AC Security Feedback Improves Pumped-Storage Grid Planning

Researchers from the Central China Branch of State Grid Corporation of China and Hunan University have developed a method for siting and sizing pumped-storage power plants that ensures high-voltage direct-current (HVDC) regional power grids remain operationally feasible under alternating current (AC) conditions. Published in Frontiers in Energy Research, the study addresses a critical gap in power-system planning where solutions derived from simplified direct-current (DC) power-flow models may fail to satisfy the complex physical constraints of actual AC networks. The proposed AC-security-feedback-based method integrates zonal source-load data with multiple HVDC infeeds to construct a joint scenario-feature set, thereby preserving the computational efficiency of mixed-integer programming while rigorously validating the plan against AC operational limits.

The research is motivated by the increasing variability of renewable generation and the growing complexity of regional grids fed by multiple HVDC transmission corridors. In such systems, renewable output, HVDC power transfers, and network constraints are strongly coupled, creating significant challenges for peak regulation and secure operation. Pumped-storage hydropower (PSH) plants are a key flexibility resource, offering peak regulation, frequency regulation, and emergency support. However, determining the optimal location and capacity of these plants requires coordinating resource conditions, power-system requirements, and network configuration, which cannot be adequately captured by system-wide flexibility requirements alone.

Traditional planning models often rely on DC power flow because of its simple structure and compatibility with integer programming. However, DC models neglect bus-voltage magnitudes, reactive power, line resistance, and active-power losses. Consequently, a plan that appears feasible under DC assumptions may result in AC overloads, voltage violations, or reactive-power deficiencies when applied to the full AC system. Directly embedding nonlinear AC power flow into multi-period capacity-expansion models substantially increases solution difficulty, often rendering them computationally intractable for complex scenario sets. The authors propose a framework that links DC power-flow planning with AC power-flow validation through an iterative feedback loop.

The methodology begins with the extraction of representative operating scenarios from historical hourly data for wind, photovoltaic (PV) generation, load, and HVDC infeeds. The team calculates the system residual load, which reflects the combined effects of regional demand, internal renewable output, and external HVDC infeeds on the flexibility required from conventional generation and PSH. To capture the spatial differences in regional sources and loads, the data are divided into feature blocks, standardized, and reduced in dimensionality using principal component analysis (PCA). K-medoids clustering is then applied to partition the historical days into representative scenarios, ensuring that each scenario corresponds to an actual historical day and preserves the inherent temporal correlations among load, renewable generation, and HVDC-infeed trajectories.

Following scenario extraction, the study formulates a two-stage mixed-integer linear programming (MILP) model for PSH siting and sizing. The first stage determines whether each candidate site is developed and its installed capacity, while the second stage optimizes the hourly operation of conventional generation, PSH, and the transmission network under different scenarios. The model accounts for integer unit configurations, multi-scenario operation, operational constraints of conventional generating units, and reserve requirements. Candidate sites are screened using hard constraints based on engineering feasibility, including rated head, conveyance-distance-to-head ratio, geological stability, and water availability. The economic parameters include fixed project investment, unit-capacity investment, and grid-connection costs.

The core innovation of the method is the AC security feedback mechanism. After solving the DC-based MILP model, the resulting planning scheme is subjected to rigorous AC power-flow validation for every scenario and time period. This validation checks bus voltages, generator reactive power, branch apparent power, active-power losses, and slack-bus adjustments. The security information obtained from the AC validation is fed back into the planning model to iteratively refine the solution. Specifically, the AC branch loading ratios are used to update the dynamic line limits in the DC planning model. If the AC loading ratio exceeds a trigger threshold, the line-limit factor is adjusted to reduce the active-power transfer space available to the DC model, thereby preventing AC overloads. A power-flow trust region is also introduced to suppress congestion shifting and maintain iterative stability.

