Balancing-oriented hydropower operation makes the clean energy transition more affordable and simultaneously boosts water security

Decarbonizing power systems requires large-scale deployment of weather-dependent VRE (solar PV and wind), whose intermittency and stochasticity cause temporal mismatches between generation and demand. Ensuring reliable supply in high-VRE systems hinges on flexible resources. Reservoir hydropower, the largest grid-connected clean technology, provides carbon-free energy and critical flexibility (fast ramping, low operating cost, long-duration storage). Yet conventional peak-shaving hydropower operations were designed for historical load patterns with low VRE shares, risking missed opportunities for VRE integration and potentially exacerbating water–energy trade-offs, especially under climate shocks (droughts/floods). Existing planning models often under-represent hydropower flexibility due to insufficient spatial, temporal, and process resolution, which can lead to over-investment in alternative flexibility (e.g., pumped storage, batteries). This study develops a coupled reservoir-operation and capacity-expansion model (PREP-SHOT) to quantify how shifting from peak-shaving (FixedHydro) to balancing-oriented hydropower operation (AdaptiveHydro) affects system costs, VRE curtailment, emissions, and water sustainability in the China Southern Power Grid (CSG) under diverse decarbonization pathways.

Prior studies have explored hydro–VRE complementarities and flexibility needs in decarbonizing systems, but most analyze energy or water sectors in isolation, lacking multi-benefit assessments that capture non-linear, dynamic water–energy interactions, especially regarding water sustainability. Energy planning models frequently treat hydropower with fixed operations, ignore head dynamics, or aggregate plants, limiting accurate representation of plant-level flexibility, cascade operations, and hourly variability. This modeling gap can bias investment decisions towards storage or thermal flexibility and understate hydropower’s role in cost-effective, environmentally friendly decarbonization. The present work addresses these limitations by explicitly coupling short-term hydropower operations (hourly to daily, plant-level head/storage, cascade topology) with long-term capacity expansion to evaluate system-wide economic and water benefits of balancing-oriented operations across uncertain decarbonization policies and hydrologic conditions.

Model: Developed PREP-SHOT (Pathways for Renewable Energy Planning coupling Short-term Hydropower OperaTion), an open-source, multi-technology, multi-node, intertemporal capacity expansion model that jointly optimizes long-term investment and short-term operations with explicit hydropower flexibility. Objective: minimize total system costs (investment, fuel, fixed and variable O&M), subject to constraints on lifetime, carbon emissions, power balance, transmission, power output and ramping, storage operation, reservoir water balance, outflows, and storage. Hydropower representation: Plant-level modeling of 46 large hydropower stations with cascade topology. Hourly hydropower power is n·ρ·g·(generation flow)·(net head). Net head is computed from forebay/tailrace levels and head losses, where forebay depends on stage–storage curves and tailrace on rating curves (piecewise linear). Because head depends nonlinearly on storage and flow, PREP-SHOT uses a simulation-based iterative procedure: initialize head; solve linear program; update head using outputs (flows, storage); iterate until relative error threshold or max iterations. Operation schemes: Two schemes compared: (1) FixedHydro (baseline), peak-shaving-oriented with predefined hydropower output obtained by minimizing sum of squared remaining demand subject to hydro constraints; (2) AdaptiveHydro (balancing-oriented), where hydropower output is a decision variable co-optimized with system investments to minimize total cost, enabling flexible dispatch to complement VRE. System scope: Technologies include coal, nuclear (dispatchable); solar PV and onshore wind (non-dispatchable with curtailment); conventional hydropower; pumped storage hydropower (PSH); and lithium-ion batteries. Multi-node power transmission (five provincial nodes: Guangdong, Guangxi, Guizhou, Yunnan, Hainan) modeled via a transportation framework with capacity limits and losses (average 94% efficiency). Time slicing: annual investment years (2018, 2025, 2030, 2035, 2040, 2045, 2050); within each year, four seasons with representative 48-hour periods (selected via k-nearest neighbors/dynamic time warping from 2018 demand), capturing hourly variability. Scenarios: 81 decarbonization scenarios with 2050 CO2 reductions from 20% to 100% relative to 2018, with a common 2018–2030 cap (2018 level) followed by linear decline to the 2050 target. Hydrology: 117 inflow scenarios—1 normal (2018 proxy), 8 wet and 8 dry (±5% to ±40% scaling), and 100 interannual variability scenarios using bootstrap sampling of seven representative years from the 1979–2018 GRADES dataset, preserving spatial correlations. Data and assumptions: Initial capacities (2018) from official statistics; technology costs from NREL ATB 2020 with projected reductions (except hydro/PSH assumed constant); UHVDC transmission costs; coal CO2 intensity extrapolated from historical trends with piecewise linear fits; coal fleet age–capacity considered. Demand: 2018 hourly load (provincial) with projected growth rates scaled to future years. VRE capacity factors derived from MERRA-2 (solar: radiation, temperature; wind: wind speeds), aggregated to province, scaled relative to 2018; applied to representative days for all years (assumed climatology unchanged). Metrics: Curtailment rate defined as curtailed VRE divided by potential generation. Water sustainability value (WSV) of VRE estimated as the slope of water savings vs total VRE generation using piecewise linear fits across four decarbonization groups (low, medium, medium-high, high); introduced a proxy marginal water value µ (irrigation water price) in the objective to reflect value of conserved water; sensitivity showed water savings are insensitive to µ within tested range. Solution approach: Linear programming solved via simplex and barrier methods (Gurobi 9.5.0 via Pyomo), selecting the faster optimal solution. Start/stop constraints and advanced river routing (impulse response) assessed in post-analysis sensitivity; main results robust with minor quantitative changes.

