Research and analysis on noise characteristics of low voltage power lines carrier communication

Noise is one of the most important factors affecting the reliability of low-voltage power line carrier communication. In order to fully grasp the noise characteristics, a channel analysis equipment has been designed to collect and analyze noise. The analysis results show that power line noise is mostly manifested as a combination of colored background noise and pulse noise, mainly concentrated in the frequency band below 700kHz. Compared with the low peak period of electricity consumption, the peak period has stronger time-varying noise, larger amplitude of background noise, richer components of pulse noise, and there is also jitter in the duration of periodic pulse noise.


Introduction
Low voltage power line communication technology can be widely applied in fields such as smart grids, smart homes, and energy internet due to its wide coverage, low vulnerability to damage, and synchronization with power grid construction [1][2][3][4][5][6][7].However, the design of power lines was originally for the transmission of electricity.The diversity of power loads and their constant access and exit on the power network make the power communication environment not ideal, in which there exists strong frequency selectivity, complex power line noise, and strong temporal variability.In particular, various noise interferences superimposed on the power line cannot be simply regarded as white noise [8,9], which will greatly increase the bit error rate of signal transmission and seriously affect the quality of communication transmission [10].Although the distribution of power line channel noise is complex and time-varying, there are still rules to follow.Reference [11] is the most representative study on power line noise, dividing it into five basic types.It points out that colored background noise and narrowband noise can maintain the noise waveform unchanged for a period of time, and can be collectively referred to as background noise; The other three types of noise, collectively known as pulse noise, undergo waveform changes within microseconds and milliseconds, and their power spectra also change accordingly.Reference [12] measured the noise of an office building in a certain city and concluded that the noise power was 20dB at a frequency of 10kHz (reference value 1Mw/MHz), but decreased to -80dB at 100MHz, and the noise attenuation reached -20dB/decade frequency.Reference [13] measured the pulse noise generated by typical household appliances and found that the time-frequency statistics of the noise were the same as the AC power frequency period, and provided a spectrogram of the comprehensive time-frequency characteristics of the noise.There are also many research achievements and related literature in this field in China.Reference [14] measured the noise in the 500kHz~10MHz frequency band of the low-voltage power grid and found that the noise power at frequencies greater than 2MHz was lower than that in the 10~450kHz frequency band.The noise amplitude showed a decreasing trend with increasing frequency, which is consistent with the conclusions of relevant foreign literature.Reference [15] measured the power line noise at different nodes in a typical urban residential area, and the measurement results showed that the noise level at low frequencies was significantly higher than that at high frequencies.The noise power ranges from 50dB to 80dBuV in the frequency range of 50kHz to 100kHz.When the frequency range is 400kHz~500kHz, the noise power is between 30~60dBuV.The maximum value of the change in noise power is between 30-40dBuV.Through the above analysis and research, it can be seen that there are many types of power line channel noise and their distribution is complex.The noise measurement results obtained at different locations and times have time-varying characteristics.These all pose significant challenges to the statistics and analysis of power line channel noise.
With the development of power communication, HPLC (High Speed Power Line Carrier) communication technology has become an important component of the power industry.The Technical Specification for Interconnection and Interoperability of Low Voltage Power Line High Speed Carrier Communication stipulates that HPLC supports communication in four frequency bands: 1.953MHz~11.96MHz(band 0), 2.441MHz~5.615MHz(band 1), 0.781MHz~2.930MHz(band 2), 1.758MHz~2.930MHz(band 3).This article conducts research and analysis on power line noise in the frequency range of 0.781MHz~11.96MHz,which can cover the four communication frequency bands of HPLC.It helps to further improve the communication performance of HPLC and provides support for optimizing the passband range of HPLC receiver filters and calibrating the threshold size of pulse noise suppression modules.
The content of other chapters in this article is as follows.In Section 1, the causes and classification of power line noise are detailed.In Section 2, a channel analysis device was introduced, which can be used for collecting noise in the 0.781MHz~11.96MHzfrequency band.The method of separating measured noise into background noise and pulse noise using simulation software was also introduced, laying the foundation for subsequent analysis of measured noise.In Section 3, the channel analysis equipment introduced earlier was used to collect 30 sets of power line noise from a certain residential area in Tai'an City during the two time periods of 10:00-11:00 and 17:00-18:00 on November 21 and November 22.Typical noise during these two time periods was exemplified and analyzed, and the noise characteristics of the two time periods were compared and analyzed at the end of Section 3. Finally, a summary and review of the article were conducted in Section 4.

