Washout-Filter Power-Sharing-Based Resilient Control Strategy for Microgrids Against False Data Injection Attacks

Abstract

Secondary control (SC) under false data injection attacks (FDIAs) in microgrids can compromise control decisions and disrupt the normal operation of the system. This paper proposes a washout-filter power-sharing-based resilient control strategy to tackle FDIAs. This strategy ensures the primary control continues to function normally by enabling the timely disconnection of the attacked SC. To address the under-rated operation state caused by the loss of SC, washout-filter power sharing is activated to restore the rated operation. Furthermore, for the FDIAs that affect both system frequency and voltage simultaneously after power sharing, a voltage compensation control loop is designed for the local voltage drop, allowing the attacked voltage value to further recover to the rated value. This strategy secures a steady frequency and enhanced voltage amplitude in the system, achieving a resilient effect against FDIAs. The proposed strategy has been validated through various simulation scenarios and FPGA-in-the-loop experiments.

1. Introduction

Microgrids are compact and manageable power networks that incorporate distributed generators (DGs), like hydrogen fuel cells, photovoltaic systems, and wind generators. They function efficiently with a hierarchical control structure, including primary, secondary, and tertiary controls [1]. Primary control (PC) typically uses droop control to promptly adjust the output power of DGs, aiming to suppress fluctuations in frequency and voltage. Secondary control (SC) addresses deviations from primary droop power sharing, after restoring frequency and voltage to rated values and obtaining grid synchronization signals to achieve grid integration. Among the various methods, distributed SC with cooperative autonomous regulation has attracted significant research interest [2,3].

In order to enhance multi-energy complementarity between regions, the concept of virtual power plants (VPPs) has been developed to integrate multiple microgrid clusters through a wide-area communications network to form a dispatchable entity with the characteristics of a “power plant”. This entity coordinates the participation of distributed resources in electricity markets and ancillary services at a higher level [4]. Its core value lies in breaking through geographical constraints, integrating microgrid distributed power generation output, energy storage regulation capability and load flexibility, and significantly reducing the uncertainty of new energy output. Reference [5] provides a systematic overview of the core concepts, diversified services, multiple control methods, differentiated optimization strategies, and practical application cases of VPPs and discusses the existing challenges and future trends in this field. Reference [6] proposes an artificial intelligence (AI)-driven integration framework for VPPs and smart grids, fusing deep reinforcement learning, blockchain, transformer, and federated learning to dynamically optimize multi-objectives, address data rigidity and privacy risks, and validate it with a Shenzhen case study that provides a scalable and secure practice path to achieve sustainable energy goals. Reference [4] systematically reviews and categorizes recent research advances on VPP modeling, focusing on its market and physical interactions, and reveals the trend of increasing complexity of VPP frameworks, extensive integration of market mechanisms, and increased energy policy adaptability through a classification framework covering model construction, computational methods, market strategies, and practical feasibility, aiming to identify efficient architectures and point out existing gaps. The aim is to identify efficient architectures and point out existing gaps. Reference [7] reviews VPPs’ development and three-layer structure (aggregation, communication, and core scheduling layers) and highlights the scheduling layer’s role as the “intelligent brain” for resource optimization. It examines current challenges and envisions VPPs’ potential in boosting renewable energy efficiency, energizing energy markets, and advancing low-carbon transitions.

Currently, there are research reports on the smart grid architecture model (SGAM), which provides a systematic framework for integrating VPPs and cybersecurity solutions into grid operations through interoperable mathematical modeling [8]. By combining Internet of Things (IoT)-enabled metering infrastructure with cloud computing and advanced metering infrastructure analytics, this approach enhances grid resilience while addressing cyber–physical threats in demand response scenarios [9]. The SGAM-based methodology offers solutions for optimizing VPP coordination and mitigating security risks. This paper also describes a study on the hierarchical defense capability of microgrids against cyberattacks within this SGAM framework.

The deep integration of power components and information technology makes microgrids, which are the underlying nodes of VPPs, vulnerable to false data injection attacks (FDIAs) [10]. FDIAs are generally meticulously designed to mislead control decisions by injecting false or manipulated data into the control system without being easily detected by the monitoring system [11,12]. Attacks on SC could cause problems like incorrect operation instructions, voltage frequency instability, and energy dispatch disorder, leading to physical equipment damage, abnormal operation, and overall system destruction.

