A Sustainable and Resilient Distribution System Restoration Framework Based on Intentional Islanding and Blockchain-Based P2P Insurance

Abstract

Extreme weather events have raised the frequency of power outages, posing critical challenges to the sustainability and resilience of modern power systems. In such cases, distributed energy resources (DERs) can effectively support the re-establishment of sustainable power supply for critical loads within the distribution network and reduce power outage losses. In this paper, a sustainable fault recovery framework based on an intentional islanding scheme is proposed to partition the distribution system in order to optimize the priority restoration of critical loads, while taking the operational constraints of the system into consideration. In addition, a blockchain-based P2P insurance mechanism is applied to mitigate the outage losses of the network’s users with a higher degree of security and transparency. By linking technical restoration decisions with financial risk-sharing mechanisms, the proposed framework improves economic sustainability and social equity among network users. For this purpose, a multi-layer, multi-objective optimization algorithm is proposed for optimal partitioning of the distribution network, management of DERs, and demand side management of flexible loads in order to minimize the outage losses and the insurance premium, while maintaining satisfactory performance of the network. To validate the feasibility of the proposed algorithm, the 45-node distribution network of Alexandria, Egypt is used. The results show that a reduction in peak load, outage losses, and operational costs are achieved, with an overall saving of 17.34%, in addition to a premium reduction of 41.3%. These results highlight the effectiveness of the proposed framework in enhancing the environmental, economic, and operational sustainability of distribution systems under outage conditions.

1. Introduction

With the widespread adoption of distributed generation (DG) and its penetration into the power system, power systems transferred from being centralized systems to more sustainable, decentralized ones. As a result, a power system can be split into several isolated subsystems or islands following a catastrophe to avoid blackouts. These controlled islands can improve the continuity of supply to local loads which improves the sustainability and reliability of the power system.

1.1. Motivation

During intentional islanding, the distribution system is divided into a number of decentralized microgrids, which are kept stable with the help of managing the DERs and energy storage systems in each island along with controlling the load demand through the application of demand side management (DSM) programs. In urban distribution networks, electric vehicles (EVs) play an important role as distributed and flexible energy storage systems when connected to the grid. The main objectives of optimally managing the DERs and loads are to restore power supply to critical loads, decrease the amount of unserved load, and decrease the outage duration, and hence, the outage losses.

1.2. Literature Review

Many studies have addressed the optimal intentional islanding operation and DG benefits during load restoration. In [1], a fault recovery approach for a distribution network was proposed considering the charging/discharging of EVs as well as the network reconfiguration. In this approach, an island partition model was developed for optimizing the priority restoration of important loads. However, the presented approach did not take the outage loss or load control through DSM programs into consideration.

Similarly, to maximize the demand’s supply, an islanding strategy was proposed in [2] considering mobile batteries placed on trucks. The interaction between the location of such mobile batteries and load switching was addressed to reach the research’s objective. In addition, the network loads were prioritized. However, in this work the loads with less priority were completely disconnected. Shifting these loads in response to available power supply by applying DSM was not considered.

An approach to find the optimal configuration of islanded microgrids in the presence of renewable energy resources was presented by the authors of [3]. The objective of this approach was to improve system resilience through a reduction in power loss. However, the proposed algorithm did not consider the continuity of the supply of critical loads or the outage losses.

In [4], the authors proposed a fault recovery strategy with the help of DG through a coordinated control of island partitioning to shorten the outage duration and minimize economic loss. However, the operational conditions of the distribution system regarding the power loss and voltage deviations were not considered. In addition, the priority distribution of the loads was not addressed.

In order to provide emergency power to restore critical loads, EVs were considered as distributed mobile energy storage units which could be dispatched to charging stations [5]. The main objective was to optimally locate the charging stations, taking into consideration the distance of traffic networks in addition to the operational conditions of the distribution network. However, the idea of dispatching EVs from their regular parking stations adds time limitations to the load restoration process, unlike making use of the stored energy of static EVs without dispatching them away from their parking location.

The authors of [6] presented a load restoration strategy incorporating both static and mobile energy storage systems. In addition, an optimal load pickup sequence was introduced, taking into consideration the priorities of the loads. The objective of the research was to maximize the total restored priority weighted power of the loads during outages. However, partitioning the distribution system into islands was not considered to lead to higher power loss.

On the other hand, in [7], a fault recovery strategy was introduced which took into consideration the presence of DERs. In addition, an island partitioning scheme was proposed which considered the important load level during the island partitioning process. The objectives were to reduce active power losses and minimize the number of switching actions. However, the load level considered represented the number of restored loads without taking the priority of these loads into consideration.

Nevertheless, there still remains a gap in the research concerning the reduction in both the curtailed loads and the outage losses during outage periods through the application of demand side management programs on flexible loads which take into consideration the loads’ priorities, while splitting the distribution network into islands depending on the presence of DERs.

Furthermore, power outages and load curtailment can cause direct economic losses (such as household labor losses) and indirect economic losses (such as production delay losses) [8,9]. To mitigate outage risks, many studies proposed outage insurance mechanisms in order to compensate users’ losses.

In [10], a post-disaster insurance mechanism was proposed, in which the distribution network losses due to disasters were claimed by insurance. The claimed insurance can provide financial support during disaster recovery.

An outage insurance mechanism was proposed in [11], such that a customer receives outage compensation based on the outage value. The distribution company signs insurance contracts with the customers defining the insurance premiums. The distribution company uses these premiums to reimburse consumers according to their outage value when the electricity outage occurred.

Similarly, a reliability insurance scheme was introduced in [12], in which electricity consumers are capable of determining their desired reliability levels and hence paying the corresponding premiums to the distribution system operator (DSO). The DSO can use the premiums to reimburse consumers according to their outage and reliability levels.

The concept of reliability insurance contracts was introduced in [13]. According to these contracts, investment incentives were calculated with respect to customer damage function. In the presence of these contracts, the revenue opportunities for distributed generation were also evaluated.

In [14], a study of the insurance effect associated with DERs with respect to distribution network reliability was presented. The relationship between the failure risk of DERs and the investment decisions made by consumers was investigated.

However, the outage insurance presented in the above-mentioned studies [10,11,12,13,14] is a direct insurance that depends on a direct contract between the insurer and the customer with insufficient participation rate. In addition, risk distribution is centralized with a lack of cost sharing and control.

On the contrary, the peer-to-peer (P2P) insurance mechanism is a decentralized mechanism with a higher participation rate and interaction between users (peers). With the help of Internet platforms, users can exchange information more easily, pool their resources to compensate each other for losses, and reduce the cost of insurance [15,16]. In such a mechanism, users are aggregated through a P2P network into mutual aid groups. Participants pay part of the total premium to the insurance company as the insured amount, and the remaining part is kept as a fund pool for the group [17]. In addition, P2P insurance aims to reduce costs as it cuts out some of the middle-level expenses, relying more on a digital platform and the self-organizing nature of the peer group [18].

P2P insurance was extended to the power industry. For example, authors in [18] proposed a P2P outage insurance mechanism for residential users based on a distribution network reliability assessment to achieve effective sharing and a reduction in outage risk. In [19], a risk management framework combining parametric insurance and peer-to-peer (P2P) risk sharing to address production uncertainty in solar electricity generation was proposed. In that framework, a complementary P2P mechanism was introduced that redistributed the remaining risk among participants.