• System cost reductions: Switching from FixedHydro to AdaptiveHydro reduces 2018–2050 total system-wide costs by up to 7.1% under a fully decarbonized 2050 grid, equivalent to US$28.2 billion. At a 20% 2050 reduction, total costs fall from US$356.1b (Fixed) to US$341.5b (Adaptive), saving US$14.6b (4.1%). Cost savings increase approximately linearly with deeper decarbonization despite nonlinear total cost growth.
• Drivers of savings: Decomposition shows major contributions from lower VRE investment and coal fuel costs. Notably, wind investment cost falls by 63%, coal fuel cost by 42%; solar investment increases by ~40% but the combined VRE investment cost is lower in AdaptiveHydro due to improved VRE integration. Overall, these categories account for ~65% of savings. Reduced VRE curtailment (0.7–3.8 percentage points lower) enables higher VRE utilization and displaces coal generation (coal fuel savings US$6.2–11.1b).
• Hydrology effects on costs: Cost savings are higher in dry years (median US$28.8b) than normal (US$28.1b) and lower in wet years (US$24.1b; 95% range US$20.0–28.4b) for a zero-carbon 2050. Incorporating interannual variability reduces median savings by 34.0%, 32.4%, and 21.2% relative to dry, normal, and wet cases, respectively, linked to smaller curtailment reductions due to seasonal inflow biases in sampled sequences.
• Water sustainability: Balancing-oriented operation plus higher VRE penetration saves substantial water for non-hydro uses by reducing hydropower generation needs and increasing operational heads. Under normal inflow and high decarbonization (80–100%), each additional 1 MWh of VRE yields annual water savings of about 320 m³. Water savings scales with VRE (volume effect) and improved timing shifts hydropower to periods of low VRE.
• WSV variability: WSV is higher in wet years than dry, with differences narrowing as decarbonization deepens. Example: under low decarbonization, median WSV drops by 238.1 m³/MWh (82%) from wet (290.5 m³/MWh) to dry (52.4 m³/MWh); under high decarbonization the difference falls to 148.5 m³/MWh (wet 473.9; dry 325.4). WSV sensitivity to inflow variability is smaller in dry (95% range 306.8–364.8; span 58 m³/MWh) than wet years (386.8–603.8; span 217 m³/MWh). Interannual variability yields WSV ranges generally between dry and wet cases, without a consistent pattern.
• Portfolio impacts: AdaptiveHydro prioritizes more solar relative to wind due to improved flexibility, reduced curtailment, and lower cost; it limits the need for storage-driven flexibility (PSH) especially in dry conditions, while FixedHydro requires proportionally more PSH to integrate VRE when hydropower flexibility is constrained.
• Dispatch dynamics: AdaptiveHydro shifts hydropower generation from daytime (peak-shaving) to nighttime to complement VRE availability, strengthening timing effects, reducing curtailment, costs, and required storage. Plants operate at higher heads, reducing water per unit electricity.
• System-level benefits: Non-monetary benefits include average annual carbon emission avoidance of about 57.2 Mt in high-reduction scenarios by adopting AdaptiveHydro. In a normal inflow year with a 100% 2050 reduction, annual water savings reach 123.8 km³—exceeding twice CSG’s agricultural water demand (60.2 km³) and over triple the Three Gorges Dam storage (39.3 km³).

