Historical Load Profile and Temperature Data Pattern Recognition of Input Time Series Autocorrelation Function, Lag Plot and Correlation Coefficients Analyses for Obtaining Relations Among Forecasting Parameters Test Data Training Data Emphasis on maintaining cyber resiliency for each device is due to the devastating effects on reliability of widespread infrastructure, given the potential cascading effect. The attack may be introduced to the grid through single or multiple smart devices and impact the harmonious operation of the entire network which may cause catastrophic grid failure and large blackouts. The existence of numerous smart devices in PEDG, provides an immense attack surface for attackers to target and compromise the system. PEDG lines of defense against cyber-attacks are illustrated in Figure 8. This section discusses the cyber-security challenges and ML-based solutions for realization of IDS and event-triggered control schemes for PEDG. A. Potential Cyber-Physical Attack Models for PEDG Types of attacks that target the smart grid can be classified into different categories. The first is the device-level attacks that aim to compromise and gain control of a device. Device-level attacks are often the initial step of a major attack where a single device is compromised and used as an entry point for the attacker to launch further attacks aiming to compromise rest of the grid. Attackers directing malware or gaining unauthorized access to the control of smart inverters are able to alter and manipulate their control functions, or spoof status information to the utilities. The second type of attacks includes data attacks. These attacks attempt to unlawfully insert, modify, or delete data or control commands in a communication network traffic in order to prompt and mislead the smart grid to make wrong decisions. Privacy attacks can be used to orchestrate a physical attack, such as a burglary, by analyzing electricity consumption data and recognizing a Developing Accurate Regression Model Deep Learning Approach for Updating the Weight Factors of the Developed Model Load Forecasting Procedure Recording Forecasted Results FIG 7 Example of load forecasting via deep learning. Potential Cyberattacks on the PEDG DoS Malware Injection DIA Man-inthe-Middle Phishing Hacking Defense-in-Depth Framework Prevent Intrusion Attack Detected? Yes Detect Intrusions First Line Restrict Implications of PEDG Operation No FIG 8 PEDG lines of defense against cyberattacks. IEEE POWER ELECTRONICS MAGAZINE Maintain Physical Security Detect Intrusions Second Line Secured PEDG Operation 34 Attack Mitigated? No z March 2021 Yes Third Line