Power line noise and its classification
The noise on the power line can be divided into non-human noise and human noise according to the noise source.Non human noise is caused by natural phenomena such as lightning.Artificial noise is generated by the irregular connection and disconnection of electrical equipment, or by a large number of household appliances, broadcast signals, switching power supplies, etc.The power line environment is complex and changeable, so the random noise can not be simply regarded as ordinary Gaussian white noise.
Through extensive research on literature and measured data, it has been found that the noise in lowvoltage power line channels can be subdivided into colored background noise, narrowband noise, asynchronous periodic pulse noise at power frequency, synchronous periodic pulse noise at power frequency, and burst pulse noise [16].Colored background noise is generated by electronic thermal noise and low-power household appliances such as hair dryers, computers, and induction cookers.Colored background noise can occupy the entire power line communication bandwidth.Compared with other noises, the power spectral density of this noise is relatively low, and the power spectral density changes slowly with time.The power spectral density of background noise generally decreases with the increase of frequency, but sometimes it also increases significantly at lower frequencies.The main sources of narrowband noise are amateur radio, medium or short wave broadcast signals, and the scanning synchronization signals of computers and televisions are also the cause of this noise.The power spectral density of narrowband noise is high and the frequency band is narrow, which is random but also changes slowly with time.The periodic pulse noise asynchronous to the power frequency is generated by the periodic switching of high-power electrical equipment, and the periodic frequency is generally between 50 and 200Hz.Periodic pulse noise synchronized with power frequency is caused by electrical equipment such as silicon controlled devices and switching devices in switching power supply.The noise frequency is 50Hz or its integer multiple.Such noise has large power and long duration, and the power spectral density decreases with the increase of frequency.Burst pulse noise is caused by short circuit fault of power line, sudden switch of electrical equipment and other reasons.It is a pulse interference or pulse interference group with large energy.The energy is relatively concentrated, the power spectral density is relatively large, the spectrum is very wide, and the time variation is strong.
Because colored background noise, narrowband noise, and periodic pulse noise asynchronous to power frequency change slowly over time and can remain stable for a relatively long time, these three types of noise can be collectively referred to as background noise for analysis.The periodic pulse noise and random pulse noise synchronized with the power frequency have strong time-varying characteristics, which can be collectively referred to as pulse noise for analysis.This can simplify the model and greatly reduce the time for analysis and calculation.

