In a groundbreaking study published in Nature Energy, researchers have unveiled a sophisticated computational framework that leverages historical national data to probabilistically forecast the global trajectories of onshore wind and solar photovoltaic (PV) power expansion. By analyzing deployment patterns across more than 200 countries, the team harnessed advanced statistical techniques and machine learning models to capture the complex diffusive behaviors of these renewable technologies and generate calibrated projections with unprecedented insight.
At the core of this research lies a novel approach to identifying the “take-off” phase in technology diffusion—a critical juncture marking the transition from isolated, sporadic deployment to widespread and sustained growth. Utilizing a Bayesian change-point detection algorithm, named FOBI, the team dissected temporal deployment data to isolate the earliest and most statistically robust inflection points where the growth trajectory sharply accelerates. This methodological innovation allows for differentiation between mere experimentation phases and meaningful, scalable adoption, enriching the precision of subsequent growth predictions.
To quantify how wind and solar technology adoption spreads geographically, the authors modeled the cumulative count of countries surpassing the take-off threshold over time, fitting logistic curves to these data. This enabled them to extract a metric termed diffusion duration, which measures the interval within which adoption expands from 10% to 90% of the potential maximum number of countries. This spatial dimension of diffusion, rarely captured with such rigor, offers a critical yardstick for benchmarking renewable energy technologies against diverse historical analogues like mobile telephony and nuclear power.
Beyond mere identification of take-off points, the study delves deeply into the dynamics of growth pulses and inflection points. By fitting logistic functions to deployment time series and introducing the concept of curve maturity—the ratio of observed deployment to the estimated upper bound—the research distinguishes accelerating growth phases from steady-state periods. Intriguingly, the analysis reveals the presence of multiple growth pulses, likely driven by policy cycles, technological improvements, or market conditions. Such granularity illuminates the nuanced patterns often obscured in aggregate diffusion narratives.
Central to the predictive power of this research is the development of PROLONG, a novel probabilistic forecasting tool that integrates national-level adoption signals with global diffusion outcomes. PROLONG’s machine learning core was trained on an extensive ensemble of over 13,000 simulated diffusion trajectories per technology, encompassing logistic and bilogistic growth shapes infused with stochastic variability. This diversity in training data equips PROLONG to accommodate real-world complexities such as policy-induced slowdowns and subsequent rebound growth, enabling it to generalize effectively beyond purely theoretical models.
PROLONG’s validation entailed rigorous hindcasting exercises employing both simulated data and actual historical observations from onshore wind, solar PV, combined cycle gas turbines, nuclear power, and mobile telecommunications. These retrospective tests demonstrated PROLONG’s superior accuracy and reduced forecasting bias compared to conventional methods—such as simple exponential extrapolations or aggregated national logistic fits—especially in early to mid phases of diffusion where standard models typically falter. Its capacity to capture complicated growth dynamics establishes a new benchmark in renewable energy forecasting.
A key strength of the PROLONG framework is its probabilistic nature, which moves beyond deterministic point forecasts to produce calibrated confidence intervals that reflect genuine uncertainties. Using a battery of metrics—including interval widths, coverage probabilities, and continuous ranked probability scores—the researchers verified that PROLONG’s uncertainty ranges appropriately expand over longer horizons while maintaining sharpness and reliability. This dimension is critical for policymakers and stakeholders who rely on robust risk assessments rather than single forecast trajectories that can misrepresent the variability inherent in complex systems.
Comparing these projections against established climate and energy scenarios, including those compiled by the IPCC and IEA, reveals compelling insights. PROLONG’s median and percentile-based forecasts align closely with a range of mitigation scenarios targeting strict climate goals, yet uniquely provide probabilistic envelopes that clarify the likelihood of more optimistic or pessimistic pathways. This capacity to assess the credibility of ambitious decarbonization efforts in probabilistic terms marks an important advance in aligning modeling tools with real-world policymaking needs.
Moreover, the research team extended their analyses to generate regionally differentiated acceleration scenarios consistent with the goal of limiting global warming to 1.5°C. By constructing early and late acceleration trajectories—distinguished by deployment ceilings and growth parameters—and distributing these across ten global regions through an optimized algorithm, the study offers actionable pathways grounded in empirical diffusion constraints. These scenarios account for technological, social, and economic heterogeneity in adoption capacity, delivering nuanced regional insights crucial for tailored climate policies.
This work also tackles meaningful practical considerations by incorporating constraints on maximum market share ceilings, annual growth rates, and acceleration limits, all derived from historical experience of large-scale technology deployment. Such empirical grounding ensures the scenario outputs remain realistic, eschewing overly optimistic or physically implausible growth assumptions that often undermine long-term energy modeling.
Highlighting the interplay between local national experiences and global diffusion patterns, the research underscores that early national adoption characteristics—such as growth rates and cumulative market shares—harbor considerable predictive information about future global trajectories. This insight challenges traditional top-down forecasting approaches and advocates for more granular, data-driven models that capture heterogeneity in adoption behavior.
Importantly, the study addresses the heteroscedastic and sometimes non-linear nature of renewable energy diffusion, acknowledging how political, economic, and technological shocks can induce complex pulses and plateaus. By accommodating bilogistic growth forms and training on mixed-mode ensembles, the model’s versatility extends to accommodate these irregularities without sacrificing statistical rigor or forecast fidelity.
The implications of this research extend far beyond the academic realm. As global leaders and energy planners grapple with accelerating the clean energy transition, tools like PROLONG offer critical foresight into the pace and scale of renewable integration, bridging the gap between historical evidence and future aspirations. This data-driven modeling suite empowers decision-makers to better evaluate risks, design adaptive policies, and monitor progress against decarbonization milestones.
With rapid climatic changes and escalating geopolitical dynamics influencing energy infrastructures worldwide, the probabilistic framing of renewable energy futures presents a timely innovation. It equips the global community with a clearer understanding of where current trajectories stand relative to needed transformations and where intensified efforts might yield the greatest impact.
Summarily, by blending statistical innovation, machine learning, and comprehensive historical datasets, this study sets a new standard for modeling energy transitions. Its findings herald a future in which renewable energy growth is not merely forecasted but understood probabilistically, enabling more resilient and responsive energy systems planning in the fight against climate change.
Subject of Research: Probabilistic forecasting of global onshore wind and solar photovoltaic power growth based on historical national deployment patterns.
Article Title: Probabilistic projections of global wind and solar power growth based on historical national experience.
Article References:
Jakhmola, A., Jewell, J., Vinichenko, V. et al. Probabilistic projections of global wind and solar power growth based on historical national experience. Nat Energy (2026). https://doi.org/10.1038/s41560-026-02021-w
Image Credits: AI Generated
DOI: https://doi.org/10.1038/s41560-026-02021-w
Tags: Bayesian change-point detection in technology adoptioncross-country analysis of clean energy growthdiffusion duration metrics in technology spreadFOBI algorithm in renewable energy analysisglobal deployment patterns of wind and solar powerglobal wind power expansion trendslogistic growth modeling for energy deploymentmachine learning in energy forecastingprobabilistic forecasting of renewable energy growthscaling phases in renewable technology adoptionsolar photovoltaic technology diffusionstatistical techniques for renewable energy projections