Microgrids are cyber–physical systems, where the SC relies on the cyber layer for information exchange. In recent years, cyberattacks targeting power systems have become increasingly frequent, with their destructive impact growing more pronounced. For example, the cyberattacks on Ukraine’s power grid in 2015 and 2016 caused widespread blackouts, clearly exposing the vulnerability of traditional power systems in terms of cybersecurity [13]. Cyberattackers often exploit open interaction features of SC for FDIAs, tampering with references in power signal transmission. Cybersecurity in renewable energy microgrids is a multi-disciplinary challenge: the energy transition increases system dynamics, and greater intelligence extends attack chains, requiring cyberresilience solutions that integrate power electronics, communications, cryptography, and intelligent algorithms. Reference [14] systematically reviews cyberresilience in renewable smart microgrids, focusing on communication technology vulnerabilities, attack mitigation solutions, and practical recommendations with future research directions. Reference [15] experimentally validates a four-layer IoT architecture (sensing, network, middleware, application) in microgrids, demonstrating its feasibility through remote monitoring. Reference [16] comprehensively analyzes microgrid cybersecurity vulnerabilities and communication protocol risks, identifies limitations in existing protections, proposes layered security strategies, and highlights critical research gaps.

In order to enhance the stability and reliability of microgrids, many studies have concentrated on implementing resilient control to counter this threat [17,18]. A fuzzy resilient control method is proposed in [17] to stabilize systems affected by parameter changes and constant power loads. In [19], a fully distributed attack-resilient adaptive control framework is presented to cope with unknown unbounded attacks, ensuring frequency and voltage stability and consensus. To enhance the anti-interference capability of islanded microgrids against FDIAs, a resilient secondary control method based on unknown input observers is put forward in [20]. An advanced sensor monitoring and decision risk assessment framework, aimed at enhancing microgrid resilience against uncertain risks, is proposed in [21]. Reference [22] proposes an energy storage configuration method based on two-stage robust optimization, considering line faults, to enhance microgrid resilience. A resilient event-driven fault detection filter is introduced in [23], aiming to enhance the fault detection resilience of DC microgrids under denial-of-service (DoS) attacks. In [24], voltage and frequency are regulated in AC microgrids under FDIAs, using a distributed convergent observer-based resilient controller. Reference [25] proposes a resilient distributed secondary control strategy utilizing extended-state observers, which enhances the resilience of islanded AC microgrids. An adaptive fuzzy logic resilient control method is proposed in [26] to ensure the stability of microgrid frequency under the influence of FDIAs and parameter uncertainties. Reference [27] proposes a method based on the W-MSR algorithm, which enhances the resilience of multi-microgrid systems against false data injection and replay attacks by isolating malicious agents. Reference [28] implements resilient control via sliding mode control and offset compensation. A logic processor is designed in [29] to capture attack durations for resilient control implementation. A distributed resilient estimator is designed in [18] to resist topology attacks.

However, when it comes to resilient control against FDIAs, much of the existing literature focuses on droop-based power sharing without explicitly mentioning washout-filter sharing. This method is similar to droop-based power distribution and does not compromise the effectiveness of frequency and voltage control [30,31,32]. This potential is crucial for mitigating FDIAs. In the context of the microgrid hierarchy, SC can be disconnected if necessary. Even after disconnection, PC can maintain stable and reliable operation independently. This analysis demonstrates the feasibility and effectiveness of timely isolating SC when this control is identified as being attacked. Moreover, a method to improve frequency and voltage on the PC layer to demonstrate the resilient effect is identified, as washout-filter power sharing proves highly competent. In addition, due to the local characteristics of the system voltage, such as voltage drop caused by line losses and load distribution, as well as both frequency and voltage being susceptible to FDIAs, the voltage value still cannot return to the rated value after power sharing switching. By analyzing the mechanism of the SC, i.e., the SC algorithm calculates the required frequency or voltage compensation and adds this to the PC to correct any deviations caused by the PC. So, is it also possible to design a control loop on the PC layer to eliminate deviations after the SC is cut off? Thus, this paper further designs a voltage compensation control loop (VCCL) for the local voltage drop, which can further return the voltage to the rated value at the PC layer.