However, authors in [18,19] assumed that policyholders belonged to the same community, grouping them together on a digital platform which relies on social bond to overcome any element of distrust. Nevertheless, with the help of emerging blockchain technology, peers of different social communities can form a mutual aid group. Blockchain technology offers a notable degree of security and transparency through the use of distributed ledgers, suitable for the exchange of payments and claims of the P2P insurance system.

Regarding energy management optimization, authors in [20] presented a hybrid metaheuristic (DE-HHO) combining differential evolution (DE) and Harris Hawks optimization (HHO) that enhances global exploration and local exploitation for microgrid energy management, achieving notable improvements in cost reduction and convergence speed. Although advanced hybrid metaheuristics can improve optimization performance, the research primarily focused on single-layer energy management problems. In contrast, the proposed work introduces a multi-layer restoration-oriented framework that jointly considers islanding, DSM/EV scheduling, and DER dispatch, in addition to integrating economic resilience mechanisms (blockchain-based P2P insurance). This establishes a clear distinction in both problem formulation and system scope, rather than only optimization techniques.

In addition, the study presented in [21] emphasizes the importance of integrating sustainability and resilience into energy-related decision-making. Motivated by that work, the proposed framework in this paper particularly addresses the integration of resilience metrics, DER utilization, and economic risk mitigation mechanisms. While the work in [21] focuses on macro-level corporate resilience, the proposed study contributes at the system-operation level by quantitatively embedding resilience through outage management, insurance mechanisms, and distributed energy coordination.

1.3. Contributions

In this paper, a multi-layer multi-objective individual-based optimization algorithm is proposed for optimal restoration of the distribution system after a loss of connection to the upstream network. In the first layer of the proposed algorithm, partitioning the distribution network into islands is considered. In the second layer, depending on the islands formed, demand side management (DSM) of flexible loads in addition to scheduling of EV charging/discharging are considered. Finally, in the third layer, the optimal dispatch of diesel generators and combined heat and power (CHP) units is considered. The objective of the proposed algorithm is to restore as much power as possible to the network loads depending on their priority. By maximizing the restored power to high priority loads, the outage losses of the system are minimized. Furthermore, to compensate the customers for their losses, a blockchain-based P2P insurance mechanism is proposed. With the minimization of outage losses taking into consideration the loads’ priorities, the premium collected is also minimized. In addition, the application of the proposed blockchain-based P2P insurance scheme further reduces the cost of insurance, while maintaining secure and transparent financial transactions.

The main contributions of this work are summarized as follows:

• Development of a multi-layer, multi-objective optimization framework for coordinated islanding based on DER availability, DSM, EV scheduling, and DG dispatch.
• Incorporation of DSM of flexible loads during outage conditions, depending on load shifting instead of complete disconnection, which improves overall load restoration and reduces unnecessary load shedding.
• Implementation of a priority-based restoration strategy to ensure that critical loads are continuously supplied through optimal management of higher-priority loads.
• Efficient utilization of EVs as stationary distributed storage units through smart charging/discharging scheduling, which overcomes delays resulting from the dispatch to charging stations and improves their response during outages.
• The proposed multi-objective algorithm explicitly considers the economic impacts of outage losses by maximizing restored power to high-priority loads and applying DSM, which many prior works have overlooked.
• A blockchain-based P2P outage insurance scheme is introduced, enabling cost-effective and decentralized risk sharing among users, in addition to transparent and secure financial transactions.
• The coupling between technical outage restoration and financial risk mitigation, which reduces both the outage losses and the insurance premiums.
• Focus on resilience enhancement during outage conditions, rather than only cost or emission optimization.

Table 1. Mapping of research gaps to proposed contributions in this paper.

Research Gap from the Literature	Limitations of Prior Works	Contributions of This Paper	Gap Addressed in This Paper
Lack of integration of DSM in restoration process.	Low-priority loads were disconnected without flexibility consideration [1,2].	Consideration of DSM of flexible loads.	Enables load shifting instead of shedding, which increases restored energy.
Lack of consideration of load priorities.	Load importance ignored or partially considered in [3,7].	Implementation of a priority-based load restoration strategy.	Enables continuous supply to critical loads.
No unified algorithm combining islanding, DSM, EVs, and DG dispatch.	Previous works considered these aspects independently: islanding in [3,4], DSM rarely considered in [1], EVs in [5], storage in [6].	Proposal of a multi-layer multi-objective optimization algorithm.	Provides coordinated consideration of network topology, demand, and generation levels.
Delayed use of EVs due to dispatch to charging stations.	Dispatch of EVs to charging stations causes delays in [5].	Use of EVs as stationary distributed storage (V2G/G2V).	Eliminates dispatch delays, hence improving time of response during outages.
Lack of explicitly considering outage losses in optimal restoration.	Restoration strategies neglect economic outage losses in [1,2,3,4,5,6,7].	Inclusion of outage loss minimization in the multi-objective function.	Reduces both direct and indirect economic outage losses.
Centralized outage insurance schemes.	Traditional insurance models in [10,11,12,13,14] rely on centralized structures with limited participation.	Introduction of a blockchain-based P2P insurance mechanism.	Enables decentralized risk sharing, transparency, and higher participation.
Lack of integration of technical restoration and financial compensation in a unified algorithm.	Restoration [1,2,3,4,5,6,7] and insurance models [10,11,12,13,14] are treated separately.	Integrated optimization of restoration process and insurance premium calculations.	Reduces insurance premiums by minimizing outage losses.
Limitations of trust and scalability in P2P insurance due to social constraints	P2P models assume same-community participation in [18,19].	Use of blockchain technology for a P2P insurance mechanism.	Ensures secure, transparent, and trustless participation.

summarizes how the proposed framework not only addresses individual limitations of prior works [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19] but also provides a unified techno-economic solution that jointly optimizes system restoration and outage risk mitigation.

2. Problem Description

In order to reduce outage losses, intentional islanding of the distribution network is considered with the help of DERs penetrating the system. Therefore, optimal partitioning of the distribution network along with optimal management of these DERs are considered. For further reduction in load shedding and hence reduction in outage losses, DSM of flexible loads along with prioritization of critical loads are considered. Then, to compensate customers for outage losses, an insurance system is considered. Finally, to reduce the cost of insurance and maintain secure financial transactions, a blockchain-based P2P insurance mechanism is adopted. In the following, a description of the proposed optimization algorithm and insurance mechanism is provided.

2.1. Multi-Layer Individual-Based Algorithm

In the proposed optimization algorithm, each individual represents a candidate solution for the restoration problem and is composed of three sequentially dependent layers, as illustrated in Figure 1. Each layer contains a set of decision variables that are optimally selected to address a specific stage of the restoration process. The first layer determines the intentional islanding configuration by assigning buses, lines, and distributed energy resources (DERs) to each formed island. Based on the island structure obtained from the first layer, the second layer optimizes demand side management (DSM) actions for flexible loads as well as electric vehicle (EV) charging/discharging schedules. The modified load demand resulting from this second layer then defines the operational limits and dispatch requirements for controllable generation resources in the third layer. Finally, the third layer determines the optimal dispatch levels of diesel generators and combined heat and power (CHP) units. Therefore, the decision variables selected in each layer directly define the feasible search space and operational boundaries of the subsequent layer, creating a hierarchical and interdependent optimization structure [22]. This type of layering is suitable for any of the individual-based metaheuristic algorithms.