Noise measurement
The collection of on-site power line environmental noise in this article is completed by the power line channel characteristic analysis equipment.The channel analysis equipment is mainly controlled by Xilinx ZYNQ7020 processor.The sampling frequency is 100MHz, covering the measurement range from 0.7MHz to 12MHz.The sampling accuracy is ±0.5mV, and the sampling range is ±10V.It can achieve real-time collection, analysis, and storage of parameters such as noise, impedance, attenuation, and delay in the low-voltage power line carrier communication station area.It has the characteristics of high precision, low cost, and easy portability.
The channel analysis device consists of a hardware acquisition device and upper computer software.When conducting noise collection, the connection of channel analysis equipment is shown in Figure 1.The hardware collection device interacts with the upper computer through a gigabit Ethernet port, and is connected to a low-voltage power line through a standard line with a shielding layer, which is not affected by environmental radiation signals during collection.When conducting noise collection work, the upper computer software sends instructions to the hardware acquisition device, uses the coupling module of the acquisition device to collect power line noise, and after processing such as gain adjustment and ADC sampling, the noise data is transmitted to the upper computer.The upper computer analyzes and processes the noise data, and displays the analysis results of the noise in an image.
The schematic diagram of power line noise collection is shown in Figure 2. is the equivalent impedance of power line, and the right side of the dashed line is the analog front-end of the collection device AFE.In the required frequency band, the equivalent resistance of AFE is far greater than the impedance of power line .Therefore, it can be seen from the voltammetry method that the voltage at both ends of BC can be equivalent to the voltage at both ends of AC.If we want to measure the noise at both ends of the AC, that is, the voltage change at both ends of AC, we can connect the collection device to both ends of the BC to obtain the voltage change at both ends of BC.In other words, in Figure 2, the right side of the dashed line is the hardware acquisition device, the left side of point B is assumed to be the live line and the left side of point C is assumed to be the zero line.The impedance of the power line is equivalent to . By clamping the collection device at both ends of BC, the voltage changes at both ends of BC can be collected.Due to the fact that the equivalent resistance of AFE is much greater than the impedance of the power line within the required frequency band, the partial voltage can be ignored.The voltage at both ends of BC is consistent with the voltage at both ends of AC, achieving the purpose of collecting noise at both ends of AC.

Noise analysis
This article uses channel analysis equipment to collect multiple sets of power line noise, save them as txt files, and simulate them on MATLAB.In noise analysis, it is not only necessary to perform timefrequency analysis on the measured noise itself, but also to separate the noise into colored background noise and pulse noise for time-frequency analysis.
In the research of existing noise separation methods, there are mainly two methods: time-domain method and frequency-domain method.The core idea is to treat noise above a certain value as pulse noise in the time or frequency domain, and the remaining part is considered background noise [12,[17][18][19][20].This article obtains background noise by averaging multiple sets of noisy data to remove random impulse noise.In practical operation, collecting noise data for a long time at power line nodes, truncating it to the same length, averaging multiple sets of noise data as the background noise of the power line channel, and then treating noise higher than a certain value of the background noise as pulse noise.Related studies have shown that extracting noise that is about 10dB higher than the background noise as pulse noise has good fitting performance.So, here we truncate the noise data for a long time into a set of 20ms, and average multiple sets of data to obtain a background noise of 20ms.After subtracting the background noise from the measured power line noise, take the noise whose amplitude is higher than the average by more than 10dB as pulse noise.

On site data analysis
The topology diagram of a certain residential area in Tai'an is shown in Figure 3.There are a total of 10 buildings, with high electricity consumption and complex electricity environment.The cables are mostly ground cables, mainly loaded with household appliances such as lamps and refrigerators.It is a typical urban residential area with severe communication environment.This section uses the channel analysis equipment introduced in the previous section to collect noise in a certain substation area of Tai'an City.The collection time is from 10:00 to 11:00 and 17:00 to 18:00 on November 21 and November 22. Three power line nodes were selected in the unit building, and a total of 30 sets of noise data with a length of 1 second were collected.The period from 10:00 to 11:00 is the low peak period for residents' electricity consumption, during which the noise collected generally has no significant change except for occasional carrier signals and random pulse noise.The period from 17:00 to 18:00 is the peak period for residents' electricity consumption.During this time period, in addition to carrier signals and random pulse noise, certain noise components will also be suddenly added at a certain time point and continue for a period of time.The typical noise between 10 and 11 o'clock and between 17 and 18 o'clock are selected for analysis and comparison in the following text.

