ElecLoad: Adaptive Electricity Load Forecasting Guided by Multiple Events

Huajian Zhang; Mingyue Cheng; Xingpeng Gao; Xiaoyu Tao; Zhiding Liu; Jiqian Yang; Zhi Li

cs.AINov 4, 2026

ElecLoad: Adaptive Electricity Load Forecasting Guided by Multiple Events

Authors: Huajian Zhang, Mingyue Cheng, Xingpeng Gao, Xiaoyu Tao, Zhiding Liu, Jiqian Yang, Zhi Li

Organizations: State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China, Hefei, China · Shenzhen International Graduate School, Tsinghua University, Shenzhen, China

Abstract

Accurate electricity load forecasting provides critical decision support for energy procurement, generation scheduling, and supply-demand balancing, contributing to the economic efficiency and operational reliability of power systems. However, future electricity demand can be affected by the complex interplay of multiple events described in Web notices, weather alerts, and event schedules. Their effects vary over time and can reinforce or offset one another, making electricity load difficult to forecast accurately. While existing methods can incorporate context as additional input, how to use the effects of multiple events and their interactions to guide the load forecasting process itself remains to be explored. To address this gap, we propose ElecLoad, a context-aware agent for multi-event electricity load forecasting that adaptively organizes the forecasting process. Specifically, we train ElecLoad to combine historical load with context to analyze each event's relevance to the target load, the timing of its effects, and potential interactions among events. Guided by this analysis, ElecLoad adaptively chooses to forecast the entire forecast horizon directly or to partition it into segments according to changes in event effects. When the horizon is partitioned, sub-agents independently generate initial forecasts for each segment. The main agent then reviews these forecasts, and the sub-agents perform selective refinement guided by its feedback. Experiments on multiple electricity load datasets show that ElecLoad achieves state-of-the-art performance in most evaluation scenarios. Code is available at https://github.com/zhjai/ElecLoad.

Figures & tables

Figure 1. Conceptual illustration of (a) load changes under multiple events over time and (b) event-guided forecast partitioning in ElecLoad.

Figure 2. ElecLoad uses multi-event analysis to adaptively choose direct or segmented forecasting. In the segmented path, the main agent selects segments for feedback and refinement. The main agent and sub-agents share the same LLM weights. Historical load and context enter ElecLoad. Event analysis guides the choice between direct forecasting of the full horizon and partitioning into segments. On the segmented path, sub-agents generate forecasts. The main agent provides feedback on selected segments; their sub-agents refine the forecasts and return them for reassessment. After stopping, segment forecasts are assembled in time order, with optional adjustments near segment boundaries. Both paths produce a final load forecast.

Source	$L/H$	Train	Val.	ID test	OOD test	Total
GEFCom2014	96/96	640	100	200	0	940
Nord	168/24	288	48	96	0	432
Nord	96/96	72	12	24	0	108
PJM	168/24	320	56	112	0	488
PJM	96/96	80	14	28	0	122
SPP	168/24	160	24	48	0	232

Table 1. Dataset sizes by forecasting window.

L/H

: history/forecast hours. ERCOT includes 71 system-level cases for the main evaluation and 14 regional sensitivity cases.

	Multi-day forecasting ( $L=96,\ H=96$ )										Day-ahead forecasting ( $L=168,\ H=24$ )
Method	GEFCom2014		Nord		PJM		SPP		AEMO †		Nord		PJM		SPP		LCL		iFlex		ERCOT †
	WAPE	NRMSE	WAPE	NRMSE	WAPE	NRMSE	WAPE	NRMSE	WAPE	NRMSE	WAPE	NRMSE	WAPE	NRMSE	WAPE	NRMSE	WAPE	NRMSE	WAPE	NRMSE	WAPE	NRMSE
ARIMAX	17.63	20.91	7.49	8.87	10.09	11.62	6.09	7.08	29.21	33.83	5.48	6.58	7.01	8.07	6.25	7.11	808.45	964.88	30.19	37.34	7.44	8.86
LightGBM	9.03	10.90	5.93	6.90	7.07	8.15	4.50	5.51	15.57	20.06	4.84	5.56	4.89	5.75	4.76	5.64	1008.97	1167.11	34.94	41.98	7.55	8.51
PatchTST	13.82	16.91	7.97	9.35	7.52	8.96	5.07	6.23	18.45	22.67	5.22	6.14	5.08	6.28	4.33	5.46	1873.00	2151.25	31.19	39.03	7.00	8.56
TimeXer	13.40	16.45	8.18	9.58	7.84	9.21	4.92	5.96	15.85	20.11	4.96	5.69	4.71	5.74	4.16	5.01	958.83	1120.30	30.49	38.58	7.18	8.47

