In the context of this PhD Thesis defense, Deep Shrestha presented his work realized at CTTC.
Power line communication (PLC) is considered as the most viable enabler of the smart grid. PLC exploits the power line infrastructure for data transmission and provides an economical communication backbone to support the requirements of smart grid applications. Though PLC brings a lot of benefits to the smart grid implementation, impairments such as frequency selective attenuation of the high-frequency communication signal, the presence of impulsive noise (IN) and the narrowband interference (NBI) from closely operating wireless communication systems, make the power line a hostile environment for reliable data transmission. Hence, the main objective of this dissertation is to design signal processing algorithms that are specifically tailored to overcome the inevitable impairments in the power line environment.
First, we propose a novel IN mitigation scheme for PLC systems. The proposed scheme actively estimates the locations of IN samples and eliminates the effect of IN only from the contaminated samples of the received signal. By doing so, the typical problem encountered while mitigating the IN is avoided by using passive IN power suppression algorithms, where samples besides the ones containing the IN are also affected creating additional distortion in the received signal.
Apart from the IN, the PLC transmission is also impaired by NBI. Exploiting the duality of the problem where the IN is impulsive in the time domain and the NBI is impulsive in the frequency domain, an extended IN mitigation algorithm is proposed in order to accurately estimate and effectively cancel both impairments from the received signal. The numerical validation of the proposed schemes shows improved BER performance of PLC systems in the presence of IN and NBI.
Secondly, we pay attention to the problem of channel estimation in the power line environment. The presence of IN makes channel estimation challenging for PLC systems. To accurately estimate the channel, two maximum likelihood (ML) channel estimators for PLC systems are proposed in this thesis. Both ML estimators exploit the estimated IN samples to determine the channel coefficients. Among the proposed channel estimators, one treats the estimated IN as a deterministic quantity, and the other assumes that the estimated IN is a random quantity. The performance of both estimators is analyzed and numerically evaluated to show the superiority of the proposed estimators in comparison to conventional channel estimation strategies in the presence of IN. Furthermore, between the two proposed estimators, the one that is based on the random approach outperforms the deterministic one in all typical PLC scenarios. However, the deterministic approach based estimator can perform consistent channel estimation regardless of the IN behavior with less computational effort and becomes an efficient channel estimation strategy in situations where high computational complexity cannot be afforded.
Finally, we propose two ML algorithms to perform a precise IN support detection. The proposed algorithms perform a greedy search of the samples in the received signal that are contaminated by IN. To design such algorithms, statistics defined for deterministic and random ML channel estimators are exploited and two multiple hypothesis tests are built according to Bonferroni and Benjamini and Hochberg design criteria. Among the proposed estimators, the random ML-based approach outperforms the deterministic ML-based approach while detecting the IN support in typical power line environment.
Hence, this thesis studies the power line environment for reliable data transmission to support smart grid. The proposed signal processing schemes are robust and allow PLC systems to effectively overcome the major impairments in an active electrical network. The efficient mitigation of IN and NBI and accurate estimation of channel enhances the applicability of PLC to support critical applications that are envisioned for the future electrical power grid.