The case study applies this method to a 500-kV equivalent receiving-end regional grid containing 35 buses and 52 transmission branches. The system includes 15 equivalent generating units, interprovincial AC tie lines, and HVDC-receiving terminals. Thirteen candidate PSH sites are evaluated, and five representative days are used in the planning and operation layer, yielding 120 operating points. The integer planning model selects five sites and installs 19 PSH units, with a total installed capacity of 4,950 MW and a total energy capacity of 29,700 MWh. The selected sites include S004, S010, S003, S013, and S005, with unit ratings of 300 MW or 150 MW depending on the site’s developable scale.

The operating characteristics of the PSH plants vary significantly across the five representative scenarios. In Scenario 3, which features high renewable output, both daily pumping energy and daily generation reach their respective maxima, with pumping power remaining between 4.0 and 4.8 GW during high-renewable periods. In Scenario 1, PSH operates predominantly in pumping mode during midday and switches to generating mode in the evening to support the system load peak. In Scenario 2, where renewable output is significantly lower, intraday PSH energy conversion is more limited, and maximum generating power is only 1,950.34 MW. These results demonstrate that PSH operation is determined jointly by renewable-output characteristics, conventional-unit operating limits, tie-line transfers, and network-flow distribution.

The AC security feedback mechanism proves essential for ensuring the feasibility of the plan. Before feedback is incorporated, branches 34, 44, and 52 remain overloaded under AC power flow, with maximum loadings exceeding 100% of their ratings. After dynamic-limit feedback and operating-schedule correction, all three branches return within their physical ratings. For instance, branch 34’s maximum loading decreases from 1.0050 to 0.9914, and branch 44’s loading drops from 1.0101 to 0.9841. These corrections demonstrate that AC overloads cannot be identified solely by comparing DC active-power flow with the branch rating, as AC apparent power is shared by active and reactive power.

The final AC power-flow validation results confirm that all 120 operating points converge strictly and satisfy the prescribed security constraints. The minimum voltage over all scenarios is 0.9619 p.u., and the maximum voltage is 1.0234 p.u., both within the allowable range of 0.95–1.05 p.u. The maximum branch loading is 99.14%, observed on branch 34 in Scenario 2, and the overload count is zero for every branch. Some generators and reactive-compensation devices reach their reactive-power limits in individual periods, but none exceed their specified limits. This confirms that the final plan maintains voltage security within the prescribed equipment capabilities.

The study highlights the importance of considering the interconnection location of PSH plants in relation to network topology and load distribution. By integrating AC security feedback into the planning model, the method adjusts operating schedules through time-varying line limits and power-flow trust regions, allowing all operating points to satisfy voltage, reactive-power, active-power-regulation, and branch-capacity constraints. This approach reconciles planning efficiency with AC operational feasibility, providing a robust framework for integrating large-scale renewable energy and HVDC transmission into regional power grids.

The authors acknowledge that the present model does not account for the effects of renewable-generation and load-forecast errors or forced unit outages on PSH operation. Future work may incorporate stochastic or robust optimization, extreme operating scenarios, and N-1 security constraints, together with reservoir operating limits and more detailed engineering-investment conditions. Adaptive update mechanisms for AC-feedback parameters also merit investigation to improve iterative efficiency and planning robustness in complex power grids. The study provides a significant methodological advance in power-system planning, demonstrating that AC security feedback can effectively eliminate overloads on critical branches while preserving the computational tractability of mixed-integer programming models.

Subject of Research: AC-security-feedback-based siting and sizing of pumped-storage power plants in multi-HVDC-infeed regional power grids

Article Title: Optimal siting and sizing of pumped-storage power plants in a multi-HVDC-infeed regional power grid via AC security feedback

Article References: Wang, Y., He, Z., Ling, X., Yu, X., Wang, J., & Zhu, L. (2026). Optimal siting and sizing of pumped-storage power plants in a multi-HVDC-infeed regional power grid via AC security feedback. Frontiers in Energy Research, 14, 2001155. https://doi.org/10.3389/fenrg.2026.2001155

Image Credits: AI Generated

DOI: 10.3389/fenrg.2026.2001155

Keywords: pumped-storage hydropower, AC power flow, HVDC transmission, grid planning, mixed-integer programming, renewable energy integration, power system security, scenario clustering, Optimal, siting, sizing, pumped-storage