New Method Reduces Database Errors, Enhancing Models to Improve Performance of Power Grids

Achievement date: 

Researchers at Northeastern University (NEU) have developed an enhanced method for detecting and correcting errors in network parameters—i.e., data that affect the results of computer models used to simulate changes in electrical power systems—giving engineers improved capability to study performance and predict disturbances in large power grids. This research is supported by the Center for Ultra-Wide-Area Resilient Electric Energy Transmission Networks (CURENT), an NSF-funded Engineering Research Center (ERC) that is headquartered at the University of Tennessee-Knoxville (UTK) with NEU as one of its partner institutions.


The quality of electrical power may be described as a set of values of parameters, such as: continuity of service (i.e., whether the power is subject to voltage drops or other conditions that can cause blackouts or brownouts); variation in the magnitude of voltage variations than can be tolerated; transient voltages and currents caused by a sudden change of state; and balance in the waveforms, which are the shape and form of signals indicating voltage changes in AC power. In energy management systems, state estimation (SE) is a key function for building real-time models to assess a variety of factors affecting power grid processes. Successful control and operation of power systems requires correct SE information as well as data on optimal power flow, security analysis, and various control and protection schemes. Traditionally, SE is implemented assuming perfect knowledge of the network model and suspecting no errors in the parameters—an assumption that is not always true. The scheme developed at NEU for network parameter error identification and correction based on SE addresses this problem. 


SE is a fundamental functionality needed to ensure smooth, reliable, and secure operation of power grids. At the transmission level, SE has a rich history. However, new developments in smart grid technologies render such existing methods inadequate, as the demands on SE are also much more stringent now and concerns such as reliability, dependability, security, and the distributed and dynamic nature of SE necessitate a paradigm shift from existing algorithms.

The core of NEU’s identification technique is the joint Normalized Lagrange Multiplier/Normalized Residual (NLMNR) test. This method has the following novel features which existing techniques lack: capabilities (a) to differentiate between parameter and measurement errors and (b) to inspect all parameters together without requiring a suspicious parameter set to be specified; and (c) high computational efficiency and low memory requirements. In addition to error identification, a multiple-measurement-scan scheme was also developed to enhance the performance of the NLMNR test for parameters whose errors may be buried in the background of measurement noise. For parameter error correction, a highly efficient and stable linear correction scheme was developed to replace existing techniques that features computational efficiency (low time and memory costs) and no need for data structures or estimation codes to be modified.