This study investigates the robustness of attacked SC removal and washout-filter-based power sharing against FDIAs, particularly analyzing the role of each DG’s voltage source converter (VSC) in AC microgrids with distributed SC architecture. Following this, the washout-filter power-sharing-based resilient control (WFPS-RC) strategy is proposed. This strategy has proven effective across multiple simulation scenarios. Its validity was further confirmed through FPGA-in-the-loop (FIL) testing, which closely replicates real-world network attack defense conditions. To conclude, the key contributions presented in this paper are as follows:

(1)

The removal of the attacked SC will avoid significant impacts on the PC and other DGs. The system will transition to an operating state with only the PC, and the washout-filter power sharing will replace the droop-based sharing, achieving system frequency and voltage recovery.

(2)

Unlike the global nature of system frequency, local voltage may still not be able to return to the rated value under the effect of washout-filter power sharing. In this case, a VCCL on the PC layer will be designed to further correct the voltage to the rated value. Moreover, the VCCL exhibits the ability to mitigate the propagation of FDIAs.

(3)

The proposed WFPS-RC strategy isolates the attacked SC, restoring frequency and voltage to its rated value at the PC layer. This enhances the system’s overall resilience, ensuring that, even after suffering FDIAs, the system can still maintain its original secondary control effectiveness.

The paper is structured as follows: Section 2 describes primary and secondary controls, Section 3 details FDIAs and the proposed strategy, Section 4 provides verification, and Section 5 presents the concluding remarks.

2. Primary and Secondary Controls

In this section, the configuration of the PC and SC will be provided, where the PC primarily handles droop control and power conversion and the SC then rectifies the frequency and voltage offsets generated by the PC. Among them, the implementation of the SC is the PC plus the compensation amount generated by the distributed control method.

2.1. Droop-Based Power Sharing at the PC Level

The droop-based power sharing control method is commonly used at the PC level. This method balances power between DGs by adjusting the output voltage amplitude and frequency of the converters, ensuring the voltage and frequency stability of the microgrid using the following formula.

$$f=f^{*}-mP$$

(1)

$$E=E^{*}-nQ$$

(2)

where $f^{*}$ and $E^{*}$ are the rated frequency and voltage, while P and Q represent the measured active and reactive power, respectively. The variables m and n denote droop coefficients.

2.2. MPC-Based Power Conversion at the PC Level

Using model predictive control (MPC)-driven power conversion has multiple advantages in the PC of microgrids, including precise modeling, intuitive control objects, and rapid response capabilities [33,34,35]. Droop control offers a voltage reference that includes frequency information for MPC-based power conversion. This reference indicates the necessary voltage level for the system to maintain autonomously. MPC then determines the optimal power conversion control instructions based on this information. This combination forms the PC architecture, and they can leverage their respective advantages. Below is the predictive model of the MPC [36,37].

$$\mathit{x}(k+1)=e^{A{T}_s}\mathit{x}(k)+{\mathbf{A}}^{-1}(e^{A{T}_s}-{\mathit{I}}_{2\times 2})\mathit{B}\mathit{u}(k)$$

(3)

where $\mathit{x}=\left[\begin{array}{l}{V}_c\\ I_f\end{array}\right]$, $\mathit{u}=\left[\begin{array}{l}{V}_i\\ I_o\end{array}\right]$, $\mathit{A}=\left[\begin{array}{cc}0& 1/C\\ -1/L& -R/L\end{array}\right]$, $\mathit{B}=\left[\begin{array}{cc}0& -1/C\\ 1/L& 0\end{array}\right]$, T_s is the sampling interval, and ${\mathit{I}}_{2\times 2}$ represents an identity matrix. In this equation, the matrices are derived from the dynamic model of microgrid interface circuit, which maps the topology of the RLC circuit connected to the DG output inverter, serving functions such as filtering and primarily supporting MPC-based power conversion control.