The layers of each individual of the proposed algorithm are as follows:

• Intentional islanding layer: The first layer is responsible for the optimal partitioning of the distribution network into islands. The parameters of this layer represent the lines and buses included in each island.
• DSM layer: The second layer is responsible for the optimal load DSM of flexible loads, in addition to optimal scheduling of EV charging/discharging in each of the islands formed in the first layer. A load shifting-based DSM is applied for peak clipping in order to reduce the amount of load shedding (and, hence, outage losses), in addition to the cost of the energy consumption of different loads. Therefore, the parameters to be selected in this layer are the amount of load to be shifted, the time of shift, the start time and duration of EV charging, and the start time and duration of EV discharge. As an outcome of this layer, the required load demand of each island is determined.
• Dispatch layer: The third layer is responsible for the optimal dispatch of diesel generators and CHPs, depending on the demand of each island determined in the second layer. Taking the supply/load balance in addition to load priority into consideration, the amount of load shedding is determined.

These layers are arranged in order to serve the multiple objectives of the optimization algorithm. The first objective considered is the minimization of power losses of the distribution network by supplying the loads locally using nearby DERs. This objective is served through optimal partitioning the network into islands (first layer). The second objective is to minimize the outage losses. That objective is targeted first by reducing the peak load through optimal load shifting-based DSM and optimal scheduling of EV charge/discharge (second layer), then by optimal prioritized load shedding while maintaining the power balance of the islands through optimal dispatch of generating units (third layer).

In addition to the two objectives previously stated, the operational costs of the system are considered. These costs represent the energy consumption costs of different loads, and the cost of energy generated by different DERs.

2.2. P2P Insurance Mechanism

The P2P insurance mechanism depends on two aspects: mutual aid groups and premium return. Participants (peers) form mutual insurance groups through online platforms. Each group contributes to a funding pool, part of which is paid to the insurer (in this case, the system operator, SO) as premiums. The other part of that fund is jointly owned by peers and is used for small claims. At the end of the insurance period, the remaining jointly owned funds in the pool are returned to all participants in the group according to a certain proportion, which indirectly reduces the premium.

In this paper, the P2P insurance mechanism is applied to power outage compensation. As most power networks are highly reliable, the outage probability is accordingly low. Therefore, in most cases, the application of the P2P insurance mechanism allows for the return of funds as the insurance probability is low.

As shown in Figure 2, in the P2P insurance mechanism, the SO acts as an insurer. The SO is responsible for policy design, insurance operation and network reliability assessment upon which the outage insurance is determined. The P2P insurance intermediary is responsible for collecting the insurance fees from the members of the mutual aid group to form a pool of funds. The funds collected are divided into two parts. The first part (m% of the fund) is paid as a premium to the SO, and the second part ($1-m$% of the fund) is managed by the P2P insurance intermediary as mutual funds for small claims and refunds.

When an outage occurs, depending on a threshold Mr, large losses costs (≥Mr) are paid directly by the SO. Small losses costs (<Mr) are paid by the P2P insurance intermediary based on the mutual fund pool. At the end of the insurance period, a certain proportion of the remaining mutual funds in the pool are returned to participants. The whole P2P insurance scheme forms a type of chain reinsurance, in which the SO acts as the insurer, the P2P insurance intermediary acts as the reinsurer, and the users act as the policyholders [18].

2.3. Blockchain-Based P2P Insurance Framework

In previous research addressing the P2P insurance mechanism [15,16,17,18,19], users forming a mutual aid group were assumed to have a social bond between them on a community basis for trusted financial transactions. However, that condition limits the rate of participation in the P2P insurance framework. To overcome this issue, a blockchain-based P2P insurance framework is proposed in this paper.

Blockchain is a distributed database or ledger of an interconnected network of users. Through this distributed ledger, every user will have the same copy of the data in a decentralized platform. Therefore, blockchain allows direct P2P transactions between users without the need for a third party. That is why it is called a ‘trustless’ technology [23,24].

Transactions are carried out following specific rules guiding their reason, amount, and involved users. Therefore, blockchain technology also uses smart contracts, consensus mechanisms and cryptographic security [25]. The smart contracts are mainly computer codes that are responsible for specifying the transaction rules and are saved in the distributed ledger of the blockchain. Miners are responsible for executing smart contracts following a specific consensus in order to validate transactions and hence add blocks to the blockchain, as shown in Figure 3. Once a smart contract is created and loaded into the blockchain, it cannot be changed. Smart contracts are created by sending a contract creation transaction to the blockchain network. The new smart contract is added to the blockchain after the network of miners verify it and reach a consensus [26].

In the proposed blockchain-based P2P insurance system, the smart contract of a mutual aid group specifies the insurance fees of each participant, the ratio of the premium to the mutual fund, and the threshold of the outage loss costs. Once an outage occurs, a transaction request is issued. The outage losses are calculated, and the claim is determined through the execution and validation of the smart contract. After validation, a block is created and added to the blockchain, and the transaction is confirmed. In addition, if a user decides to quit the mutual aid group halfway through, their premium will be partially refunded based on a premium sharing scheme after deducing a certain liquidated damage fee that will be paid to the P2P insurance intermediary.

Transaction validation and block creation mainly pose mathematical problems to be solved following a distributed “trustless” consensus. Blockchain technology has different types of consensuses such as Proof of Authority (PoA), Proof of Stake (PoS) and Proof of Work (PoW). In each of these consensuses, miners are responsible for solving this mathematical problem, i.e., mining. In PoW, all participants are allowed to perform the mining process. The first participant who can solve the problem is considered the miner. However, this type of mining consumes a large amount of energy, in addition to the possibility of one participant of large mining power possessing the ability to hijack the system and compromise the integrity of transactions. On the other hand, in PoS, the mining rights are permissioned to a number of participants based on their percentage stake in the network [27]. In The P2P insurance system, these participants could be the ones with higher loads or load priorities.

However, in this paper, the third consensus mechanism, namely Proof of Authority (PoA), is adopted to implement a permissioned blockchain. In this approach, transaction validation is not performed by a single entity, but rather by a set of pre-approved and trusted validator nodes within the network. PoA is a reputation-based consensus mechanism widely used in private and consortium blockchains, where authorized validators collectively verify transactions and generate new blocks. This structure provides high throughput, low latency, and energy efficiency, making it suitable for enterprise-level energy management and financial applications.

In the proposed framework, a subset of participants in the mutual insurance group are designated as authority nodes, and the role of block proposer is periodically rotated among these validators. This rotation mechanism, combined with distributed validation, prevents any single node from having persistent control over the system and mitigates the risk of manipulation or single-point-of-failure. Additionally, all transactions are recorded on a transparent and tamper-evident ledger, ensuring accountability and enabling verification by other authorized participants.

Open source or special blockchain platforms could be used to implement the blockchain network. Ethereum is one of the open source platforms in which Cloud service such as Azure is used to host the private blockchain network. A PoA blockchain network can be set up using Ethereum’s Clique consensus engine, in which Solidity is the smart contract language [28]. In this paper, customized blockchain platforms are adopted.

It is important to note that the proposed blockchain-based peer-to-peer (P2P) insurance mechanism is implemented as an economic evaluation and compensation layer that operates alongside the physical restoration framework. The insurance model does not directly influence the operational decision variables of the restoration process, such as island formation, load scheduling, or DER dispatch. Instead, it utilizes the outputs of the restoration model—specifically outage duration and unserved energy—to quantify compensation among participants.