Table 2. Overall performance in multi-day and day-ahead load forecasting. WAPE and NRMSE are reported as percentages. Best results are bold; second-best results are underlined.

L/H

: history length/forecast horizon in hours.

Figure 3. Forecasting on GEFCom2014 synthetic event cases. (a) WAPE for four event conditions; $0$ retains the original target without added events. (b) IEE grouped by zero/nonzero target interaction. (c,d) WAPE and IEE across generalization settings.

Group	Variant	GEFCom2014		PJM		LCL
Group	Variant	WAPE	NRMSE	WAPE	NRMSE	WAPE	NRMSE
Reference	ElecLoad (Full)	6.68	8.42	3.13	3.86	41.94	77.24
Inputs	w/o event text	8.46	10.63	3.51	4.37	46.27	83.16
Inputs	w/o covariates	7.39	9.36	3.79	4.71	44.12	80.58
Planning	w/ direct only	7.82	9.91	3.48	4.32	42.91	78.83
	w/ segmented only	6.97	8.86	3.39	4.21	45.36	82.07

Table 3. Ablation study (

L=96

H=96

for GEFCom2014;

L=168

H=24

for PJM and LCL).

Figure 4. WAPE versus normalized token cost on GEFCom2014: (a) routing; (b) refinement and reconciliation. Two square scatter plots show WAPE (%) vertically against normalized token cost horizontally for GEFCom2014 policies. Adaptive routing compares direct-only, segmented-only, and ElecLoad (Full). Segmented only compares segmented-only, no refinement, random refinement, no reconciliation, and forced reconciliation on the same cases under forced segmented forecasting. Points are unconnected, and segmented-only has identical coordinates in both panels.

Figure 5. ElecLoad case: (a) T1 planning and T2 initial forecasts; (b) T3-guided refinement and T4 boundary reconciliation. Two panels show a 96-step electricity load forecast, with forecast step on the horizontal axes and load in arbitrary units on the vertical axes. Panel a shows a heatwave from steps 24 to 65 and demand response from steps 36 to 60 above three event-guided segments, S1, S2, and S3. The lower plot compares Ground truth with Initial prediction, which is too high within S2. Panel b compares Ground truth, Initial, After T3, and Final. Purple shading identifies S2 for refinement, and green bands with T4 arrows identify reconciliation near its boundaries. The summary reports WAPE values of 4.17\%, 2.07\%, and 1.91\% and cumulative costs of 19.5k, 25.9k, and 26.2k tokens for Initial, After T3, and Final, respectively. — Figure 5. ElecLoad case: (a) T1 planning and T2 initial forecasts; (b) T3-guided refinement and T4 boundary reconciliation. Two panels show a 96-step electricity load forecast, with forecast step on the horizontal axes and load in arbitrary units on the vertical axes. Panel a shows a heatwave from steps 24 to 65 and demand response from steps 36 to 60 above three event-guided segments, S1, S2, and S3. The lower plot compares Ground truth with Initial prediction, which is too high within S2. Panel b compares Ground truth, Initial, After T3, and Final. Purple shading identifies S2 for refinement, and green bands with T4 arrows identify reconciliation near its boundaries. The summary reports WAPE values of 4.17%, 2.07%, and 1.91% and cumulative costs of 19.5k, 25.9k, and 26.2k tokens for Initial, After T3, and Final, respectively.

