Decoding the 'Nature Encoded' Messages for Distributed Energy Generation Control in Microgrid

Shuping Gong, Husheng Li, Lifeng Lai, Robert Caiming QiuICC 2010 · DOI · OpenReview

Abstract

The communication for the control of distributed energy generation (DEG) in microgrid is discussed. Due to the requirement of realtime transmission, weak or no explicit channel coding is used for the message of system state. To protect the reliability of the uncoded or weakly encoded messages, the system dynamics are considered as a 'nature encoding' similar to convolution code, due to its redundancy in time. For systems with or without explicit channel coding, two decoding procedures based on Kalman filtering and Pearl's Belief Propagation, in a similar manner to Turbo processing in traditional data communication systems, are proposed. Numerical simulations have demonstrated the validity of the schemes, using a linear model of electric generator dynamic system.

Introduction

In the evolving landscape of microgrids, maintaining stability requires precise, real-time control of distributed energy generators (DEGs) such as solar arrays and wind turbines. A central controller relies on frequent state updates from these units. However, traditional error-correction methods (channel coding) introduce latency by requiring long data sequences to achieve high reliability. This creates a critical "delay-reliability" trade-off: while robust coding protects data, the resulting delay can compromise the stability of fast-acting physical systems. Consequently, microgrid sensors often transmit weakly coded or uncoded data. This paper proposes a novel solution: leveraging the inherent physical dynamics of the system as a "natural" form of error correction.

The Concept of "Nature Encoding"

The core innovation of this research is treating physical laws as an implicit "nature encoding." In a physical system, the current state ($x_t$) is mathematically dependent on its previous state ($x_{t-1}$). This temporal redundancy is functionally analogous to a convolutional code, which uses previous bits to determine the current bit. While traditional digital communication encodes messages in a Galois field, physical dynamics perform a similar operation in the real field.

Concatenated code structure representing nature encoding as an outer code.

By viewing the physical system as an "outer code" and the radio channel as an "inner code," the researchers establish a framework for joint decoding that mirrors the architecture of Turbo codes.

Proposed Decoding Architectures

The authors explore two primary methods for exploiting this natural redundancy:

  1. Kalman Filtering Based Heuristic: This approach utilizes a standard Kalman filter to predict the next system state. This prediction serves as a "prior" probability, aiding the receiver during the soft demodulation of the incoming radio signal.
  2. Pearl’s Belief Propagation (BP): This more advanced method models the entire process—from physical movement to bit transmission—as a single Bayesian network. Using an iterative "Turbo-like" scheme, the communication decoder and the physical state estimator exchange "beliefs" (probabilities) until they converge on the most likely state and message.

Simulation and Results

The proposed schemes were evaluated using a numerical simulation of a 7-dimensional linear dynamic model of an electric generator. The model tracks physical states such as rotor swings, excitation circuits, and voltage controllers over 1,000 time slots with varying signal-to-noise ratios ($E_b/N_0$).

Mean Square Error performance comparison between Kalman Filter and Pearl Belief Propagation.

The results demonstrate that the Pearl BP approach significantly outperforms standard Kalman filtering once the signal strength reaches a threshold (approximately 5 dB). At higher signal strengths, the BP method achieves a Mean Square Error (MSE) several orders of magnitude lower than traditional methods that separate state estimation from decoding.

Discussion and Insights

The research highlights a "Turbo" analogy where the physical world provides the redundancy needed for reliable communication. However, this synergy is a double-edged sword. At low signal-to-noise ratios, a "cascading effect" occurs: a decoding error leads to a poor state estimate, which in turn degrades subsequent decoding. This results in a performance "cliff" where the proposed algorithms may underperform traditional methods under extremely noisy conditions.

Conclusion

This paper successfully demonstrates that "nature encoding" can be harnessed to improve communication reliability in microgrids without the latency of traditional channel coding. Future work is needed to extend these techniques to non-linear system models and to optimize the computational overhead of iterative Belief Propagation for real-time, low-power sensor applications.

Summary updated

Abstract

The communication for the control of distributed energy generation (DEG) in microgrid is discussed. Due to the requirement of realtime transmission, weak or no explicit channel coding is used for the message of system state. To protect the reliability of the uncoded or weakly encoded messages, the system dynamics are considered as a `nature encoding' similar to convolution code, due to its redundancy in time. For systems with or without explicit channel coding, two decoding procedures based on Kalman filtering and Pearl's Belief Propagation, in a similar manner to Turbo processing in traditional data communication systems, are proposed. Numerical simulations have demonstrated the validity of the schemes, using a linear model of electric generator dynamic system.
BibTeX
@inproceedings{gong2010decoding,
  author = {Shuping Gong and Husheng Li and Lifeng Lai and Robert Caiming Qiu},
  title = {{Decoding the 'Nature Encoded' Messages for Distributed Energy Generation Control in Microgrid}},
  booktitle = {ICC},
  year = {2010},
  doi = {10.1109/icc.2011.5963048},
}
Figure