Meanwhile, the cost function J_VI utilized to determine the optimal switching variable follows this formula.

$$J_{VI}=w_V{(V_{c\alpha ,\beta }^{*}-V_{c\alpha ,\beta }(k+1))}^2+w_I{(I_{f\alpha ,\beta }^{*}-I_{f\alpha ,\beta }(k+1))}^2$$

(4)

where w_V and w_I are weight coefficients. V_c and I_f are the capacitor voltage and filter inductor current, respectively. The ones marked with * are their references. α and β denote the three-phase variables transformed using the Clarke transformation. The cost function solution yields optimal pulse width modulation (PWM) signals for the VSCs.

2.3. Graph-Based Distributed Cooperative Control at the SC Level

The SC of microgrids uses graph theory to show the connection relationships between DGs [38]. It employs a cooperative algorithm that allows intelligent agents to keep system parameters stable without the need for central coordination. Taking DG as a node, if node j transmits data to adjacent node i, node j is defined as a neighbor of i. The set of neighbors for node i is represented as $N_i=\left\{j\right|(v_j,v_i)\}$, with an associated weight of $a_{ij}=1$. In all other instances, $a_{ij}=0$. The adjacency matrix $\mathit{A}={[a_{ij}]}_{N\times N}$, composed of a_ij, outlines the communication pathways for DGs and the reference transmission in the SC.

Distributed control strategies synchronize DGs to improve frequency and voltage regulation. By applying input–output feedback linearization, the complex system is simplified into linear subsystems [36,37].

$${\dot{f}}_i=\mathrm{\Delta }{\dot{f}}_i-m_i{\dot{P}}_i\equiv u_{fi}=e_{fi}{d}_f$$

(5)

$${\dot{E}}_i=\mathrm{\Delta }{\dot{E}}_i-n_i{\dot{Q}}_i\equiv u_{Ei}=e_{vi}{d}_v$$

(6)

where $·$ refers to the differential operation.

Consider the global characteristics of frequency and distribution of active and reactive power. The subsequent formula is applicable.

$$m_i{\dot{P}}_i\equiv u_{pi}=m_i(P_i {\displaystyle \frac{1}{{\omega }_{c}s+1}}-P_i)d_p$$

(7)

$${\dot{Q}}_i={\displaystyle \frac{1}{{\omega }_c}}(Q_i-Q_i {\displaystyle \frac{1}{{\omega }_{c}s+1}})$$

(8)

where ${\omega }_c$ represents the cutoff frequency. The above d_f, d_v, and d_p are the relevant control coefficients.

The tracking errors, denoted as e_fi and e_vi, are formulated as

$$e_{fi}={\displaystyle \sum _{j\in N_i}{a}_{ij}}(f_j-f_i)+g_i(f^{*}-f_i)$$

(9)

$$e_{vi}={\displaystyle \sum _{j\in N_i}{a}_{ij}}(V_{cj}-V_{ci})+g_i(V_c^{*}-V_i)$$

(10)

where g_i is the pinning gain, indicating the DG is a root node when 1 and not when 0.

The protocols for distributed SC cover both frequency and voltage, specifying the compensation added to the PC as

$$f_c=(u_{fi}+u_{pi})\cdot {\displaystyle \frac{1}{s}}$$

(11)

$$E_c=(n_i\stackrel{·}{Q_i}+u_{Ei})\cdot {\displaystyle \frac{1}{s}}$$

(12)

Below is the complete formulation for droop-based power sharing with SC effect compensation.

$$f=f^{*}-mP+f_c$$

(13)

$$E=E^{*}-nQ+E_c$$

(14)

The diagram of the PC and SC structures and connections in one DG as well as the communication topology of the SC layer for four DGs is shown in Figure 1. This figure illustrates that the compensation amount, derived from a distributed secondary control structure, is integrated into the droop controller within the primary control. This integration results in the elimination of steady-state errors for both frequency and voltage. The communication architecture within the secondary distributed system employs a sequential transmission of reference values, originating from DG1 and progressing to DG4. In this architecture, DG1 serves as the root node that can obtain the rated values. The FDIA operates on the channel through which reference values are transmitted, such as the frequency and voltage amplitude f₁ and E₁ given from DG1 to DG2.

