Accelerating the energy transition towards photovoltaic and wind in China

Ambitions to achieve carbon neutrality are needed in all nations to limit global warming to below 2 °C in the Paris Agreement. Accelerating the penetration of renewables is a key pillar in climate mitigation. Global decarbonization is not progressing fast enough to meet Paris goals; the world is probably on track for 2.8 °C of warming by 2100 on current policies. COP27 called for US$4–6 trillion per year of investments to accelerate renewables, but details on how to allocate these funds among technologies remain unclear, requiring advanced spatially explicit models to optimize power systems with geospatial details and coordinating infrastructure. Since 2000, rising global CO2 emissions have been driven mainly by growing energy demand in developing countries. Decarbonization can be more challenging in developing countries, yet their mitigation is indispensable to meet climate goals. China, responsible for 28% of global CO2 emissions and home to 18% of the global population, has announced a carbon neutrality target by 2060. Among renewables, PV and wind have broad applicability, impose fewer food and ecosystem trade-offs than bioenergy, and likely entail lower costs than CCS. Achieving Chinese carbon neutrality requires scaling PV and wind from 1 to 10–15 PWh year−1 by 2060. However, extrapolations of 2010–2020 growth (100 TWh year−1), CFED plans, and a recent high-resolution model suggest only 5–9.5 PWh year−1 by 2060. Growth could also slow due to declining subsidies, limited transmission infrastructure, and land-use protections. Hence, a spatially explicit optimization of generation, transmission, and consumption is needed for a country as vast as China. Methods exist for Europe and the USA that address spatial heterogeneity of resources, transmission and storage, but studies for China have rarely accounted for power-load flexibility and intertemporal learning dynamics. This study develops a unified optimization framework that incorporates geospatial siting capacities for new PV and wind, expansion of UHV transmission, energy storage, flexible power loads, and learning dynamics, highlighting the need for system upgrades and the co-benefits of increasing resident incomes.

The study situates itself within literature noting insufficient global decarbonization progress, large investment needs to meet Paris goals, and challenges specific to developing countries. Prior projections for China indicate PV and wind capacity of only 5–9.5 PWh year−1 by 2060 based on historical growth, CFED plans, or high-resolution models. Spatially explicit optimization frameworks have been applied in Europe and the USA to incorporate resource heterogeneity, transmission, and storage, yet Chinese studies have seldom included demand-side flexibility and learning-by-doing. Existing pathways to carbon neutrality often rely heavily on CCS, which faces economic, geological, and biomass constraints. The paper builds on and extends this body of work by integrating load flexibility, UHV expansion, storage, and endogenous learning in a comprehensive geospatial optimization for China.

The authors develop a unified, spatially explicit optimization framework to plan the location, capacity, and construction timing (by decade, 2021–2060) of new utility-scale PV and wind plants (>10 MW), minimizing levelized cost of electricity (LCOE). LCOE includes present-valued costs of capital investment, operation and maintenance, land acquisition, UHV transmission, and energy storage, divided by generated electricity over a 25-year plant lifetime. The model selects pixels for new installations to minimize system LCOE, subject to constraints on resource availability, administrative boundaries, land suitability and restrictions, terrain slope, land cover, latitude/longitude, ecological conservation (terrestrial and marine), offshore water depth and shipping routes, solar irradiance, wind power density, and air temperature. The framework coordinates generation with expansion of UHV transmission and deployment of energy storage, and includes flexible power-load management whereby end users shift hourly demand to better match supply (except for inflexible end-uses such as heating/cooling and electric vehicles; these constitute 12% of total power demand by 2060). Seasonal and diurnal profiles of PV and wind generation are modeled and contrasted with demand, with flexibility shifting daytime loads to align with PV production peaks. The optimization internalizes learning-by-doing dynamics, using empirically estimated learning rates in China (2000–2020), which are higher than historical international averages. By optimizing plant construction timing across decades, the model exploits cost declines to reduce overall LCOE. Electrification of non-power sectors is modeled to reach 58% by 2060. Transmission capacity expansion (~6.4 TW) and storage deployment (~1.3 TW power capacity) are evaluated for effects on power-use efficiency. Sensitivity analysis uses a baseline scenario (plant capacity capped at 10 GW; no transmission; no storage; constant growth; no learning) and sequentially adds: A) higher individual plant capacity limit (10→100 GW), B) new UHV lines, C) energy storage, D) higher electrification (0→58%), and E) demand-side flexibility. Case E becomes the optimal path when plant construction timing is further optimized with learning. The model computes plant-level marginal abatement costs (MAC) and aggregates emissions reductions by MAC ranges. It also analyzes land requirements, cost components, and financial impacts (initial investment, O&M, fossil fuel savings, and carbon cost savings at representative carbon prices). For carbon neutrality implications, the framework estimates residual CCS demand in 2060 by combining optimal PV and wind deployment with exogenous projections of other renewables (hydro, nuclear, biomass, hydrogen) and terrestrial carbon sinks, assuming coal meets residual demand and computing CCS needs. Co-benefits for poverty alleviation are assessed by quantifying revenues from PV and wind under carbon prices ($0–$100/tCO2), modeling financial flows embodied in interregional electricity transmission, and estimating changes in per-capita income distributions, poverty counts, and Gini coefficients via Monte Carlo uncertainty analysis.