2.4. Practical Implementation Considerations

The implementation of the proposed framework in real distribution systems requires coordination among multiple components, including islanding control, demand side management (DSM), electric vehicle (EV) scheduling, distributed energy resource (DER) dispatch, and the insurance platform. This coordination can be achieved through a hierarchical control architecture integrated within existing distribution management systems. In such a structure, a central controller usually performs day-ahead optimization, while local controllers associated with DERs, EVs, and flexible loads execute control actions based on received setpoints.

The communication requirements of the proposed framework are moderate, as the exchanged data mainly include load demand, DER generation capacity, EV state of charge, and control signals. These data exchanges can be supported by existing advanced metering infrastructure and standard communication technologies. Moreover, since the proposed framework operates on a planning basis rather than real-time control, communication latency constraints are not stringent.

Although the integration of multiple components increases coordination complexity, the proposed multi-layer structure decomposes the overall problem into smaller, manageable sub-problems, facilitating implementation and scalability. Furthermore, the blockchain-based P2P insurance mechanism operates independently from the real-time control layer, thereby avoiding additional burden on system operation.

Although no explicit behavioral feedback from insurance payouts to restoration decisions is modeled, an indirect coupling exists through the objective function, where minimizing the outage losses cost simultaneously reduces compensation costs. This ensures that economically efficient restoration strategies are inherently favored.

3. Problem Formulation

3.1. Objective Function

The overall objective function of the proposed multi-layer multi-objective optimization algorithm is a cost function, which is to be minimized. This cost function is composed of the cost of outage losses, the cost of energy consumption of different loads, the cost of energy supplied by dispatchable DERs in addition to EVs, and the cost of energy loss for all islands. As stated earlier, these objectives are accomplished through optimal islanding, optimal dispatch of generating units, optimal scheduling of EV charge/discharge, and optimal load shifting-based DSM, while maximizing the use of renewable energy resources. Therefore, the proposed objective function is as given in (1):

$$\mathit{min}\hspace{0.33em}f=C_{outage}+C_{EV}+C_{DER}+C_{load}+C_{loss}.$$

(1)

The expected outage losses cost depends on the average cost per kWh of unsupplied energy (interruption energy assessment rate, i.e., IEAR) in addition to the amount of load shedding at each bus of the network ($P_{shed}$), taking into consideration the outage rate ($\gamma$) and the equivalent time of that outage ($d$) at each bus, as provided in (2). The IEAR in turn is dependent on the expected energy not supplied per hour (EENS) and the expected interruption cost per hour (ECOST) at each bus, as given in (4)–(6).

$$C_{outage}={\sum }_{i=1}^N{IEAR}_i{\gamma }_i{u}_i{P}_{shed-i}.$$

(2)

$$u_i={\displaystyle \frac{{\sum }_{j=1}^{NS}{d}_{ij}}{{\gamma }_i}}.$$

(3)

$${IEAR}_i={\displaystyle \frac{{ECOST}_i}{{EENS}_i}}.$$

(4)

$${EENS}_i={\displaystyle \frac{P_{shed-i}{\sum }_{j=1}^{NS}{d}_{ij}}{SD}}.$$

(5)

$${ECOST}_i={\displaystyle \frac{P_{shed-i}{\sum }_{j=1}^{NS}{co}_j}{SD}}.$$

(6)

The cost of energy supplied by dispatchable DERs is given in (7). As provided in (7), the generation costs considered for minimization are of the diesel and CHP units. The generation costs of PV and wind are not included as the objective is to maximize the use of renewable energy resources for sustainability and environmental considerations. The cost functions of diesel and CHP units are given in (8) and (9), respectively.

The cost function of CHPs follows a quadratic form for both power and heat generation [29]. However, the heat-related part of the cost is not included as the heat energy of the system is not considered for management in this work. It is assumed that CHP units included in the study operate in environments where sufficient thermal demand exists (e.g., residential or commercial areas), allowing flexible electrical dispatch within their rated limits.

$$C_{DER}=C_D+C_{CHP}.$$

(7)

$$C_D={\sum }_{t=1}^{NS}\left[{\sum }_{i=1}^{ND}(a_i+b_i{P}_{D-i}(t)+c_i{P}_{D-i}(t{)}^2)\right].$$

(8)

$$C_{CHP}={\sum }_{t=1}^{NS}\left[{\sum }_{i=1}^{NCHP}({\alpha }_i+{\beta }_i{P}_{CHP-i}(t)+{\epsilon }_i{P}_{CHP-i}(t{)}^2)\right].$$

(9)

The cost function of EV operation is given in (10), including a priority coefficient for each vehicle ($c_{pr}$), while the cost of load consumption at each node and the cost of power loss of the system are given in (11) and (12), respectively.

$$C_{EV}={\sum }_{t=1}^{NS}{\sum }_{i=1}^{NEV}\left[E_{ch-i}\right(t)-E_{dis-i}(t)]c_{EV-i}{c}_{pr-i}.$$

(10)

$$C_{load}={\sum }_{t=1}^{NS}{E}_{load}\left(t\right)c_{load}\left(t\right).$$

(11)

$$C_{loss}={\sum }_{t=1}^{NS}{E}_{loss}\left(t\right)c_{loss}\left(t\right).$$

(12)

3.2. Operational Constraints

3.2.1. Island Formation Constraint

While partitioning the system into islands, it must be taken into consideration that each island contains a source of energy and is not solely composed of load buses. In addition, since the studied distribution system is originally radial and islanding is performed only via line-opening actions, all resulting islands inherently maintain a radial topology.

Therefore, the constraint given in (13) must be fulfilled when forming an island:

N_DER-i ≥ 1.

(13)

3.2.2. Power Balance Constraint of Each Island

In order to form a reliable and stable island, the power balance constraints given in (14) should be considered:

$$P_D\left(t\right)+P_{PV}\left(t\right)+P_W\left(t\right)+P_{CHP}\left(t\right)+P_{EV-dis}\left(t\right)-P_{EV-ch}\left(t\right)-P_{loss}\left(t\right)-P_{load}\left(t\right)=0.$$

(14)

3.2.3. Load Shedding Constraint of Each Island

The load shedding should be considered whenever there is a mismatch between the available generated power and the load power while forming an island. Therefore, to achieve stable islanding load shedding is performed in each island to satisfy the constraint given in (15). The loads’ priorities are considered while load shedding such that the shedding starts with loads of lower priorities.

$$P_{shed}\left(t\right)=P_{load}\left(t\right)-P_{DER}\left(t\right)+P_{loss}\left(t\right).$$

(15)

3.2.4. Diesel Generator Constraint

$$P_D^{min}\le P_D(t)\le P_D^{max}.$$

(16)

3.2.5. CHP Generator Constraint

$$P_{CHP}^{min}\le P_{CHP}(t)\le P_{CHP}^{max}.$$

(17)

3.2.6. Power Flow Constraints

$$P_i-P_{load-i}={\sum }_j\left|V_i\right|\left|V_j\right|Y_{ij}\mathit{cos}({\theta }_{ij}+{\delta }_j-{\delta }_i),$$

(18)

$$P_i-P_{load-i}={\sum }_j\left|V_i\right|\left|V_j\right|Y_{ij}\mathit{cos}({\theta }_{ij}+{\delta }_j-{\delta }_i),$$

(19)

where V is the bus voltage and δ is its angle, and $Y_{ij}$ is the element of bus admittance matrix and ${\theta }_{ij}$ is its angle. P and Q are the real and reactive power generation at that bus, respectively. $P_{load}$ and $Q_{load}$ are the real and reactive load power at that bus, respectively.