Group	Variant	GEFCom2014			PJM			LCL
Group	Variant	$E$	$C$	$G$	$E$	$C$	$G$	$E$	$C$	$G$
Routing	w/ direct only	7.82	7.8	—	3.48	7.0	—	42.91	6.5	—
	w/ segmented only	6.97	24.0	0.052	3.39	22.0	0.006	45.36	20.0	-0.181
	ElecLoad (Full)	6.68	18.7	0.105	3.13	15.3	0.042	41.94	12.7	0.156
Refinement	w/o refinement	7.24	15.0	—	3.37	12.2	—	44.63	10.0	—
	w/ random refinement	6.94	18.6	0.083	3.32	15.2	0.017	43.06	12.8	0.561

Table 5. Ablations of routing, refinement, and reconciliation.

Appendix figures & tables11 assets

Supplementary material from the paper’s appendix.

Appendix

Source	Split	Date range
GEFCom2014	Train	2005-01-05–2009-10-31
	Val.	2009-11-08–2010-11-11
	Test	2010-11-19–2011-11-26
Nord	Train	2015-06-30–2022-12-30
	Val.	2023-01-01–2023-12-05
	Test	2024-01-01–2024-11-19

Appendix

Table 6. Calendar date ranges by data source and split.

Split	$N$	0	H	I	1	$\geq$ 2	Overlap
Train	640	105	15	80	250	190	130 (68.4)
Val.	100	15	2	20	36	27	20 (74.1)
ID test	200	30	4	40	72	54	40 (74.1)

Appendix

Table 7. GEFCom2014 case composition (

L=96

H=96

			Event descriptions
Source	$L/H$	$N$	0	1	2	$\geq$ 3	Overlap	Incomp.
ID test
GEFCom2014 original	96/96	30	30	0	0	0	0	0
GEFCom2014 synthetic	96/96	170	0	102	58	10	40 (23.5%)	0
Nord	168/24	96	0	54	30	12	30 (31.2%)	16
Nord	96/96	24	0	15	5	4	3 (12.5%)	11

Appendix

Table 8. Test-set event descriptions and temporal overlap.

Dataset	Event Condition	$N$	Full	Direct	No text	$\Delta$
GEFCom2014	Single Event	72	6.42	7.90	8.20	-1.48
GEFCom2014	Multiple Events	54	7.61	9.50	10.80	-1.89
PJM	No Event	35	2.63	2.80	2.41	-0.17
	Single Event	45	3.12	3.42	3.57	-0.30
	Multiple Events	32	3.69	4.31	4.63	-0.62
SPP	Single Event	18	2.74	2.91	3.08	-0.17

Appendix

Table 9. WAPE by number of events (

L=96

H=96

for GEFCom2014;

L=168

H=24

for PJM and SPP).

Setting	Value
Agent RL batch; epochs; learning rate	$16;\ 2;\ 10^{-6}$
Train/validation; PPO clip; KL	$1800/300;\ 0.2;\ 0.001$
Prompt/response/context tokens	$32768/4096/36864$
Train/validation temperature	$1/0$
Max segments; minimum steps	$4;\ 6$
Refinements; repairs/call	$2;\ 1$

Appendix

Table 10. Runtime settings.

Setting	GEFCom2014	PJM	LCL
ElecLoad 0.8B	8.14	3.37	47.36
ElecLoad 4B	6.68	3.13	41.94
ElecLoad 9B	6.63	3.16	41.81
w/o Reasoning init.	7.42	3.46	45.08
w/o Agent training	8.05	3.83	47.18
w/o Agent SFT	7.19	3.42	44.28

Appendix

Table 11. Model scale and training.

Group	Removed component	GEFCom2014	PJM	LCL	kTok
Reasoning	Numerical term	8.13	3.83	48.01	–
	Improvement bonus	7.52	3.53	44.97	–
	Scale penalty	7.62	3.66	47.27	–
	Extreme penalty	7.56	3.52	46.87	–
	Format penalty	7.45	3.52	45.03	–
Agent	Final accuracy credit	7.48	3.58	45.72	17.7

Appendix

Table 12. Training-component ablations.

