Data-driven Approach for State Prediction and Detection of False Data Injection Attacks in Smart Grid

In a smart grid,state estimation(SE)is a very important component of energy management system.Its main functions include system SE and detection of cyber anomalies.Recently,it has been shown that conventional SE techniques are vulnerable to false data injection(FDI)attack,which is a sophisticated new class of attacks on data integrity in smart grid.The main contribution of this paper is to propose a new FDI attack detection technique using a new data-driven SE model,which is different from the traditional weighted least square based SE model.This SE model has a number of unique advantages compared with traditional SE models.First,the prediction technique can better maintain the inherent temporal correlations among consecutive measurement vectors.Second,the proposed SE model can learn the actual power system states.Finally,this paper shows that this SE model can be effectively used to detect FDI attacks that otherwise remain stealthy to traditional SE-based bad data detectors.The proposed FDI attack detection technique is evaluated on a number of standard bus systems.The performance of state prediction and the accuracy of FDI attack detection are benchmarked against the state-ofthe-art techniques.Experimental results show that the proposed FDI attack detection technique has a higher detection rate compared with the existing techniques while reducing the false alarms significantly.

A VAE-Bayesian deep learning scheme for solar power generation forecasting based on dimensionality reduction

The advancements in distributed generation(DG)technologies such as solar panels have led to a widespread integration of renewable power generation in modern power systems.However,the intermittent nature of renewable energy poses new challenges to the network operational planning with underlying uncertainties.This paper proposes a novel probabilistic scheme for renewable solar power generation forecasting by addressing data and model parameter uncertainties using Bayesian bidirectional long short-term memory(BiLSTM)neural networks,while handling the high dimensionality in weight parameters using variational auto-encoders(VAE).The forecasting performance of the proposed method is evaluated using various deterministic and probabilistic evaluation metrics such as root-mean square error(RMSE),Pinball loss,etc.Furthermore,reconstruction error and computational time are also monitored to evaluate the dimensionality reduction using the VAE component.When compared with benchmark methods,the proposed method leads to significant improvements in weight reduction,i.e.,from 76,4224 to 2,022 number of weight parameters,quantifying to 97.35%improvement in weight parameters reduction and 37.93%improvement in computational time for 6 months of solar power generation data.

FedDiSC:A computation-efficient federated learning framework for power systems disturbance and cyber attack discrimination

With the growing concern about the security and privacy of smart grid systems,cyberattacks on critical power grid components,such as state estimation,have proven to be one of the top-priority cyber-related issues and have received significant attention in recent years.However,cyberattack detection in smart grids now faces new challenges,including privacy preservation and decentralized power zones with strategic data owners.To address these technical bottlenecks,this paper proposes a novel Federated Learning-based privacy-preserving and communication-efficient attack detection framework,known as FedDiSC,that enables Discrimination between power System disturbances and Cyberattacks.Specifically,we first propose a Federated Learning approach to enable Supervisory Control and Data Acquisition subsystems of decentralized power grid zones to collaboratively train an attack detection model without sharing sensitive power related data.Secondly,we put forward a representation learning-based Deep Auto-Encoder network to accurately detect power system and cybersecurity anomalies.Lastly,to adapt our proposed framework to the timeliness of real-world cyberattack detection in SGs,we leverage the use of a gradient privacy-preserving quantization scheme known as DP-SIGNSGD to improve its communication efficiency.Extensive simulations of the proposed framework on publicly available Industrial Control Systems datasets demonstrate that the proposed framework can achieve superior detection accuracy while preserving the privacy of sensitive power grid related information.Furthermore,we find that the gradient quantization scheme utilized improves communication efficiency by 40%when compared to a traditional federated learning approach without gradient quantization which suggests suitability in a real-world scenario.

Total Harmonics Distortion Prediction at the Point of Common Coupling of industrial load with the grid using Artificial Neural Network

Power quality challenges have generated a lot of disputes between utilities,customers,network operators,and equipment manufacturers around the world as regards the share of responsibility for power quality solutions,this results in different levels of financial and technical losses for both the network operators and the customers.One of the major consequences of the operation of heavy-duty factories globally is the corruption of power quality at the point of common coupling(PCC).In order to quantify the harmonics contribution at the PCC by industrial consumers,this paper presents three-phase total harmonics distortion of current(THDi)prediction model at the PCC.The proposed artificial neural network(ANN)models use a multilayer perceptron neural network(MLPN)to predict three-phase total harmonic distortion.The input parameter used in the models is easily measured with basic power meters.The model was trained with input parameters captured at 33 kV and 132 kV voltage levels using power quality meters at five(5)different steel manufacturing plants.Eight(8)different models were designed,trained,validated,and tested with different combinations of input parameters,number of hidden layers,and number of neurons in the hidden layer.The results show that the model with two hidden layers which uses four major power parameters(Current,apparent power,reactive and active power)as input parameters in the training model had the best performance with a 95.5%coefficient of correlation between the measured THDi and the predicted THDi.