3.2.7. Thermal Constraint of Network Feeders

$$\sqrt{P_{ij}^2+Q_{ij}^2}\le S_{ij}^{max},$$

(20)

where $Pij$ and $Qij$ are the active and reactive power flow in line ij, respectively, while $S_{ij}^{max}$ is the maximum thermal limit of that line.

3.2.8. Bus Voltage Constraint

$$V^{min}\le V_i\le V^{max}.$$

(21)

3.2.9. Electric Vehicle Model and Constraints

In this paper, two categories of electric vehicles (EVs) are considered: commercial and residential. Each category has its pattern of arrival and departure times, which influence their patterns of charge/discharge. In addition, EV owners have the freedom to decide whether to share their EV’s stored energy with the network. This decision is simulated by a willingness factor ($W_f$) assigned to each vehicle. Furthermore, the minimum SOC of each vehicle is assumed to depend on its next journey information given by the vehicle’s owner. For that reason, a factor (${\mathrm{J}}_{EV-i}$) is given to each vehicle to determine its allowable minimum SOC. Taking these factors into consideration, the state of charge (SOC) of each vehicle is updated according to the constraints given in (22)–(26):

$${\mathrm{S}\mathrm{O}\mathrm{C}}_{EV}(\mathrm{t}+1)E_{EV}^{max}={\mathrm{S}\mathrm{O}\mathrm{C}}_{EV}(\mathrm{t})E_{EV}^{max}+{\eta }_{EV}\left[M_{EV}(P_{EV-ch}(t+1){\displaystyle \frac{M_{EV}+1}{2}}+P_{EV-dis}(t+1){\displaystyle \frac{1-M_{EV}}{2}})W_f\right]∆t,$$

(22)

$$M_{EV}=\left\{\begin{array}{c}1 \mathrm{if} \mathrm{the} \mathrm{EV} \mathrm{is} \mathrm{charging}\\ -1 \mathrm{if} \mathrm{the} \mathrm{EV} \mathrm{is} \mathrm{discharging}\\ 0 \mathrm{if} \mathrm{the} \mathrm{EV} \mathrm{is} \mathrm{idle}\end{array}\right.,$$

(23)

$${SOC}_{EV-i}^{min}={SOC}_{EV}^{min}\ast {\mathrm{J}}_{EV-i},$$

(24)

where

$${\mathrm{J}}_{EV-i} \ge 1 \mathrm{and} W_f\in [0,1].$$

(25)

The willingness factor ($W_f$) is independently assigned to each EV, allowing the model to capture varying levels of participation rather than assuming full cooperation. A value of $W_f$ = 0 represents no participation, while $W_f$ = 1 represents full willingness, with intermediate values reflecting partial contribution. Therefore, the EV participation is probabilistic and limited, not fully cooperative, in which the aggregated contribution of EVs is scaled by the willingness factor such that the effective available energy from EV fleets is reduced proportionally. This ensures that the restoration strategy does not rely on unrealistically high EV support.

$${\eta }_{EV}=\left\{\begin{array}{c}{\eta }_{EV}^{ch} \mathrm{if} \mathrm{the} \mathrm{EV} \mathrm{is} \mathrm{charging}\\ {\displaystyle \frac{1}{{\eta }_{EV}^{dis}}} \mathrm{if} \mathrm{the} \mathrm{EV} \mathrm{is} \mathrm{discharging}\end{array}\right..$$

(26)

The limitations on the SOC and the charging/discharging rates of each EV in addition to the charging/discharging intervals are given in (27):

$$\left\{\begin{array}{c}{SOC}_{EV-i}^{min}\le {\mathrm{S}\mathrm{O}\mathrm{C}}_{EV-i}(\mathrm{t})\le {SOC}_{EV}^{max}\\ {\mathrm{P}}_{ch}(\mathrm{t})\le {\mathrm{R}}_{EV}^{ch}\\ {\mathrm{P}}_{dis}(\mathrm{t})\le {\mathrm{R}}_{EV}^{dis}\\ T_{arv}\le t_{ch}^{start}\le T_{dep}-1\\ t_{ch}^{start}+1\le t_{ch}^{stop}\le T_{dep}\\ t_{ch}^{stop}+1\le t_{dis}^{stop}\le T_{dep}-1\\ t_{dis}^{start}+1\le t_{dis}^{stop}\le T_{dep}\end{array}\right..$$

(27)

Therefore, the consideration of constraints (25) and (27) demonstrates that EV participation is already constrained through arrival/departure time patterns for residential and commercial EVs, and a willingness factor ($W_f$) that probabilistically limits participation, in which lower values of $W_f$ can be interpreted as reflecting user hesitation arising from degradation costs, lack of incentives, or other behavioral factors. However, the explicit consideration of battery degradation costs is not included in the economic model.

The formulated constrained optimization problem is solved within the proposed multi-layer metaheuristic framework. Specifically, candidate solutions (individuals) are first generated according to the decision variables of the three optimization layers: islanding configuration, DSM/EV scheduling, and DER dispatch introduced in Section 2.1. For each candidate solution, the objective function is evaluated only after verifying all operational and technical constraints.

The constraints are handled during the optimization process through two complementary mechanisms:

• Feasibility checking: Each candidate solution is tested against network operational constraints, including power balance, voltage limits, line loading capacities, DER generation limits, EV charging/discharging boundaries, and load restoration requirements.
• Rejection-based constraint handling: If a candidate solution violates any constraint, it is rejected and another solution is generated to make sure that each population is formed only of accepted solutions (individuals). This guides the metaheuristic algorithms toward feasible and operationally acceptable regions of the search space.

In addition, the hierarchical structure of the proposed algorithm naturally reduces complexity because each optimization layer limits the feasible decisions of the following layer. For example, the islanding decisions in Layer 1 define the operational boundaries for DSM and EV scheduling in Layer 2, which subsequently determine dispatch requirements in Layer 3.

3.3. Premium Setting

A premium is the payment made by an insured party in exchange for the insurer’s agreement to provide coverage against potential losses. In the P2P insurance scheme, the SO collects premiums from P2P insurance intermediaries. The premium is composed of two parts; pure premium $Pr$, and expense loading ${Pr}_{exp}$. The pure premium is the part that is directly related to losses, while expense loading is an additional cost to maintain normal operations of the insurer, which is calculated as percentage ($\mu$) of the pure premium [18]. Moreover, with different categories of customers, each category will have a different priority and hence different insurance rate. In order to represent these priorities, a proportion coefficient ($\beta$) is introduced when calculating the premium of each customer. Therefore, the premium can be calculated as in (28):

$${TPr}_i={Pr}_i+{Pr}_{exp-i}=\left(1+\mu \right){Pr}_i(1+{\beta }_i).$$

(28)

The total fund ($M$) to be collected by the P2P insurance intermediary from the mutual aid group users is given as in (29):

$$M={\sum }_{i=1}^N{TPr}_i.$$

(29)

As explained in Section 2.2, a percentage ($m$%) of the collected fund ($M$) will be paid as a premium to the SO, while the remaining fund will be used as a mutual aid fund pool. If no outages occur to any of the users during the insurance claim period, a percentage $(100-m)$% of ($M$) will be returned to users in proportion to their contribution to the fund pool. Therefore, each participant will receive a return fund ($RF$) as given in (30):

$${RF}_i=\left(1-{\displaystyle \frac{m}{100}}\right)M{\displaystyle \frac{{TPr}_i}{M}}=\left(1-{\displaystyle \frac{m}{100}}\right){TPr}_i.$$

(30)

On the other hand, if outages occur during the insurance claim period, payouts take place according the threshold ($Mr$). Similarly, the remaining funds in the pool ($M_{rem})$, after satisfying the small claims of the mutual aid group ($Cs$), are returned to group members depending on their participation ratios as given in (31):

$${RF}_i=M_{rem}{\displaystyle \frac{{TPr}_i}{M}}.$$

(31)

$$M_{rem}=\left(1-{\displaystyle \frac{m}{100}}\right)M-{\sum }_{i=1}^N{Cs}_i.$$

(32)

Therefore, the P2P insurance intermediary is responsible for selecting the appropriate values of the percentage $(m\%)$ and threshold ($Mr$) to guarantee that the mutual funds pool is sufficient enough to cover small payouts. In case the pool amount is insufficient to cover such payouts, the P2P insurance intermediary is contractually responsible for paying the remaining claims.