The study demonstrates that explicitly leveraging hydropower’s operational flexibility to balance VRE delivers significant, previously underappreciated economic and environmental gains in high-VRE systems. By coupling short-term dispatch with long-term expansion, PREP-SHOT captures two mechanisms: (1) volume effects—greater VRE supply directly displacing fossil generation; and (2) timing effects—adaptive hydropower shifting output to low-VRE periods, cutting curtailment and enabling more VRE uptake. These mechanisms amplify with deeper decarbonization, explaining the near-linear growth in cost savings and substantial water savings per MWh of VRE under AdaptiveHydro. Hydrologic conditions modulate benefits: water savings are larger in wet years (greater inflows) while cost savings are larger in dry years because flexible hydropower reduces the need for costly storage expansion when generation-driven flexibility is limited. This highlights trade-offs between economic efficiency and water sustainability under climate variability and underscores the need to design portfolios and operations resilient to inflow uncertainty. The findings show that aligning hydropower operations with VRE balancing is a low-regret strategy that reduces system costs, emissions, curtailment, and water use, and is path-dependent—earlier and deeper decarbonization magnifies benefits.

By shifting from conventional peak-shaving to balancing-oriented hydropower operation and co-optimizing with capacity expansion, power systems can achieve sizable cost savings (up to 7.1% over 2018–2050 in the CSG), lower VRE curtailment, reduce reliance on fossil fuels and storage, abate emissions (~57.2 Mt/yr in high-reduction cases), and realize large water savings (e.g., 123.8 km³/yr under full decarbonization with normal inflow). The PREP-SHOT framework advances planning by explicitly modeling plant-level head dynamics and cascade operations at hourly resolution within long-term expansion decisions, enabling robust quantification of hydropower’s flexibility value. Future work should incorporate improved inflow forecasting and uncertainty, sedimentation impacts on storage, environmental operational constraints (e.g., gas supersaturation limits), dynamic interactions between water savings and head changes, and electricity market incentives for hydropower operators to adopt adaptive operations. Extending PREP-SHOT to other renewable-rich regions and transboundary basins can guide least-cost decarbonization pathways and sustainable water management, with enhanced cross-border coordination where needed.

Key limitations include: exclusion of small-reservoir and run-of-river plants and of reduced thermal-cooling water use from coal phase-out (likely understating water savings); omission of reservoir sedimentation impacts that reduce storage and flexibility (could overstate benefits); deterministic operations assuming perfect inflow forecasts—real-time uncertainty and flood-control priorities may constrain generation and diminish achievable benefits; unmodeled environmental constraints (e.g., total dissolved gas limits) that can cap discharges; no explicit coupling between water savings and head changes; market dynamics and incentive structures not modeled, potentially overestimating feasibility and benefits of AdaptiveHydro; simplified river routing with constant travel times (impulse response sensitivity shows main conclusions robust but at higher computational cost); not enforcing unit commitment start/stop and ramping costs in the core model (post-analysis suggests small reductions in cost savings by ~3.7–6.4% and negligible effect on water savings); reliance on seasonal representative 48-hour periods chosen from electricity data, which may not fully reflect water-use seasonality.