- Optimal deployment raises PV and wind generation potential to 15 PWh year−1 by 2060 (vs 9 PWh year−1 under CFED), while lowering average abatement cost from $97/tCO2 (CFED) to $6/tCO2. - System LCOE falls from $0.067/kWh to $0.046/kWh by optimizing construction timing and leveraging learning dynamics (5% discount rate). - Required annual PV+wind investment increases from ~$127 billion/year in the 2020s to ~$426 billion/year in the 2050s (current level ~$77 billion/year in 2020). - Transmission expansion (~6.4 TW) and energy storage deployment (~1.3 TW) markedly improve power-use efficiency; adding storage contributes the largest capacity gain (+6.4 PWh year−1 versus transmission-only), while learning optimization yields the largest cost reduction (−$115 billion/year relative to a non-learning flexible-load case). - Plant-level MACs in 2060 vary from $166/tCO2 to $106/tCO2 across sites in the optimal path; 77% of PV and wind generation could be competitive against nuclear at lower MACs. For 9.5 PWh of generation, the average abatement cost is −$4.5/tCO2 (lower than prior estimate $27/tCO2 under 80% renewables). - Emissions abatement is boosted by storage (+3.5 Gt CO2 for plants with MAC < $100/tCO2) and by learning optimization (+3.5 Gt CO2 for plants with MAC < $0/tCO2). - Cost sensitivities: Average abatement cost increases from $2 to $14/tCO2 when the discount rate rises from 3% to 7%; it decreases from $22 to $0/tCO2 when plant lifetime increases from 15 to 35 years. Cost composition shifts from transformers and O&M toward modules and land acquisition along the path to optimal deployment. - Land and siting: Scaling PV and wind from ~1 to 15 PWh year−1 requires ~585,000 km2 for PV panels and ~672,000 km2 for wind areas, with 33% in deserts, 35% grassland, 16% ocean, and 6% cropland. - Annualized system costs: Initial investments ~$201 billion/year and O&M ~$47 billion/year (together ~7% of China’s 2020 public finance spending), partially offset by fossil fuel savings (~$223 billion/year) and carbon cost savings (~$399 billion/year at $100/tCO2). - Deployment scale: 3,844 plants identified (2,767 PV; 1,066 onshore wind; 11 offshore wind); 183 plants exceed 10 GW. Increasing the plant capacity limit from 0.1 to 10 GW reduces the average abatement cost from $62 to $6/tCO2. - System contribution: PV and wind can supply up to 59% of projected 2060 power demand, complementing other non-fossil sources (hydrogen, nuclear, biomass). - Carbon neutrality implications: Transitioning from CFED to the optimal path reduces CCS power generation needs in 2060 from 8.9 to 2.8 PWh year−1. PV+wind share of total power supply rises from 12% to 59% during 2021–2060, with annual share increases of ~1.8%, 1.4%, 1.0%, and 0.7% in the 2020s, 2030s, 2040s, and 2050s, respectively. - Poverty alleviation co-benefits: At $100/tCO2 in 2060, revenue redistribution via electricity transmission yields a financial flow of ~$1,055 billion, moving ~21 million people out of low-income (<$5,000/year) and adding ~6 million to high-income (>$20,000/year) groups; the Gini coefficient decreases from 0.453 to 0.441. The largest finance flow is from East China to Northwest China (~$223 billion/year). Regional outcomes: generation concentrated in Northwest (5.9 PWh/year) and North (5.2 PWh/year); consumption concentrated in East (5.7 PWh/year) and Central (4.3 PWh/year). Per-capita income increases from $29,000 to $34,400 in North China and from $29,100 to $30,600 in Northwest China at $100/tCO2.