3.4. Blockchain-Based P2P Insurance Implementation

Figure 4 shows how the blockchain-based P2P insurance process is carried out. The steps of transactions for a claim or a refund are as follows:

• The P2P insurance intermediary uploads smart contracts to the blockchain’s distributed ledger.
• After an outage occurrence, an insurance claim is issued into a MAG by the user.
• According to the uploaded smart contracts, the payout amount and classification are determined.
• The validation of the transaction is carried out by a PoA consensus. In such PoA consensus, the elected P2P insurance intermediary acts as the miner of the blockchain.
• The transaction is carried out either by the funding pool using the mutual fund, or by the system operator (i.e., insurer), depending on the amount of payout relative to the threshold (Mr).
• At the end of the insurance period, transactions of the remaining fund in the funding pool are validated.
• The refund transactions are carried out by the funding pool to participants of the mutual aid group.

It is worth noting that P2P intermediaries charge the policyholders a certain mining fee, which is a percentage of each transaction.

For practical implementation, some aspects are considered. The proposed blockchain-based P2P insurance mechanism is designed as a lightweight, permissioned blockchain framework, where transactions are limited to insurance-related events (e.g., premium payments and compensation settlements), resulting in relatively low computational and communication overhead compared to public blockchain networks.

Regarding transaction latency, the proposed application is not time-critical (i.e., insurance settlement occurs after outage events rather than in real-time control), making typical confirmation delays acceptable within the operational context.

In terms of scalability, the system is implemented at the distribution network level, where the number of participants is moderate. In addition, off-chain data management and aggregation techniques can be employed to further reduce on-chain transaction volume.

4. Case Study

4.1. System Description

The proposed individual-based optimization algorithm is applied to the 45-bus distribution system of Alexandria, Egypt. The system is assumed to lose connection to the upstream national network due to rough weather conditions during a day in January. Therefore, the simulation period of 24 h will be considered, with a time step of 15 min. The distribution system under study includes different types of users, residential, commercial and industrial, which have different loading patterns and priorities. In addition, the system has different types of DERs and different types of EV parking garages. The location, type and capacity of DERs are given in

Table 2. The location, type and capacity of DERs.

Location (Bus)	DER Type *	Capacity (MW)	Garage Capacity (Vehicles)	Garage Type **
2	D	100	-	-
3	G	-	200	R
4	CHP	100	-	-
5	D	220	-	-
6	PV	98	-	-
7	CHP	80	-	-
9	D	180	-	-
10	W	51	-	-
12	D	80	-	-
13	CHP	80	-	-
14	D	60	-	-
15	CHP	60	-	-
17	PV	55	-	-
21	D	80	-	-
25	D	40	-	-
29	CHP	30	-	-
34	CHP	180	-	-
35	G	-	140	R
36	G	-	200	C
37	G	-	160	C
40	W	90	-	-
41	CHP	100	-	-
42	G	-	180	C
44	D	120	-	-

* G = EV garage, PV = photovoltaic units, W = wind units, D = diesel units, CHP = combined heat and power units. ** R = residential, C = commercial.

, while the diesel and CHP generator data are given in

Table 3. Diesel and CHP generator data.

Location	Fuel Cost Coefficients
	$\mathbf{a}/\mathit{\alpha }$ ($)	$\mathbf{b}/\mathit{\beta }$ ($/MW)	$\mathbf{c}/\mathit{\epsilon }$ ($/MW2)
2	1000	10.67	0.0051
4	460	11.67	0.0053
5	1000	10.67	0.0051
7	460	11.67	0.0053
9	1000	10.67	0.0051
12	680	16.5	0.00211
13	460	11.67	0.0053
14	680	16.5	0.00211
15	460	11.67	0.0053
21	680	16.5	0.00211
25	640	15.5	0.0021
29	300	16.75	0.0127
34	580	12.43	0.0032
41	460	11.67	0.0053
44	1000	10.67	0.0051

. The per unit output power of both wind turbines and PV units due to predicted average solar irradiation [30] and wind speed [31] during an average day in January are shown in Figure 5.

Regarding EV parking garages, residential and commercial garage users have different patterns depending on their occupation, which affect their arrival and departure times. The arrival/departure patterns of these vehicles are considered to have a normal distribution, but different mean and standard deviations, as presented in [20]. The percentage of available EVs in each type of parking garage is presented in Figure 6. During the periods of EV availability, vehicle users can charge or discharge their vehicle and share their stored energy with the distribution network, depending on the SOC of their vehicles.

Although the 45-node distribution system of Alexandria is used for the proposed framework validation, the adopted network represents a realistic configuration with practical operational characteristics. The scalability of the proposed approach is supported by the use of population-based optimization algorithms, which allows efficient handling of large solution spaces and supports parallel computation.

It is also important to note that the proposed framework is intended for planning rather than real-time control, where computational time constraints are less restrictive. For real-time applications, the framework can be extended using hierarchical or distributed control strategies and simplified decision-making models.

4.2. Optimal Island Partitioning

Applying the proposed multi-objective, multi-layer optimization algorithm to the first layer of the test system, the network is partitioned into six islands, as presented in Figure 7. As shown in Figure 7, each island is formed of a number of load buses in addition to DER buses, satisfying the islanding constraint given in (13).

4.3. Optimal Energy Management—DSM and DER Dispatch

According to the islands formed in the first layer of the optimization algorithm, optimal energy management of the islands’ loads and DERs are considered in the second layer, taking the operational constraints of each island into consideration.

Load shifting-based DSM of islands’ loads is carried out in order to minimize energy consumption costs as the wholesale energy price varies during the day, as shown in Figure 8. In this paper, the three types of loads (i.e., residential, commercial and industrial) are assumed to have the same energy price.

With load shifting, the use of EV stored energy and the optimal dispatch of diesel and CHP generating units, both the generation cost and load shedding will be reduced. Figure 9 presents the total load of the whole system before and after DSM. As shown in Figure 9, the application of optimal DSM resulted in peak reduction and load shifting into the intervals of lower energy price.

Table 4. The system performance with and without the application of optimal islanding, energy management, and load priority.