3. FDIAs and Proposed WFPS-RC Strategy

In this section, the FDIAs will be first described, then the washout filter’s recovery ability will be introduced, and its formula will be compared with droop control. Following that, the VCCL at the PC layer will also be presented, with its main purpose being to further compensate voltage. Finally, the proposed WFPS-RC strategy, which includes the above two parts, as well as how to specifically implement it, will be given.

3.1. FDIAs on SC

The FDIAs can compromise transmitted references within the network, potentially destabilizing the microgrid system [39,40]. The core idea of FDIAs is to inject a manipulated signal into the original using the following expression.

$$\widehat{d}(t)=d(t)+d_r(t)$$

(15)

$$d_r(t)=\left\{\begin{array}{l}0, for t\notin {\tau }_r\hfill \\ {\lambda }_{r}F(\cdot ), for t\in {\tau }_r\hfill \end{array}\right.$$

(16)

where $\widehat{d}(t)$, d(t), and d_r(t) are the disrupted, original, and tampered signals, respectively. τ_r is the attack duration. F( ) is the main function for attack effect, while λ_r modifies its strength. Regarding SC, d(t) symbolizes the transfer of frequency and voltage references.

The root node typically maintains frequency and voltage references using fixed controller values. For the ith node, the frequency and voltage values from the jth node are assumed to be impacted by FDIAs, outlined as follows.

$${\widehat{e}}_{fi}={\displaystyle \sum _{j\in N_i}{a}_{ij}}({\widehat{f}}_j-f_i)+g_i(f^{*}-f_i)$$

(17)

$${\widehat{e}}_{vi}={\displaystyle \sum _{j\in N_i}{a}_{ij}}({\widehat{V}}_{cj}-V_{ci})+g_i(V_c^{*}-V_i)$$

(18)

The impact of the FDIA-affected system extends from the SC layer to executive levels, posing risks to operational stability and safety.

3.2. Washout-Filter Power Sharing

This method merges droop-based power sharing with SC in the PC layer, preserving the distributed control structure. This can be described as [32]:

$$f_c={\displaystyle \frac{k_{if}}{s}}(f^{*}-f)$$

(19)

$$E_c={\displaystyle \frac{k_{iE}}{s}}(E^{*}-E)$$

(20)

where k_if and k_iE represent the integral coefficients of linear SC and are also the frequency and voltage regulation coefficients of the washout filter. Substitute the f and E defined by Formulas (13) and (14) into the above f_c and E_c, amalgamate similar terms, and solve for the relationship between f_c and P as well as E_c and Q. Then substitute back into Formulas (13) and (14) to derive the following washout-filter-based expressions.

$$f=f^{*}-{\displaystyle \frac{s}{s+k_{if}}}mP$$

(21)

$$E=E^{*}-{\displaystyle \frac{s}{s+k_{iE}}}nQ$$

(22)

The washout filter, a potential replacement for the droop-based method, has reshaped power sharing. Compared to droop control, it can be seen that a first-order high-pass filter transfer function is multiplied before the droop coefficient. In fact, the washout filter functions as this kind of filter, effectively eliminating the DC component while allowing the transient component of the signal to pass through. Its control performance under FDI will be discussed next.

To a certain extent, washout filters can influence the transient response due to their high-pass characteristics. However, they do not affect the steady-state power sharing accuracy because they attenuate DC components, including both FDIA biases and steady-state deviations, to zero. The transient response can be adjusted by appropriately selecting the coefficients k_if and k_iE, enabling the system to preserve normal load dynamics while effectively attenuating relatively slow-varying FDIA signals. It is noteworthy that the proposed strategy exhibits a balanced coordination between enhanced robustness against attacks and the maintenance of power sharing performance.

3.3. VCCL

To accurately adjust the voltage to its rated value, this subsection will introduce the voltage compensation control loop (VCCL) which is implemented at the PC layer. Unlike the system frequency, which has a global characteristic, local voltage may still deviate from its rated value due to the influence of washout-filter power sharing. The SC algorithm determines the necessary frequency or voltage adjustments and incorporates these modifications into the PC to rectify any deviations introduced by the PC. This mechanism serves as a reference for the further recovery of voltage to its rated value after the washout filter is engaged. The following control loop is designed to achieve this task, added to the PC control layer as compensation.

