State Estimation and Forecasting
NREL researchers are developing advanced data analytics for estimating and forecasting grid conditions to support operations and planning on power systems with high penetrations of renewable energy resources.
The power system was designed with dispatchable generation to provide electricity to meet demand and additional reserves to meet contingencies. But better methods are needed to estimate and forecast grid conditions to optimally manage flexible resources and mitigate the risks associated with variable generation, such as wind and solar.
NREL is develop estimation and forecasting methods by leveraging data-driven approaches and physical models to provide real-time and predictive situational awareness and inform decision-making in electric power systems.
- State estimation with limited measurements
- Machine learning for short-term state forecasting
- Predictive state estimation-enabled optimization
With the rising penetration of distributed energy resources, distribution system control and enabling techniques such as state estimation have become essential to distribution system operation. However, traditional state estimation techniques have difficulty with the low-observability conditions often present on distribution systems because of a paucity of sensors and heterogeneity of measurements. To address these limitations, NREL researchers are developing distribution system state estimation methods that employ data-driven matrix/tensor completion augmented with power flow constraints to recover the operation states of an entire system. These methods operate in systems that lack full observability and where standard least-squares-based methods cannot operate, and flexibly incorporates any network quantities measured in the field.
Because of the variability of renewable generation, grid conditions vary both spatially and temporally. To accommodate high penetrations of renewables, state forecasting is crucial for utilities and grid operators to dispatch controllable resources, prepare for changing grid conditions, and reduce operation costs. NREL researchers develop machine learning-based methods for predicting system states in the short-term future by employing neural networks, decision tree-based approaches, and ensemble learning. With forecasted system states, grid operators are able to better coordinate control efforts and prioritize control needs—therefore improving grid reliability, resilience, and economic efficiency.
Decentralized Low-Rank State Estimation for Power Distribution Systems, IEEE Transactions on Smart Grid (2020)
Machine Learning-Based Prediction of Distribution Network Voltage and Sensor Allocation, IEEE Power and Energy Society General Meeting (2020)
Matrix Completion for Low-Observability Voltage Estimation, IEEE Transactions on Smart Grid (2020)
Predictive Analytics for Comprehensive Energy Systems State Estimation, Big Data Application in Power Systems (2018)
Short-Term Distribution System State Forecast Based on Optimal Synchrophasor Sensor Placement and Extreme Learning Machine, IEEE Power and Energy Society General Meeting (2016)