Why do I see different measured values for frequency of the same three phase electrical signal while using different PLL & PMU block types?

Hi Community,

I have modelled a grid using a programmable three phase AC voltage source and there are three loads attached to it. Model & Parameter file is attached. The loads 2 and 3 are switched off at T = 2 secs and switched back on at T = 4 secs.

I want to measure the frequency deviation in the 3-ph electrical voltage (magnitude, direction and time duration of instability) during sudden load shedding & load reconnection.

For this I used PLL blocks. Now, Simulink offers a few options for PLL & PLL based PMU blocks and my issue is that each of these blocks gives me a different result for the exact same scenario. I want to know how much frequency deviation is happening, but with each block giving me varying results, how do I conclude which result is accurate?

Frequency Meausurement Outputs

Types of PLL & PMU blocks used

Method 1: Sinusoidal Measurement (PLL, Three-Phase)

Method 2: PLL (3ph)

Method 3: Phasor Measurement Unit (PMU, Three-Phase)

Method 4: PMU (PLL-Based, Positive-Sequence)

Main Model

My base assumption is - frequency deviation in voltage is inherent attribute of the voltage. No matter the measuring mechanism, a frequency of 50 Hz should be measured as 50 Hz +- tolerance.

The same conceptual doubt also extends to the gains of the PLL block. Modifying the gain will affect one or more of frequency deviation characteristics (magnitude, direction and time duration of instability). So how do I narrow down on what gain values would be appropriate?

Answer 1

[Sabin]

A PLL is a closed-loop feedback algorithm that can lock the output frequency to match the input frequency. There are a variety of algorithms that can be used, and tuning is very important because by adjusting the PLL parameters we will change the accuracy, convergence speed, and stability. From what I see in the plot all four blocks are outputting a frequency that is close to the reference, but they have a different response to changes in the input signal. Rather than comparing four different algorithms it would be better to create a test model where you tune and test that the algorithm you use is giving the expected results for different use cases.