024-0326545053611822540
2024-042664485579942364
2024-0529044350610872430
2024-062652944098481905
2024-072844455308002161
2024-0827447857211552586
2024-0927548054410152388
2024-1027244948412612566
2024-1127545248712202536
2024-1224743851211842474

Suos dem

  1. Ministry of Energy and Energy Industries, Consolidated Monthly Bulletins, January–December 2023, Table 3B, amended 28 November 2025
  2. Ministry of Energy and Energy Industries, Consolidated Monthly Bulletins, January–December 2024, Table 3B, amended 28 November 2025
  3. S. Ramaswamy, R. Rastogi and K. Shim (2000), Efficient algorithms for mining outliers from large data sets

Weh di data come fram · Ministry of Energy and Energy Industries, Table 3B, 2023 and 2024 bulletins, amended 28 November 2025. CAIA calculations.

Dounload di data an di analysis code (ZIP)

Workin piepa · In-house review · How wi check di risorch