Setting	GEFCom2014	PJM	LCL	kTok
Default	6.68	3.13	41.94	18.7
At most 2 segments	6.95	3.26	42.05	15.9
At most 3 segments	6.73	3.14	41.98	17.5
Force 4 segments	7.06	3.33	44.93	26.1
Refine first segment	6.98	3.35	43.46	18.5
Force 1 refinement	6.85	3.21	42.73	17.3

Appendix

Table 13. Execution variants.

Figure 6. Mean WAPE change after adding irrelevant text on GEFCom2014 (pp: percentage points).

	No events		Relevant events
Type	E	D	E	D
Administrative	+0.012	-0.023	+0.045	-0.015
Wrong target	+0.032	-0.013	+0.061	-0.011
Outside horizon	-0.005	+0.027	-0.005	-0.007
Cancelled	+0.028	+0.057	+0.045	-0.086
Superseded	-0.032	-0.033	-0.043	-0.043

Appendix

Table 14. Irrelevant-text effects (

L=96

H=96

Source	$\Delta$ WAPE	95% interval	kTok	P99
GEFCom2014	-1.14	[-1.31, -0.98]	18.7	16.18
PJM	-0.35	[-0.58, -0.12]	15.3	8.50
LCL	-0.97	[-1.95, +0.07]	12.7	79.88

Appendix

Table 15. Paired WAPE differences.

Electricity Load Forecasting

Electricity load forecasting predicts future power demand to support grid operation, planning, and energy management.

See all 50 papers

From Shared Demand Patterns to Local Uncertainty: Probabilistic Load Forecasting by Mixing Compact Adaptations
Oct 6, 2026Haoran Li, Zhe Cheng, Yang WengElectricity Load Forecasting Probabilistic Time Series Forecasting
PeakFocus: Bridging Peak Localization and Intensity Regression via a Unified Multi-Scale Framework for Electricity Load Forecasting
May 20, 2026Wangzhi Yu, Peng Zhu, Qing Zhao +2Time Series Forecasting Electricity Load Forecasting
School of Computer Science and Technology, Tongji University, Shanghai, China · Big Data Center, State Grid Corporation of China, Beijing, China
Behaviour-Conditioned Neural Processes for Adaptive Residential Short-Term Load Forecasting
Jul 17, 2026Ramin Soleimani, Andrea Visentin, Dirk PeschNeural Processes Time Series Forecasting
School of Computer Science and Information Technology, University College Cork, Cork, Ireland

Oct 6, 2026J/K move · Enter open · S save

From Shared Demand Patterns to Local Uncertainty: Probabilistic Load Forecasting by Mixing Compact Adaptations

Haoran Li, Zhe Cheng, Yang Weng

Electricity Load Forecasting Probabilistic Time Series Forecasting Parameter-Efficient Fine-Tuning

Open paper AskPDF arXiv

Probabilistic load forecasting has been widely studied for power-system operation and planning, but customer- and transformer-level forecasting introduces a distinct scalability challenge. At these levels, load uncertainty is strongly affected by customer behavior, weather, and mixed load composition, making it difficult for a single shared model to capture heterogeneous patterns. Using separate probabilistic models can improve local accuracy, but becomes costly to train, store, update, and validate at scale. To address this challenge, we develop a scalable customer-aware forecasting framework that learns common demand behavior through a shared model while adapting only a compact subset of parameters. Rather than using an independent model for each load or assigning each load to a specialized model, the proposed design learns a small bank of low-dimensional adaptation components and allows each load to combine them according to its forecasting characteristics. This preserves shared knowledge across customers while providing sufficient flexibility for heterogeneous and mixed load compositions. Experiments on 590 load profiles from the SMART-DS dataset show consistent improvements in deterministic accuracy and probabilistic quality over statistical, neural-network, Transformer-based, and pretrained time-series baselines, while retaining low storage and inference costs.

Abstract

Electricity Load Forecasting

Explore similar work

From Shared Demand Patterns to Local Uncertainty: Probabilistic Load Forecasting by Mixing Compact Adaptations

PeakFocus: Bridging Peak Localization and Intensity Regression via a Unified Multi-Scale Framework for Electricity Load Forecasting

Behaviour-Conditioned Neural Processes for Adaptive Residential Short-Term Load Forecasting