Performance Aspect	Islanded System	Non-Islanded System	Reduction (%)
Peak demand (MW)	1.9528 × 103	2.1548 × 103	9.37
Energy loss (MWh)	449.1494	710.7368	36.81
Unserved energy (MWh)	2.8358 × 103	10.209 × 103	72.22
Cost and Cost Saving
Energy consumption cost ($)	5.0638 × 106	5.2052 × 106	27.2
Energy losses cost ($)	1.6524 × 104	2.5995 × 104	36.44
Generation cost ($)	2.96 × 106	3.0633 × 106	3.33
Outage losses cost ($)	8.0378 × 106	11.1565 × 106	38.8
Total cost ($)	16.078 × 106	19.451 × 106	17.34

compares the system performance with and without the application of optimal islanding and energy management and without load priority consideration. The system performance is compared in terms of peak demand, energy consumption cost, energy loss, unserved energy, and generation cost. As shown in Table 4, application of the proposed algorithm resulted in a peak reduction of 9.37%, in addition to a reduction of 27.2% in the cost of energy consumption. In each island, the available DERs are responsible for supplying the island’s load. As a result, a reduction of 3.33% in generation cost is achieved due to optimal island formation.

The apparent discrepancy between the 27.2% reduction in energy consumption cost and the 3.33% reduction in generation cost arises from the fact that these two cost components represent different economic quantities within the proposed framework, and are influenced by different mechanisms. The energy consumption cost reflects the total cost incurred to supply the demand, which includes the effects of demand-side management (DSM), EV scheduling, load shifting, and peak reduction. The generation cost, on the other hand, represents only the operational cost of dispatchable local generators (e.g., diesel generators and CHP units), excluding the contribution of renewable DERs and demand reduction. In addition, dispatchable generators are mainly used as supporting resources after DSM and renewable utilization. Their operating range is already relatively constrained, and their cost coefficients do not vary significantly across the compared scenarios. Therefore, energy consumption cost captures system-level economic benefits, whereas generation cost reflects only local dispatchable unit operation.

Furthermore, after splitting the system into islands, the unserved energy—and hence the outage losses cost—reduced by 72.22% and 38.8%, respectively. The reduction in outage losses cost resulted from considering the load shedding of lower priority loads first. Furthermore, splitting the system into islands resulted in a reduction in energy loss of 36.81%, and hence a reduction in the energy losses cost by 36.44%. From the point of view of the overall saving resulted from applying the optimal islanding algorithm along with optimal energy management, the proposed algorithm resulted in an overall saving of 3372.87 × 10³$ (i.e., 17.34%).

Figure 10 presents the share of each DER in an island to the power supplied in that island, in addition to the power required for the island’s loads. As shown in Figure 10, in some islands, the load demand at some intervals is higher than the available generated power, which leads to load shedding to satisfy the power balance constraint given in (14). In addition, although both island 2 and island 3 have two parking garages each, EVs have no contribution to the islands’ power supply because all of the discharged energy was used internally in the garages to charge other vehicles as shown in Figure 11. On the other hand, at some time intervals, the parking garage in island 1 has a surplus of discharged energy to share among the island’s loads. However, sharing the discharged energy internally in the garages helps reduce part of the island load by supplying charging EVs.

As described in this work, the proposed multi-layer optimization framework can be effectively solved using individual-based metaheuristic algorithms. In this study, four widely used algorithms—Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Artificial Bee Colony (ABC), and Firefly Algorithm (FA)—are employed to solve the formulated problem.

A population size of 200 individuals and a maximum of 10,000 iterations are considered for all algorithms. The convergence behavior of the objective function is illustrated in Figure 12, while the corresponding computational times and minimum objective values are summarized in

Table 5. Computational times and minimum values of different algorithms.

Algorithm	Computational Time (Hours)	Minimum Value ($)
GA	2.8	16.078 × 106
PSO	1.75	18.1 × 106
FA	1.5	18.8 × 106
ABC	0.9	20.5 × 106

.

Based on implementation on an Intel^® Core^TM i7-7500U CPU, 2.70 GHz with 8 GB RAM, the computational time ranges from approximately 1 to 4 h depending on the selected optimization algorithm. Among the tested methods, GA requires the highest computational time due to its intensive search mechanism, while swarm-based algorithms such as PSO and FA achieve near-optimal solutions with lower processing time. The ABC algorithm exhibits the lowest computational time but with slightly reduced solution quality.

The results indicate that GA achieves the lowest total cost value compared to the other algorithms; however, it requires the highest computational time due to the complexity of the problem. PSO and FA exhibit similar convergence characteristics and reach near-optimal solutions, with FA demonstrating slightly lower computational time. The ABC algorithm also achieves convergence with relatively low computational time, although the obtained solution is slightly inferior compared to PSO and FA.

Overall, swarm-based algorithms such as PSO and FA provide a favorable balance between solution quality and computational efficiency. Nevertheless, since the proposed framework is designed for planning applications, computational time does not represent a critical limitation.

4.4. Application of Blockchain-Based P2P Insurance Mechanism

As shown in Figure 7, the distribution network under consideration has three types of customers: residential, commercial and industrial customers. Each type of user can aggregate into a separate mutual aid group with different insurance rates. Therefore, each island is assumed to have three mutual aid groups with P2P insurance blockchains and PoA consensus.

It should be noted that two distinct economic parameters are considered in this study: energy price and interruption cost (IEAR). The energy price is used within demand-side management (DSM) optimization to represent the cost of electricity consumption, and is assumed to be uniform across different customer sectors for modeling simplicity and to focus on system-level operational behavior. In contrast, the interruption cost values presented in

Table 6. Interruption costs by sector.

Sector	Interruption Time and Cost ($/kW)
	1 min	20 min	60 min	240 min	480 min	1440 min
Commercial	0.381	2.969	8.552	31.32	83.01	488.3
Residential	0.001	0.093	0.482	4.914	15.69	117.91
Govt./Inst.	0.044	0.369	1.492	6.558	26.04	253.6

[32] are sector-dependent and reflect the economic impact of service interruptions for residential, commercial, and governmental/institutional loads. These interruption costs are not used in the DSM scheduling process but are instead applied in the evaluation of unserved energy and outage-related costs, as well as in prioritizing load restoration. Therefore, the use of a uniform energy price does not contradict the differentiated interruption costs, as each parameter serves a distinct role within the overall optimization and assessment framework.

The reliability assessment indices for each island are given in

Table 7. The reliability assessment indices for each island.

Island 1
Bus Number	1		10		31		36		43		44	45
u (h)	12.5		12.5		12.5		12.5		12.5		6.25	12.5
d (h/year)	4		4		4		4		4		2	4
IEAR ($/kWh)	1.23		1.23		1.23		1.23		7.83		0.24	1.23
Island 2
Bus Number	19		9		28		30		35		37	41
u (h)	12.5		12.5		12.5		12.5		12.5		12.5	6.25
d (h/year)	4		4		4		4		4		4	2
IEAR ($/kWh)	1.23		1.23		1.23		1.23		1.23		1.23	0.24
Island 3
Bus Number	18	2	3	11	12	13	20	21	29	32	38	42
u (h)	12.5	12.5	12.5	12.5	12.5	12.5	12.5	12.5	12.5	12.5	9.375	12.5
d (h/year)	4	4	4	4	4	4	4	4	4	4	3	4
IEAR ($/kWh)	1.23	1.23	1.23	1.64	7.83	7.83	1.23	1.23	1.23	1.23	0.16	7.83
Island 4
Bus Number	22		4		14		15		24		25	33
u (h)	12.5		6.25		12.5		12.5		12.5		12.5	12.5
d (h/year)	4		2		4		4		4		4	4
IEAR ($/kWh)	1.23		0.24		7.83		1.64		1.64		1.64	1.23
Island 5
Bus Number	8		5			27		34			39
u (h)	7.81		12.5			12.5		12.5			12.5
d (h/year)	2.5		4			4		4			4
IEAR ($/kWh)	0.597		1.23			1.64		1.23			1.23
Island 6
Bus Number	16		6		7		17		23		26	40
u (h)	6.25		12.5		12.5		12.5		12.5		12.5	12.5
d (h/year)	2		4		4		4		4		4	4
IEAR ($/kWh)	0.746		7.83		1.64		1.23		1.23		7.83	1.23

, in which ($u$) is the equivalent outage time in a year and ($d$) is the average annual outage time of each bus. In this paper, the equivalent outage rate ($\gamma$) is assumed to be 0.32 (times/year) for all buses as the upstream national network is considered a highly reliable network. The outage rate is assumed constant as an exogenous parameter, as the proposed islanding strategy primarily impacts outage duration rather than outage frequency.