Typical noise analysis from 10:00 to 11:00
The noise during this time period is relatively stable.Figure 4    From the time domain diagram Figure 4, the measured noise is manifested as pulse noise and colored background noise, with noise intensity generally below 0.2V.From the frequency domain diagram Figure 5, the overall pattern shows a gradual decrease in noise intensity as the frequency increases.The maximum value of the spectrum is around 900kHz, with a maximum amplitude of 69 dBuV.Separate the measured noise from background noise and pulse noise.The background noise and its spectrum are shown in Figure 6 and Figure 7, and the pulse noise and its spectrum are shown in Figure 8 and Figure 9.As shown in Figure 6, the amplitude of background noise ranges from -0.035V to -0.015V, while in Figure 8, the amplitude of pulse noise is relatively large.The background noise spectrum in Figure 7 is roughly the same as the measured noise trend in Figure 5.The spectrum curve of pulse noise in Figure 9 is smoother, and the energy of pulse noise is smaller in the communication frequency range of 700kHz~12MHz.Locally amplify the pulse noise as shown in Figure 10.As shown in Figure 10, the separated pulse noise includes periodic pulse noise and random pulse noise.The period of periodic pulse is 20ms, the amplitude is between 0.1V and 0.2V, the duration is 6.5ms, and the duration interval is 13.5ms.The amplitude of random pulse noise is around 0.1V.

Typical noise analysis from 17:00 to 18:00
This period is the peak period of electricity consumption, and residents will generate various types of noise when using electricity, with strong time-varying noise.A group of noise from this time period is selected and compared with the noise from 10:00 to 11:00.
From the time domain diagram Figure 11, the measured noise is manifested as pulse noise and colored background noise, with noise intensity generally below 0.3V.From the frequency domain diagram Figure 12, the overall pattern shows a gradual decrease in noise intensity as the frequency increases.The maximum value of the spectrum is around 900kHz, with a maximum amplitude of 70dBuV.Compared with the noise in Figure 5, the noise in this time period is relatively high at 7MHz to 12MHz, manifested in the form of harmonics, with a harmonic period of 56.3kHz.As shown in Figure 13, the background noise amplitude is between -0.07 and 0.01V, while the pulse noise amplitude in Figure 15 is relatively large.The background noise in Figure 14 has high energy in the frequency range of 700kHz to 1.7MHz, as well as in the frequency range of 2.1MHz to 2.8MHz.It is believed that this background noise contains narrowband noise and is caused by unsuccessful carrier communication.In the high-frequency range, from 7MHz to 12MHz, it exhibits the same harmonic form as the original measured noise.Locally amplify the pulse noise as shown in Figure 16.The separated pulse noise in Figure 17 contains two sets of periodic pulse groups.The first set of periodic pulses has a period of 10ms, an amplitude of 0.25V, a duration of 3.5ms, and a duration interval of 6.5ms.The second group of periodic pulses has a period of 20ms, an amplitude of 0.13V, a duration of 6.5ms, and a duration interval of 13.5ms.In addition, there is also random pulse noise, with an amplitude of around 0.15V or 0.4V.