$$\left\{\begin{array}{l}{e}_{Ecl}=\left|V_c-E\right|\hfill \\ E_{cl}=\left(k_{pEcl}+k_{iEcl}\cdot {\displaystyle \frac{1}{s}}\right)\cdot e_{Ecl}\hfill \end{array}\right.$$

(23)

This voltage compensation can be used to address strong FDIAs on the system, such as simultaneous attacks on frequency and voltage, to compensate for the inability of the washout filter to achieve rated value recovery. Given that the VCCL utilizes an integral controller, a limiting module must be incorporated at the controller output to prevent integral saturation. The limiting module serves as a critical component for maintaining system stability in the presence of FDIAs, particularly during the transient phase before the defense algorithm is fully activated. By constraining the impact of compromised signals, it helps prevent the propagation of malicious data and enhances the system’s resilience under attack conditions.

3.4. Proposed WFPS-RC Strategy

To fully utilize the washout-filter features and achieve a resilient control effect under the FDIA, the following protocol will be designed.

$$f=f^{*}-g_{s1}mP-g_{s2}{\displaystyle \frac{s}{s+k_{if}} }mP$$

(24)

$$E=E^{*}-g_{s3}nQ-g_{s4}{\displaystyle \frac{s}{s+k_{iE}} }nQ+E_{cl}$$

(25)

where g_s1, g_s2 and g_s3, g_s4 represent two sets of control gains, specifically associated with frequency and voltage compensations, respectively.

The proposed WFPS-RC strategy contains two parts. The first part is to cut off SC, meaning the compensation produced by distributed SC is set to zero. The second part is to switch the power sharing mode in the PC layer. Therefore, the following rules for triggering and assignment are hereby established.

$$\left\{\begin{array}{l}{f}_c=E_c=1, t\in {{\rm T}}_1\hfill \\ g_{s1}=g_{s3}=1, g_{s2}=g_{s4}=0, t\in {{\rm T}}_2\hfill \end{array}\right. : \mathbf{if} \mathrm{FDIA} \mathrm{is} false$$

(26)

$$\left\{\begin{array}{l}{f}_c=E_c=0, t\in {{\rm T}}_3\hfill \\ g_{s1}=g_{s3}=0, g_{s2}=g_{s4}=1, t\in {{\rm T}}_4\hfill \end{array}\right. : \mathbf{if} \mathrm{FDIA} \mathrm{is} true$$

(27)

The time intervals T₁ and T₂ represent periods without FDIA, whereas T₃ and T₄ correspond to periods during which FDIA is present. To flexibly handle network attacks, T₃ and T₄ can commence simultaneously. Alternatively, T₃ can initiate first to disconnect the under-attack SC, followed by switching the power sharing in T₄ to achieve resilient effects. Here, to ensure the integrity of the system power sharing and power conversion control architecture between different operations, while maintaining simple control logic, the deactivation of SC and activation of the washout filter are configured to occur simultaneously. This simultaneous operation is achieved through synchronized control logic.

The proposed WFPS-RC strategy is shown in Figure 2a. The figure illustrates that, when a DG experiences the FDIA, the proposed control strategy initiates the appropriate control algorithm. This process involves transitioning from droop power sharing to a washout filter and implementing a VCCL for voltage stabilization. Subsequently, MPC-based power conversion is used to generate the required PWM control signals for the VSC. Figure 2b shows the FIL platform constructed in this paper, with a USB Blaster cable connecting a computer to an FPGA board. The FPGA will be manually operated to generate FDIAs, then the produced attacks are injected into the simulation model. The WFPS-RC strategy proposed in this paper is designed to implement the defense and elimination of external network attacks within the model. This FIL configuration, leveraging the FPGA’s hardware acceleration capabilities to generate network attack signals in real time, more closely resembles real-world network attacks. This approach makes the attack vector more explicitly transmitted through communication lines into the system model. Meanwhile, the proposed algorithm operates at the control level to defend against and eliminate incoming attacks.

