2026 Theses Doctoral

Strategic and Resilient Energy Storage Operations: Economics, Cybersecurity, and Risk Management in Modern Power Grids

Wu, Yiqian

The rapid deployment of renewable energy resources is fundamentally transforming the structure and operation of power systems. Energy storage has emerged as a central enabling technology for this transition, bridging the temporal mismatch between intermittent generation and consumer demand. Yet as storage units grow into significant electricity market participants, their forecast-dependent, autonomous operation raises a set of interconnected challenges spanning market economics, operational risk, and cybersecurity. This thesis addresses in a unified framework grounded in the physics of power system operations.

The first part of the thesis investigates the strategic and economic dimensions of energy storage market participation. A bilevel optimization framework is developed to model market power exercise by generation companies under offer mitigation mechanisms, characterizing how strategic participants can exploit network congestion even within regulated markets. This framework is then extended to energy storage units, whose intertemporal capacity constraints create an ambiguity between legitimate arbitrage and market power exercise. Analytical conditions are derived that allow a market operator to distinguish between the two. The analysis concludes with a risk-aware arbitrage formulation accommodating robust, ellipsoidal, and chance-constrained uncertainty models, with risk-profit efficient frontiers providing systematic comparison across strategies.

The second part develops an online, distribution-free risk management framework for storage arbitrage under imperfect price forecasts. Electricity prices are non-stationary and difficult to predict, making static risk models inadequate. A conformal risk-aware arbitrage pipeline is proposed that dynamically calibrates dispatch conservativeness using the temporal difference error in the opportunity value function as a real-time proxy for forecast quality. Two calibration strategies are designed and proved to satisfy formal cumulative risk bounds, providing statistical guarantees without any assumptions on the price distribution. Numerical experiments on NYISO real-time market data demonstrate that the proposed framework recovers significant profit under poor forecast accuracy while preserving protection against downside risk.

The third part studies the cybersecurity vulnerabilities that arise from the data-driven nature of storage dispatch. A novel stealthy intraday false data injection (FDI) attack is developed that targets DER dispatch signals during renewable ramping periods to exhaust system reserves, and a convolutional neural network (CNN)-based sequential detection model is proposed that issues timely security alerts by monitoring multi-interval dispatch patterns. Validated on a real-world 187-bus distribution system with year-long operational data, the detection model demonstrates strong performance under both complete and incomplete attacker information. A second, related threat is then identified: forecast-induced coordinated attacks that manipulate shared price forecast signals to synchronize BTM storage dispatch at scale, without any direct device access. The attacker's strategy is formulated as a bilevel optimization problem, and a graph convolutional network (GCN)-based anomaly detection framework is proposed that localizes attack-induced load anomalies using spatial correlations across the grid. Results demonstrate that such attacks can exhaust system reserves and trigger blackouts, while the proposed detection framework achieves strong localization performance.

Together, the contributions of this thesis establish a comprehensive analytical and learning-based foundation for understanding how energy storage can be operated strategically, resiliently, and securely in modern power grids under the joint pressures of market competition, forecast uncertainty, and cyber-physical threats.