Comparison of noise between two time periods
During the two time periods, 15 sets of noise were collected, and in the previous section, a typical set of noise was selected for analysis.It can basically represent the characteristics of noise within the current time period.
Among the 15 sets of noise from 10:00 to 11:00, the measured noise in the time domain is characterized by pulse noise and colored background noise, with noise intensity generally below 0.2V.In addition, the first seven sets of noise data all have narrowband interference caused by unsuccessful carrier communication, with amplitudes ranging from 0.3V to 0.4V.In the frequency domain, it exhibits high energy in the 700kHz to 1.7MHz and 2.1MHz to 2.8MHz frequency bands.From the frequency domain perspective, the overall pattern shows a gradual decrease in noise intensity as the frequency increases.The maximum value of the spectrum is around 750kHz or 900kHz, and the maximum value is within the range of 69 dBuV to 70 dBuV.After separating the measured noise into background noise and pulse noise, in the time domain, the peak to peak value of the background noise does not exceed 0.04V.The separated pulse noise contains periodic pulse noise and random pulse noise.The period of periodic pulse is 20ms, the amplitude is between 0.1V and 0.2V, the duration is 6.5ms, and the duration interval is 13.5ms.The amplitude of random pulse noise is between 0.3V and 0.4V.
Among the 15 sets of noise from 17:00 to 18:00, in terms of time domain, the measured noise is manifested as pulse noise and colored background noise, with noise intensity generally ranging from 0.1V to 0.3V.In addition, some noise data may contain carrier signals or random pulse noise, with amplitudes of around 0.5V, 1V, or 5V.From the frequency domain perspective, the overall pattern shows a gradual decrease in noise intensity as the frequency increases.The maximum value of the spectrum is around 750kHz or 900kHz, and the maximum value is within the range of 67dBuV to 73dBuV.In the higher frequency range, the noise energy is greater than during the low peak period of electricity consumption.After separating the measured noise into background noise and pulse noise, in the time domain, the peak to peak value of the background noise does not exceed 0.25V.Analyze the separated pulse noise.In the 1st to 4th set of data, only one set of periodic pulse noise is included, with a period of 20ms, an amplitude of 0.1V to 0.15V, and a duration of 6.5ms to 9ms for jitter.In the subsequent data, in addition to the periodic pulses mentioned above, a new set of periodic pulses with a period of 10ms is added, and the pulse group amplitude ranges from 0.22 to 0.3V, with a duration of jitter between 3.5ms-4ms.In addition, there is random pulse noise.
Therefore, compared to the low peak period of electricity consumption, the noise during the peak period of electricity consumption has stronger time-varying characteristics, larger amplitude of background noise, larger amplitude of pulse noise, richer components of pulse noise, and there is also jitter in the duration of periodic pulse noise.

Conclusions
The noise distribution in low-voltage power line channels is complex and time-varying, which can seriously affect the quality of power line communication.However, through a large number of literatures on the measurement and analysis of power line noise, it has been found that although power line noise cannot be described by a simple Gaussian distribution, there are also rules to follow.In addition, since HPLC communication technology has become an important part of the power industry, in order to improve the efficiency of HPLC communication and continuously optimize the HPLC module, this paper measured and analyzed the noise in the low-voltage broadband over power lines broadband carrier communication frequency band (0.781MHz~11.96MHz).
Firstly, a channel analysis device is introduced, which can be used for measuring noise and storing data within the carrier communication frequency band.Secondly, a two-day noise collection was conducted using channel analysis equipment in a typical residential area with severe communication environment, and simulation software was used to separate the measured data into background noise and pulse noise.Finally, statistical analysis was conducted on the collected data, and the conclusions obtained are as follows: from the time domain perspective, noise data generally manifests as a combination of colored background noise and impulse noise, where the amplitude of colored background noise is small and includes background noise and narrowband noise, the amplitude of pulse noise is relatively large, generally including periodic pulse noise and random pulse noise.The period of periodic pulse noise is generally 10ms or 20ms, which is related to the power frequency period.And due to the varying power loads connected to the power network at different times, there may be more than one set of periodic pulse noise, and the duration within a cycle is uncertain.From the frequency domain perspective, the measured noise exhibits a characteristic of higher frequency and lower noise amplitude in the frequency domain.The maximum value of the spectrum is generally within the frequency range of 700kHz~900kHz, with an amplitude of around 70dBuV, and the amplitude is even greater during peak power consumption periods.The energy of pulse noise is relatively small in the communication frequency band (700kHz~12MHz), and the energy is mainly concentrated below the low-frequency 700kHz.Compared with the low peak period of electricity consumption, the peak period has stronger time-varying noise, larger amplitude of background noise, richer components of pulse noise, and there is also jitter in the duration of periodic pulse noise.

Figure 2 .
Figure 2. Principle of power line noise acquisition.

Figure 3 .
Figure 3. Topological map of Tai'an District.
selects a set of noise data without carrier signals as an example to analyze the noise during this time period.

Figure 12 .
Figure 11.Noise signal at 17:12 on Nov. 22.Figure 12. Noise spectrum at 17:12 on Nov. 22.Separate the measured noise from background noise and pulse noise.The background noise and its spectrum are shown in Figures 13 and Figure 14, and the pulse noise and its spectrum are shown in Figures 15 andFigure16.