It must be noted that the outage duration of some buses is lower in the case of the non-islanded system, while other buses have a prolonged outage duration, as the load shedding starts from the first bus up to the remaining buses of the system without taking the loads’ priorities into consideration. Therefore, the outage losses and their costs were higher than those of the islanded system, which affect the total premium of the system, as will be discussed later.

According to the outage duration of each bus, the premium for each user can be calculated using (29), while the overall premium of each mutual aid group is calculated using (30). In this paper, each customer type has their own insurance rate. Therefore, commercial and industrial customers are considered to pay higher premiums than residential users by 10% and 15%, respectively. In addition, the expense loading proportion coefficient ($\mu$) is set to 0.1, while the fund allocation ratio ($m\%$) is set to 80% for all mutual aid groups. To allocate the payout to the insurer or to the mutual fund pool, a threshold ($Mr$) is set to be equal to the expected average outage loss cost of the group. Applying this premium setting to the system with and without the consideration of network splitting, energy management, and load priority resulted in a total premium of 4.02 × 10⁷$ for the islanded system and of 5.681 × 10⁷$ for the non-islanded system. Therefore, applying the proposed multi-layer, multi-objective optimization algorithm resulted in a reduction of 41.3% in the total premium.

To verify the effectiveness of applying the blockchain-based P2P insurance mechanism, an analysis of island 1 is introduced. Depending on the reliability assessment indices given in Table 7, the outage losses cost (payouts), mining fees, total group premium, total mutual fund, source of payouts, and total funds returned at the end of the claim period are given in

Table 8. Blockchain-based P2P insurance mechanism applied to island 1.

Island 1
Bus Number	1	10	31	36	43	44	45
Payouts ($ × 103)	135.7	108.6	466.6	55.7	1842.1	14.3	273.2
Source of payouts *	MFP	MFP	SO	MFP	SO	MFP	MFP
Blockchain mining fees ($)	135.7	108.6	466.6	55.7	1842.1	14.3	273.2
Total SO premium ($ × 103)	2935.7
Total mutual fund ($ × 103)	733.92
Total funds returned ($ × 103)	146.5
Threshold—$\mathit{M}\mathit{r}$ ($ × 103)	413.74

* SO = system operator, MFP = mutual fund pool.

.

As shown in Table 8, the total premium insured to the SO is 2935.7 × 10³$, while the total mutual fund remaining in the fund pool is 733.92 × 10³$, making the overall premium collected by the P2P insurance intermediary equal to 3.67 × 10⁶$. When outage events take place, certain payouts are to be claimed. According to the proposed P2P insurance scheme, these payouts are compared to the payout threshold (Mr) to determine who is responsible for such payouts. As shown in Table 6, the payout required for the outage at bus 1 is 135.7 × 10³$, which is less than the Mr. Therefore, the payout was paid by the P2P insurance intermediary through the MFP after charging 0.1% (i.e., 135.7$) for mining fees. On the other hand, for an outage event at bus 31, the payout required is 466.6 × 10³$, which exceeds the payout threshold. Therefore, the payout was paid by the P2P insurance intermediary through the SO after charging 0.1% (i.e., 466.6$) for mining fees.

Through the whole claim period, the total outage insurance that was paid by the SO was 2.3087 × 10³$, while the P2P insurance intermediary needed to make small payments of 587.5 × 10³$ total through the MFP. Therefore, the SO made a profit of 627 × 10³$ after paying these claims, while there was a remaining 146.5 × 10³$ in the MFP. The remaining funds are returned to policyholders in proportion to their contribution to the fund pool as given in (31)–(32). Therefore, it can be said that the equivalent total premium for P2P insurance is further reduced to 3.67 × 10⁶ − 146.5 × 10³ = 3.52 × 10⁶$. On that basis, the average payouts, revenue and fund return of all islands are given in

Table 9. Average payouts, revenue and fund return of all islands.

Island	1	2	3	4	5	6
Average SO payout ($ × 106)	2.31	1.58	11.61	3.71	0.285	6.27
Average SO revenue ($ × 103)	627	395.83	3266.5	855.83	95.02	1160.3
Average P2P insurance intermediary payout ($ × 103)	587.42	364.3	3648.5	875	77.27	834.83
Average P2P insurance intermediary surplus ($ × 103)	146.5	130.87	70.1	266.91	17.61	1021.9

.

5. Conclusions

In this paper, a sustainable and resilient fault recovery framework based on an intentional islanding scheme was proposed to partition the distribution system in order to optimize the priority restoration of critical loads, while taking the operational constraints of the system into consideration. The proposed approach enhanced the sustainability of power system operation by improving the utilization of distributed energy resources (DERs), reducing energy losses, and ensuring reliable supply to essential loads during outage conditions. In addition, a blockchain-based P2P insurance mechanism was applied to further mitigate the outage losses of the network’s users with a higher degree of security and transparency. This mechanism contributed to economic and social sustainability by enabling fair risk sharing among users, reducing reliance on centralized insurance structures, and increasing participation through trustless transactions.

For this purpose, a multi-layer, multi-objective optimization algorithm was introduced for optimal partitioning of the distribution network, management of DERs, and demand side management (DSM) of flexible loads. The objectives of this algorithm were to minimize the cost of energy consumption of different loads, the cost of energy supplied by dispatchable DERs in addition to EVs, the cost of energy loss, and the cost of outage losses, hence the insurance premium. These objectives were accomplished, while maintaining satisfactory performance of the network in terms of network power loss, voltage deviations, line loading, and critical loads supply. To validate the feasibility of the proposed algorithm, it was applied to the 45-node distribution network of Alexandria, Egypt. The results show that reduction in peak load, cost of energy consumption, outage losses, energy loss and generation cost were achieved, with an overall saving of 17.34%. In addition, the application of the proposed blockchain-based P2P insurance mechanism resulted in an overall saving of 1.6539 × 10⁶$ (i.e., 4.11%) of the insurance cost in the form of returned funds, while the proposed optimization algorithm resulted in a premium reduction of 41.3%.

Although the proposed framework is validated through simulation, the adopted 45-node system represents a realistic distribution network based on actual data from Alexandria, Egypt, including practical topology and load characteristics. Therefore, the obtained results provide meaningful insights into real-world system performance. Furthermore, the proposed approach is compatible with existing distribution management systems and can be implemented using current communication and control infrastructures. Future work will focus on real-time validation using hardware-in-the-loop simulations and pilot-scale deployments to further demonstrate practical feasibility.

Overall, the proposed framework provides a comprehensive techno-economic solution that enhances system reliability, operational efficiency, and economic performance, while supporting fair compensation mechanisms for users. These features contribute to the development of more resilient and efficient distribution systems.