Demand response programs are essential tools for managing peak electricity demand and supporting grid stability, but their effectiveness depends on accurate load forecasting and protection against emerging cyber threats. A new research framework addresses these challenges by integrating three critical components: consumption prediction, behavioral modeling, and attack detection.
The core innovation uses a hybrid convolutional-bidirectional long short-term memory (CB-LSTM) neural network to forecast electricity consumption with higher accuracy than traditional methods. This improved forecasting enables utilities to segment customers into usage categories—low, medium, and high consumers—and assign appropriate demand response strategies to each group. The system allocates time-of-use pricing and direct load control schemes dynamically based on predicted behavior, optimizing cost-effectiveness across different customer types.
To account for consumer participation uncertainty, the framework employs a Z-number possibilistic-probabilistic approach that captures both the likelihood and confidence of customer engagement. This dual representation produces fairer incentive allocations and more realistic estimates of program effectiveness than conventional probabilistic models alone.
The cybersecurity dimension introduces a detection mechanism specifically designed to identify false data injection (FDI) attacks—malicious efforts to manipulate load profiles and corrupt incentive payments. The CB-LSTM-assisted deviation-bound-based detection and correction method achieved a 98% detection rate with only 1.1% false positives in simulations, successfully identifying compromised data while minimizing legitimate customer disruptions.
Simulation results demonstrate the integrated framework reduces peak demand by 24–27% while maintaining program cost efficiency. The approach represents a significant advancement for utilities implementing demand response, providing a pathway to deploy DRPs with confidence in their operational integrity and resilience against both technical uncertainties and coordinated cyberattacks. This work is particularly relevant for grid operators seeking to modernize demand-side management infrastructure.
