Power utilities worldwide experience problems related to harmonic distortion of voltage waveforms. A non-linear element will cause a non-sinusoidal current to be drawn from the power supply, which can lead to capacitor failures and interference with communication circuits – not to mention catastrophic failures.
To ensure adequate quality of supply, utilities try to keep harmonic distortion below levels specified by published standards, such as the Australian Standard 2279 and the American IEEE 519 Standard. If the levels specified in the standard are exceeded – even if nothing fails – field investigations are performed to locate offending harmonic sources, and these investigations require a combination of power engineering and signal analysis skills and tools to match.
The investigative team at the Queensland Electricity Commission (QEC) used DADiSP, a graphical display and data management softwaretool, to determine the cause of the failures. With DADiSP and some creative problem-solving techniques, the engineers identified the troublesome harmonics in the system and traced their source.
In a power system, synchronous generators produce energy at a fundamental frequency of 50 or 60 Hz. Non-linear loads convert this energy into useful output and into harmonic energy which flows back into the power system to be dissipated. Since harmonic power flow is from the non-linear load back 'towards' the power system, it is possible to locate an offending harmonic source by determining the direction of the harmonic power flow. The goal of the DADiSP processing was, therefore, to determine the direction of harmonic power flow.
About 30 current and voltage signals were measured using a data acquisition system controlled by a personal computer. The data acquisition system was triggered by high current levels in the filter bank which indicated a sixth harmonic resonance. Recording duration was about two seconds, with a sampling rate of 2,560Hz per channel. Recorded data were translated from HP format into a binary file and imported into DADiSP for post-processing, the voltage and current harmonic components being extracted via Fast Fourier Transforms (FFTs). A DADiSP macro, which applied a frequency-related phase shift to the phase output of the FFT, was written and used to compensate for skewing errors caused by the sequential sampling of the signals.
Since data had been acquired at a fixed sampling rate, even though the power system frequency varied by as much as 0.2Hz, spectral leakage posed a serious threat to accuracy. The investigative team reduced leakage errors to sufficiently low levels using DADiSP's Hanning window function, then verified them using test signals created with DADiSP's 'G' (Generate) functions. Another macro was written which acted on two series - a voltage and a current - to produce a power flow versus frequency output. When the problems were isolated it was found that the transformers under investigation had gone into saturation in the presence of non-symmetric load currents – currents containing both even and odd harmonics. The sixth harmonic components were found to be exciting the resonant condition and particularly bad resonances were found to occur when the large power transformers were energised in the vicinity of the filter bank. These resonances produced high voltage stresses on the capacitors thus contributing to their failure.
One of the investigators on the team, in particular, appreciated DADiSP's help with the project. In a report on the investigation he wrote: "The ability of DADiSP to automate the bulk of the signal processing through the use of macros and pre-defined spreadsheets contributed to tremendous savings in manpower."
The team is likely to use DADiSP as a standard tool to analyse similar problems in the future – though there’s no expectation of further catastrophic failures in the system!