4. Verification

In this section, a microgrid system of four DGs is constructed. Here, the algorithm development is carried out on a host computer (Intel Core i7 CPU, 40 GB RAM, Windows 11) using MATLAB/Simulink R2022b. The reference value transmission paths of these four DGs are as described earlier. In the following simulation scenarios, the frequency and voltage reference values received by DG2 from DG1 will be first considered to have been subjected to an FDIA. The system operates at a rated frequency of 50 Hz and the root mean square value of the line rated voltage is 380 V. The frequency f and voltage V_c are both collected from the filter output of DG VSC. In the following validation scenarios, the VCCL is equipped on DG2 to DG4, while DG1 is not configured for comparison. To distinctly demonstrate the compensation effect of the VCCL in the proposed strategy, its control performance is showcased starting from scenario 4.3.

4.1. Situation of DG2 Suffering from FDIA

First, the FDIA will increase the frequency reference transmitted from DG1 to DG2 by 100 units at 1 s, as displayed in Figure 3. Figure 3a shows the attacked effect without any defense algorithms. DG3 and DG4, both located behind DG2, lose control under the attack. Although DG1 is not directly targeted and continues to operate, it still experiences side effects. These include fluctuations in frequency and gradual oscillations in voltage amplitude. As shown in Figure 3b, DG2 initiates the soft disconnection (SC disconnection) at 1 s using the proposed WFPS-RC strategy. Subsequently, at 4 s, the droop control mode transitions to a washout-filter-based power sharing method. All subsequent simulations presented in this study follow this predefined sequence of control actions, ensuring consistency in evaluating the system’s dynamic response and stability. After SC is cut off, system frequency gradually returns to the rated value due to its global nature and local recovery ability. However, the system frequency returns back to the rated value without steady-state error after 4 s, owing to the washout filter. At 1 s, V_c decreases but can eventually recover to the rated state.

Next, the FDIA will increase the voltage reference transmitted from DG1 to DG2 by 100 units, and the result is depicted in Figure 4. In Figure 4a, the voltage suffers the most direct attack, with its amplitude experiencing a steady deviation, but it remains stable after more than 1 s. Meanwhile, the frequency is also affected, although it did not receive a direct attack; the frequency undergoes a transient decrease for more than 1 s, after which it returns to its rated value. Figure 4b shows that the frequency is almost unaffected by using the proposed WFPS-RC strategy. Despite a drop after 1 s, the voltage recovers to its rated value by 4 s, ensuring stable operation. At the same time, to distinguish the control effect of the VCCL, V_c will be presented in two separate diagrams.

Then, the FDIA will further increase 100 units each in the frequency and voltage references transmitted from DG1 to DG2. The results are displayed in Figure 5. It reveals that the FDIA’s comprehensive impact on the system has increased in comparison to previous attacks that only targeted frequency and voltage, as evidenced by more intense fluctuations in frequency and voltage. However, fortunately with the help of the proposed WFPS-RC strategy, the system can restore frequency to the rated value and achieve some voltage compensations close to the rated value. Although the voltage stabilizes below rated values after 4 s, the system mitigates the attack impact to an acceptable range (significantly reduced from a maximum of about 1.86 V when under attack to about 0.25 V after using the proposed strategy). Also, no major voltage fluctuations occur when the system is stable. This frequency recovery and voltage compensation brought about by the switching of power sharing methods brings the system’s operating state closer to its rated state, which demonstrates the effect of resilient control.

4.2. Situation of DG2 to DG4 Suffering from FDIA

To further verify the effectiveness of the proposed WFPS-RC strategy, the FDIA will simultaneously and suddenly increase the frequency and voltage references by 100 units from the upper reference signal received by DG2 all the way to that received by DG4. Figure 6 compares the relevant results. In Figure 6a, it is shown that the attack’s impact on the entire system has increased significantly compared to the case where only a single DG receiving channel is subjected to an FDIA. Nevertheless, by using the proposed strategy, the system maintains its capacity to effectively mitigate this impact. As shown in Figure 6b, the system successfully restores the frequency to its rated value within approximately 5.5 s. Meanwhile, the voltage remains stably regulated near the nominal level, demonstrating the effectiveness of the control strategy in maintaining both frequency and voltage stability under the given operating conditions.

