Minimum-Sufficient Knowledge and Computation for Wind Energy Predictive Analysis

Purbak Sengupta; Aparna S. varde; Emil Jørgensen Njor

cs.LGNov 5, 2026

Minimum-Sufficient Knowledge and Computation for Wind Energy Predictive Analysis

Authors: Purbak Sengupta, Aparna S. varde, Emil Jørgensen Njor

Organizations: SDU Center for Software Technology University of Southern Denmark (SDU) Vejle, Denmark

Abstract

Resource-constrained cyber-physical decision systems often apply fixed sensing, knowledge, and computational pipelines despite state-dependent information needs. We introduce Minimum-Sufficient Knowledge and Computation (MSKC), a resource-aware computing principle: select the least-costly route meeting decision-risk and deployment constraints, escalating only when cheaper routes are insufficient. As proof of concept, we isolate adaptive knowledge routing, the decision underlying Knowledge-Guided Machine Learning (KGML), which conventionally injects external knowledge unconditionally rather than only where needed. We instantiate this for wind ramp-event predictive analysis, an edge-oriented decision-support task, evaluating local, relational, and physics-informed routes with a state-aware gate on publicly available real data from two wind farms. We establish conditions for feasibility preservation, cost dominance, and strict expected savings over always-on routing. At a 35.9% relational-knowledge injection rate, adaptive routing is statistically indistinguishable from always-on injection on the reported metrics while reducing knowledge calls by 64%, supporting sustainable, Green-IT-aligned deployment. These results support MSKC's premise that uniformly maximal intelligence is unnecessary. This offers intelligent data analysis for big data streams: the analysis itself decides, per state, how much knowledge and computation the data warrant.

Figures & tables

Fig. 1: Overview of the MSKC selection process: candidate routes are filtered by certified sufficiency, resource, and admissibility checks into the feasible set $\widehat{\mathcal{A}}(s)$ , from which the minimum-carbon route $r^{*}(s)$ is selected, with fallback to $r_{H}$ or deferral as described in Section III . The highlighted region marks the two-route knowledge-acquisition instance evaluated empirically in this paper.

Fig. 2: The minimum-sufficient route selection procedure. Starting from the cheapest candidate route, the system checks accuracy, resource fit, and safety together; a failed check escalates to the next costlier route, while a route that passes all three is used immediately. If no route in the menu ever passes, the system defers the decision rather than acting on an unsupported one.

Fig. 3: This figure explains why the selected route can never cost more than the fallback $r_{H}$ . Ordinary routes ( $r_{0},r_{1},r_{2},\ldots$ ) must pass all three checks, accurate enough, fits resources, safe, to reach the feasible set; some fail and drop out. $r_{H}$ is assumed to always pass and reaches the feasible set by a guaranteed path regardless. Because the final selection $r^{*}$ is always the cheapest option inside this same feasible set, and $r_{H}$ is always one of its members, $r^{*}$ can never be more expensive than $r_{H}$ .

	Kelmarsh	Penmanshiel
Turbines	6	5 (of 14)
Rated power (kW)	2050	2050
Rotor diameter (m)	92.0	82.0
Hub height (m)	68.5–78.5	59.0
Period	2024 (1 year)
Resolution	10 min

TABLE I: Wind farm specifications and dataset summary after cleaning. Ranges are taken across each site’s turbines.

Region	Share of labeled rows
Severe down-ramp	1.7%
Moderate down-ramp	10.7%
Stable	75.1%
Moderate up-ramp	10.8%
Severe up-ramp	1.7%

TABLE II: Ordinal ramp-region label distribution, pooled across both sites.

Fig. 4: Validation of the ordinal ramp-region labeling mechanism: power trace shaded by assigned region (top), the underlying change signal against its thresholds (middle), and the full-year, all-turbine distribution of that signal (bottom).

Fig. 5: Macro-F1 as a function of injection rate under the swept gate threshold.

Fig. 6: Spatial layout and neighbor availability at both sites; a four-day trace of true region, predicted region, and injection decision; the deployed gate’s confusion matrix; within-site versus cross-site severe recall; and the learned KUS gating-strategy comparison at matched injection rate.

