research highlights
Secure Outsourcing of Power System Data Analysis and Computation for Efficient
Operations of The Grid
by Dr. Kai Sun ([email protected]) and Dr. Jinyuan Stella Sun ([email protected])
At present, the control center of a power utility company or transmission operator is the unique location to gather and
analyze raw measurement data from the grid and to perform computation essential to grid operations, e.g., state estimation,
contingency analysis, oscillation mode analysis, and time-domain power system simulation. With the growing penetration of
intermittent renewable resources and responsive loads, the grid for real-time monitoring and control will significantly expand
in three dimensions: the complexity of the system model, the volume of online data, and the length of the required simulation
period. Computational burdens will be exponentially increased if all analysis and computation are handled at control center.
At CURENT, we developed a viable solution leveraging the cloud to provide computational resources needed by utilities
in a secure manner, i.e., outsourcing fully protected data from one or multiple utility companies while allowing the cloud
to perform big data analysis and computation over such protected data. This technology, if deployed at a utility company,
will greatly improve the company’s online data analysis and computing capabilities for grid operations without incurring
high cost or data security breaches. Specifically, we developed novel outsourcing algorithms for two representative types
of power system computation involving algebraic equations and differential-algebraic equations: Cross-utility data analysis
and computation, two examples of which include real-time multi-company inter-area oscillation analysis and multi-company
collaborative state estimation and powerflow-based contingency analysis; and Real-time time-domain power system
simulation.
Fig. 1 shows the experimental results for power system time-domain simulation where computationally expensive differential
equations are outsourced. We implemented the injective mapping-based outsourcing technique with Power System Toolbox
on the NPCC 48-machine, 140-bus power system model. Fig. 2 shows the results for inter-area oscillation analysis where
synchrophasor data is needed from multiple PMU clusters owned by different utilities and is securely outsourced for spectral
estimation. We implemented the homomorphic encryption-based outsourcing technique with Java’s BigInteger library and
Paillier cryptosystem. The spectrograms of the encrypted and recovered data with meaningful oscillation modes are shown in
Fig. 2.
(a) Original machine angles
b) Disguised machine angles
(c) Recovered machine angles
Fig. 1: Injective mapping-based outsourcing technique for power system time-domain simulation.
a) Encrypted spectrogram of the cluster angle
(b) Recovered spectrogram of the cluster angle
Fig. 2: Homomorphic encryption-based outsourcing technique for spectral estimation in oscillation analysis.
newsletter Summer 2015
6