Learnings

Outcomes

• A detailed technical report for WP1 has been completed and submitted to National Grid ESO. 
• A joint technical report for WP2 and WP3 is being compiled and with the aim to submit to National Grid ESO during May 2024. 
• An academic paper on the impedance/admittance model identification is being drafted and with the aim to submit later in 2024. 
   

Lessons Learnt

Lessons learnt from the work in WP1 are: 

• Free oscillations in power systems are becoming a major concern around the world. The dissipative energy flow (DEF) method, together with its variants, for system oscillation tracing is not guaranteed to be correct for all cases, especially for oscillations where IBRs are dominating.  
• The impedance participation analysis can indicate the root-cause correctly, but their calculation requires knowledge of both the left eigenvector (representing the input to the system) and the right eigenvector (representing the output from the system, also known as the mode shape). Even if the output can be acquired from the PMU data, it will remain difficult to obtain knowledge of input since there is no input for free oscillations. Further examination is needed to determine the useful information that can be extracted from the mode shape, and if it possible to introduce a known input disturbance in order to calculate the participation factor.  
• The present tools for oscillation tracing are limited. Even if a method indicates specific participants, the lack of a known ground-truth makes it difficult to validate the inferences. A proper small-signal model which could reproduce the oscillations and can offer theorical insights would greatly enhance the study of oscillations and their tracing.  
  
  

Lessons learnt from the work in WP2 and WP3 are: 

• Impedance/admittance model identification is useful for analysing potential oscillations and for post-event analysis to identify the root-cause. Both the time-domain identification and frequency-domain identification methods can be used. Whereas the proper equipment for system frequency scan is still under study and is absent for now, the time-domain identification making use of responses during events like step changes in the voltage or in the load would be an appropriate solution. Various algorithms can be chosen for time-domain identification, such as least-square method and eigensystem realization algorithms. 
• For time-domain impedance (or admittance) identification, it is important to know the details of the system input, including: the location where the event happened, the size of the change, and, preferably, a record of the waveform of the input. It also requires knowledge of system power flow, especially the information of loads and lines. It is preferable to know the parameters of synchronous generators (SG) so that analytical models of SGs can be applied as a priori information. A difficult case is buses which are not monitored by PMUs but with IBR connected where some assumptions have to make, such as using a generic model to represent the IBR, or, in a very simplified case, using an ideal current source. Most of the above knowledge will be accessible by the system operator so the issues raised can be dealt with.  
• When the input location is far from the locations measured, the estimation of the relevant off-diagonal elements of the whole-system impedance matrix will be inaccurate. Multiple inputs at a variety of locations during a certain period would be useful to improve the accuracy but a proper method to merge the over-determined systems is needed. This needs further study. 
• For a long-term system monitoring purposes, special equipment designed for frequency sweep, used in conjunction with frequency-domain identification, will be needed.