Population	Route	Macro F1	Sev. recall	Comm. cost
Neighbor-avail.	Base ( $r_{0}$ )	0.280	0.701	none
Neighbor-avail.	Relational ( $r_{1}$ )	0.273	0.751	required
Full test	Base ( $r_{0}$ )	0.263	0.694	none
Full test	Physics ( $r_{2}$ )	0.266	0.705	none

TABLE III: Knowledge route scores, reported separately for each route’s own paired evaluation population.

Comparison	Macro F1	Sev. recall	False-safe
$r_{1}-r_{0}$	$-0.0068$	$+0.0511$	$-0.0017$
	[ $-0.011$ , $-0.003$ ]	[ $+0.034$ , $+0.068$ ]	[ $-0.008$ , $+0.005$ ]
$r_{2}-r_{0}$	$+0.0036$	$+0.0115$	$-0.0009$
	[ $+0.002$ , $+0.006$ ]	[ $+0.003$ , $+0.020$ ]	[ $-0.004$ , $+0.003$ ]

TABLE IV: Paired bootstrap comparisons, 95% confidence intervals.

Strategy	Macro F1	Sev. recall	False-safe
Random (mean)	0.277	0.719	0.032
Heuristic	0.273	0.752	0.030
Learned gate	0.273	0.750	0.031

TABLE V: Gating strategies at matched injection rate.

Trees	Latency (ms)	Macro F1	Sev. recall
300	12.0	0.327	0.53
50	2.2	0.299	0.65
15	$<$ 2	0.280	0.70

TABLE VI: Model weight versus latency and decision quality (base route).

Knowledge-Guided ML

See all 8 papers

XWind: A Cross-site Router for Large Language Model Inference Serving at Renewable Energy Farms
May 22, 2026Tella Rajashekhar Reddy, Atharva Deshmukh, Liangcheng Yu +7Energy-Efficient ML LLM Inference Serving
Microsoft
Energy-Aware Wind-Resilient Routing for Truck-Assisted Multi-UAV Delivery under Wind Uncertainty
Aug 12, 2026Tianshun Li, Yanggang Sheng, Hongliang Lu +3UAV Navigation Vehicle Routing
The Hong Kong University of Science and Technology (Guangzhou), China · Southern University of Science and Technology, China · Intelligent Transportation Thrust, Systems Hub, The Hong Kong University of Science and Technology (Guangzhou), China
Flexible Routing via Uncertainty Decomposition
May 8, 2026Charlotte Peale, Siddartha Devic, Parikshit Gopalan +2Uncertainty Quantification Adaptive Model Routing
1Stanford University · University of Southern California · 3Apple

May 22, 2026J/K move · Enter open · S save

XWind: A Cross-site Router for Large Language Model Inference Serving at Renewable Energy Farms

Tella Rajashekhar Reddy, Atharva Deshmukh, Liangcheng Yu, Chaojie Zhang+6

Microsoft

Energy-Efficient ML LLM Inference Serving LLM Routing Edge-Cloud Computing

Open paper AskPDF arXiv

AI power demand is growing at an unprecedented rate while power grids are often ailing and struggle to keep up. Grid expansion comes with high capital expenditure and long-distance transmission losses, yet there is abundant renewable energy at the source, just not matched to demand. This paper proposes a complementary AI infrastructure deployment model, AI Greeninferencing, that brings modular AI compute to renewable energy sources, focusing on wind, allowing AI footprint expansion, generating local behind-the-meter demand for renewable sites, and helping ease the growing strain on power utilities. Our feasibility analysis shows that 890+ GW of wind capacity lies within 50 ms network round trip time of Azure data centers, and that site-wise right-sizing combined with spatial complementarity of wind energy keeps aggregate fleet utilization on par with traditional deployments. To serve inference requests under variable wind power, we build XWind, a lightweight, reactive, and workload-agnostic AI inference router that uses only real-time signals: inference latency, KV-cache utilization, and queue depth, to dynamically configure sites and distribute requests. Evaluated on a real 64-GPU A100 testbed emulating three wind-powered sites with Azure production traces, XWind reduces P99 end-to-end latency by up to 52% over the strongest contender (also our idea) and by up to 98% over baselines such as power-capping and GPU idling, with consistent gains across workload types, load levels, and GPU generations.

Abstract

Knowledge-Guided ML

Explore similar work

XWind: A Cross-site Router for Large Language Model Inference Serving at Renewable Energy Farms

Energy-Aware Wind-Resilient Routing for Truck-Assisted Multi-UAV Delivery under Wind Uncertainty

Flexible Routing via Uncertainty Decomposition