CAISO Day-Ahead Load Forecast

CAISO's day-ahead demand forecast, one row per forecast hour with a system total and one column per TAC area.

Load data CAISO market

Query with the agent

Native record

one row per forecast hour, wide by TAC area

hourly (day-ahead run) publication cadence

Evidence

Forecast data

Forward-looking values with an explicit issue time

Time basis

Interval-start records

UTC · event_time

Overview

75,727 hours, 2018-01 through the next delivery day. Column-for-column identical to CAISO Load (caiso_load) on purpose: system_mw, pge_mw, sce_mw, sdge_mw, vea_mw, mwd_mw. Forecast-versus-actual is therefore a plain join on event_time with no reshaping on either side. CAISO publishes no issue-time column for this feed, so there is no vintage axis here — this is the day-ahead market run for each hour, not a curve of successive forecasts. CAISO's other horizons (two-day, seven-day, real-time) live in caiso.load_forecast_horizons and only begin in May 2026.

Query contract

event_time is the hour being forecast. There is exactly one day-ahead forecast row per hour and no publication column, so no vintage selection is possible or needed.

Analysis rules

Take every value for one hour from that hour's single row.
Do not sum the five TAC-area columns together with system_mw.

Related analysis

day-ahead forecast error · join on event_time; both tables carry the same six columns

Interpretation

This is the DAY-AHEAD run only. It cannot answer what CAISO expected an hour before delivery, and calling it 'the forecast' without the horizon overstates it.
There is no issue-time column, so this table cannot support any question about how a forecast was revised as delivery approached.

Access

Query from your own tools

Use the guarded HTTP endpoint directly, or connect an MCP client and load the dataset contract with describe_data before calling query_data.

CAISO MCP API reference

HTTP endpoint

POST /api/v1/datasets/caiso_load_forecast/query

Need the full history of this dataset in Snowflake, Databricks or your own storage? Data services delivers it on a schedule.

Schema7

Column

Type

event_timeDateTime64(6, 'UTC')

system_mwNullable(Float64)

pge_mwNullable(Float64)

sce_mwNullable(Float64)

sdge_mwNullable(Float64)

vea_mwNullable(Float64)

mwd_mwNullable(Float64)

Sample data

event_time	system_mw	pge_mw	sce_mw	sdge_mw	vea_mw	mwd_mw
2026-10-13 06:00:00.000000	22683.36	10001	10296.47	2154.26	52.86	178.77
2026-10-13 05:00:00.000000	24063.65	10655.65	10880.86	2293.26	55.89	178
2026-10-13 04:00:00.000000	25306.38	11282.79	11325.56	2460.68	59.79	177.55
2026-10-13 03:00:00.000000	25960.06	11700.66	11450.97	2567.34	62.89	178.2
2026-10-13 02:00:00.000000	26635.04	12056.05	11655.98	2679.95	65.75	177.31
2026-10-13 01:00:00.000000	26454.67	12082.44	11487.49	2642.02	65.23	177.5
2026-10-13 00:00:00.000000	25232.85	11391.25	11167.8	2433.91	62.71	177.18
2026-10-12 23:00:00.000000	23668.14	10390.45	10841.97	2198.06	61.07	176.59
2026-10-12 22:00:00.000000	22563.67	9639.59	10641.05	2047.7	57.48	177.86
2026-10-12 21:00:00.000000	21794.96	9014.71	10605.42	1943.73	53.13	177.97
2026-10-12 20:00:00.000000	21477.97	8705.69	10642.45	1901.99	49.58	178.27
2026-10-12 19:00:00.000000	21534.6	8641.49	10770.71	1896.13	47.76	178.5
2026-10-12 18:00:00.000000	22012.59	8918.65	10921.09	1947.23	47.29	178.33
2026-10-12 17:00:00.000000	23037.68	9528.61	11173.42	2110.17	48.93	176.55
2026-10-12 16:00:00.000000	24136.98	10284.83	11367.56	2256.92	50.82	176.86
2026-10-12 15:00:00.000000	24882.18	11069.56	11264.23	2317.71	53.22	177.46
2026-10-12 14:00:00.000000	24795.82	11307.5	10918.75	2339.7	52.55	177.31
2026-10-12 13:00:00.000000	23308.69	10473.53	10362.6	2242.44	52.21	177.91
2026-10-12 12:00:00.000000	21335.41	9518.58	9506.43	2084.35	48.37	177.69
2026-10-12 11:00:00.000000	20169.49	8953.78	8994.02	1997.83	45.24	178.62
2026-10-12 10:00:00.000000	19867.51	8785.1	8842.1	2018	44	178.32
2026-10-12 09:00:00.000000	20204.94	8883.19	8997.49	2101.12	44.13	179.01
2026-10-12 08:00:00.000000	20894.39	9103.31	9352.18	2213.72	44.96	180.22
2026-10-12 07:00:00.000000	21716.18	9393.91	9797.49	2297.2	46.54	181.03
2026-10-12 06:00:00.000000	22467.33	9675.95	10277.52	2292.43	44.61	176.82

Coverage

100.0% complete0 gap days3,208 days with data

Show coverage history

Coverage: 100.0%Days with data: 3,208Gaps: 0

2026

2025

2024

2023

2022

2021

2020

2019

2018

gaplessmore

Related load data

Related load data
Dataset	Description
CAISO Hourly Actual LoadCAISO · hourly	Actual hourly electricity demand in CAISO, one row per hour with a system total and one column per TAC area. 75,689 hours, 2018-01 to now. Columns: system_mw plus pge_mw, sce_mw, sdge_mw, vea_mw, mwd_mw. system_mw is CAISO's own CA ISO-TAC total. The five TAC areas are the CAISO balancing area's own partition. Wide rather than long, so "load right now" and "the 5 TAC area shares of it" are one row, not 6. Every market's load table in this catalog shares the system_mw column and the same hourly grain, so a cross-market demand comparison needs no reshaping.
Five-Minute Demand OutlookCAISO · 5-minute	CAISO system demand, day-ahead and hour-ahead demand forecasts, and demand response from the live demand Outlook. Every capture is retained, with Pacific snapshot date and DST ambiguity flags.
Load ActualsCAISO · hourly	Hourly actual electricity demand from CAISO OASIS, reported for the CAISO system total and TAC load areas such as PG&E, SCE, and SDG&E. This settlements-grade series is published after the operating interval rather than as a live five-minute reading. It supports forecast benchmarking, historical load shape, and aligned hourly net-load analysis. The official source is CAISO OASIS.
Load Forecast HorizonsCAISO · 5-minute forecast	CAISO load forecasts for the system and TAC load areas. The RTM five-minute label is the rolling VSTLF forecast used for each real-time dispatch operating interval, not a day-ahead peak forecast. event_time is the delivery interval and source_time is the publication time. For an honest historical forecast comparison, select the last source_time at or before event_time; for a forward interval select the latest source_time available at the analysis cutoff. The public source is CAISO OASIS.