4.3. VCCL Effect

From the previous simulation results, it can be observed that, when the system is subjected to relatively severe FDIAs, such as when one or multiple DGs’ frequency and voltage are affected, using the power sharing method in the proposed strategy is not sufficient to restore the voltage to its rated value without deviation. Therefore, benefiting from the proposed VCCL method, the system voltage will receive further compensation, bringing it to its rated value. Figure 7 compares results when the VCCL activates at 7 s, based on previous simulations. From Figure 7, it can be observed that, even though the power sharing was switched to the washout-filter method, the voltage remained stable in the period of 4–7 s but continued to operate below the rated level. Subsequently, after 7 s, the voltage profile exhibits a further increase, stabilizing at the rated value. Regardless of whether one DG or multiple DGs experience FDIAs, the proposed VCCL can effectively restore both frequency and voltage to their rated values.

4.4. System Performance After the SC Delay Cutoff

In the previous simulation, facing the FDIA, the immediate removal of SC was first simulated. However, in some cases, there might be a delay in the removal of SC, so it is crucial to assess the system’s performance under conditions where there is a delay in SC cutoff. Therefore, in this scenario, not removing SC at the moment of attack but rather allowing the attack to continue for a short period, followed by the removal of SC and the implementation of the proposed strategy, will be evaluated. The comparison of the system’s control performance and effectiveness is shown in Figure 8. Here, the FDIA initiates at 1 s and persists for 1 s; the system’s performance without and with the VCCL is presented. As can be seen from Figure 8, when using the VCCL method, the system maintains control during the ongoing attack, in contrast to the situation without VCCL implementation. This is due to the limiting module in the VCCL. When the system reconnects to the proposed complete strategy after 2 s, it can quickly recover to its previous rated value. Thus, the VCCL can also incorporate functionality to mitigate the propagation of attack effects.

4.5. Verification of Random Attacks by FIL Experiments

With the FIL setup in Figure 2b, the FPGA will randomly implement an FDIA, where SC removal and power sharing switch occur simultaneously. The most severe attack begins, with four DGs concurrently subjected to FDIA, including modifications to the constant reference acquired by DG1. DG1 is not equipped with a VCCL, while the other three DGs are all configured with it. As shown in Figure 9, the attack starts just after 2 s and lasts for a very short time. This causes fluctuations in the system frequency and voltage under the proposed WFPS-RC strategy, but the system is still able to operate stably. Specifically, the FDIA temporarily decreases DG1’s frequency and then it regains rated value, while other DGs’ frequencies remain stable. Additionally, the FDIA causes a steady decrease of about 0.5 V in DG1’s voltage and slight drops in other DGs’ voltages. The FIL experiment here further demonstrates the effectiveness of the proposed WFPS-RC strategy.

5. Conclusions

From the perspective of designing control algorithms within the controller, this study proposes an effective WFPS-RC defense strategy against FDIAs that affects the SC layer of microgrids. The strategy involves rapid disconnection of the attacked control layer, thereby preventing the cascading spread of security threats to the PC layer. By isolating the compromised component, the strategy ensures the continued stable operation of unaffected DGs and contributes to the robustness and fault tolerance of the overall microgrid control architecture. It adopts a power sharing mechanism using washout filters. This allows for frequency and voltage compensation, making sure the system can swiftly return to its standard operating state. Furthermore, in order to compensate for the residual voltage deviations not addressed by the washout filter, the proposed VCCL effectively mitigates these errors. This enhancement ensures that the system maintains operation close to the rated voltage level, improving the regulation accuracy at the nominal operating point. The VCCL also helps stop FDIAs from spreading, especially during the initial phase when the defense algorithm has not yet fully activated in response to an ongoing attack. Thus, the stability and resilience of the microgrid are maintained. This WFPS-RC strategy has proven to be effective with various simulation scenarios and is robust enough to handle physical external injection attacks via the FIL.

Although the WFPS-RC strategy proposed in this study effectively defends against FDIAs on the SC layer, future work will focus on more complex attack vectors and system topologies. First, the WFPS-RC framework will be extended to address a broader range of attack types beyond FDIAs, such as DoS attacks, and its robustness will be evaluated under multi-layer coordinated threats. Second, the resilient control mechanism will be further developed to enhance its adaptability in advanced operational scenarios. For instance, in systems assisted by machine learning, online learning techniques will be integrated to dynamically adjust control parameters, enabling the system to respond effectively to unknown or evolving attack patterns.