The study addresses how China can feasibly scale PV and wind to achieve carbon neutrality targets cost-effectively. By jointly optimizing siting, capacity, and timing with coordinated UHV transmission, storage, and demand-side flexibility, the framework substantially increases achievable PV and wind output and reduces costs compared with CFED trajectories. Learning-by-doing and strategic timing of investments lower LCOE and abatement costs while storage and transmission mitigate temporal and spatial mismatches between variable renewable generation and demand. These system upgrades decrease reliance on CCS, alleviating economic and geological constraints associated with heavy CCS deployment. The findings emphasize the importance of policy interventions—investment in grid integration, storage, large utility-scale projects, and demand management—to overcome techno-economic limits as penetration rises. Beyond mitigation, optimized deployment can drive equitable economic benefits by channeling revenues from load centers to resource-rich, less-developed regions, reducing poverty and income inequality. The approach provides insights transferable to other large developing economies with vast resource heterogeneity, indicating that holistic power-system optimization can unlock higher renewable shares at lower societal costs and support broader socio-economic objectives.

The paper demonstrates that China can raise PV and wind generation to 15 PWh year−1 by 2060 with substantially reduced abatement costs through a unified, spatially explicit optimization that coordinates siting, UHV transmission, storage, demand-side flexibility, and learning dynamics. This pathway lowers LCOE, reduces CCS reliance, and yields significant co-benefits for poverty alleviation via interregional financial flows. Key contributions include: integrating intertemporal learning with geospatial system design; quantifying the roles of transmission, storage, and load flexibility; and mapping socio-economic benefits alongside climate gains. Future work could refine demand flexibility modeling across sectors, deepen representation of grid stability and ancillary services, expand uncertainty analysis of learning and finance costs, and extend the framework to other regions and to broader portfolios including sector coupling (heat, transport) and additional storage technologies.

- Affordability and feasibility of very large plants: While 183 plants >10 GW are identified and precedents exist (for example, Jiuquan wind at 20 GW; Yanchi PV at 1 GW), realizing this scale requires substantial grid integration, transmission buildout, and high upfront investments. Non-economic factors—ecological preservation, engineering feasibility, and political constraints—may limit implementation. - Cost sensitivity: Results depend on financial assumptions (discount rates, capital costs), plant lifetimes, and learning rates. Higher discount rates, higher capital costs, or shorter lifetimes increase abatement costs; international (lower) learning rates also raise costs. - System assumptions: The analysis assumes significant transmission expansion (UHV) and storage deployment, and a 58% electrification rate of non-power sectors by 2060; deviations could alter achievable capacities, costs, and abatement. - Scope of technologies: Other renewables (hydro, nuclear, biomass, hydrogen) are taken from external scenarios for carbon neutrality accounting; detailed co-optimization with these sources, grid reliability constraints, and ancillary services is not fully represented in the summary. - Market and policy risks: Potential slowdown from declining subsidies, infrastructure bottlenecks, and evolving supply-chain dynamics can affect deployment pace and costs, though the study suggests global supply-chain impacts on module prices may be moderate for China given rapid domestic declines.