Blockchain-Driven Peer-to-Peer Smart Energy Trading for Decentralized Power Networks

The most recent trends in the operation and design of a power grid include an influx of distributed energy resources (DERs) from rooftop solar panels, wind turbines, and home battery systems; as the number of DERs continues to grow, so too does the spectrum of risk and reward for both consumers and utilities. This development has allowed those who produce their own electricity (prosumers) to participate in local energy markets and therefore decrease their reliance on large centralized electricity suppliers. Another benefit of allowing prosumers to directly sell surplus energy generated by their DER is that grid operators have greater capabilities to enhance grid resilience and flexibility [1]. Conversely, establishing a market structure that meets the needs of all participants while simultaneously providing adequate security and privacy for distributed energy generation and storage [2].

Peer-to-peer (P2P) energy trading allows for transactions between prosumers and avoids the costs of relying on a centralized third party for these transactions [3]. Prosumer-to-prosumer transactions improve the local utilization of renewable energy generation and minimize the amount of energy lost during the long-distance transmission of electricity, as well as provide opportunities for a more competitive and consumer-centric pricing model. Therefore, to create an efficient, automated, trustworthy P2P energy market, a reliable e-transaction platform must also possess transparency, low latency, privacy, and auditability simultaneously. No currently available e-Transaction platforms possess all of the features outlined above [4]. Blockchain technology provides the foundational building blocks for P2P energy trading markets with capabilities such as decentralization, immutability, automated instructions through smart contracts, etc. The results of proof-of-concept projects and pilot programs demonstrate blockchain technology's ability to alleviate specific trust issues while allowing bid/offer matching automation, settlements, and reputation accounting to be implemented automatically [5]. Nevertheless, Blockchain Technology faces many real-life challenges when it comes to implementing successful deployments. Public Blockchains are characterized by high latency and cost per transaction, which serve as hurdles to successful implementations. Second, purely singular layered blockchains will encounter challenges due to throughput and privacy concerns. Lastly, many proposed schemes do not address regulatory compliance and long-term auditing [6].

There have been many recent publications regarding the various challenges mentioned above. Specifically, cryptographic primitives such as commitments and zero-knowledge proofs are now being incorporated into privacy-enhancing technologies utilizing sufficiently secure meter-level consumption data, bid information, and verifiable settlement capabilities [7]. By using advanced forecasting techniques based on artificial intelligence (AI) and advanced time series models (e.g., LSTM and Transformer), traders can take advantage of better-informed decisions in order to increase their pre-emptive trades. In addition, through the use of machine learning algorithms, energy companies and blockchain transaction network (graph) service providers have successfully detected fraudulent and abnormal activities occurring through their energy and blockchain networks [8]. Additionally, machine-learning algorithms have been utilized to provide energy companies with a means to identify thefts, tampering, and manipulation of transactions that occur on their blockchain transaction network [9].

While advances have been made in developing P2P trading technology for urban microgrid energy systems, this research identifies three significant gaps: Many research efforts have focused on one primary component without developing a cohesive architecture to satisfy all three components of latency, scalability, and regulatory auditing. There is a lack of available solutions that include the integration of predictive analytics into smart contracts, enabling preemptive matching and grid-aware trading that is based on short-term forecast data and actual network constraints. And comprehensive countermeasures that combine methods for cryptographic privacy, reputation/incentive design, and machine learning-based fraud detection in a unified and operational framework are extremely limited [10-12].

The above-mentioned issues ultimately result in a barrier for the practical use of P2P trading technology within the scope of urban microgrids/community energy systems. Also, due to these identified barriers, this paper is proposing the PBET-Chain, a new architecture that simultaneously addresses the following five critical objectives: latency, privacy, auditability, predictive optimization, and security. The PBET-Chain architecture has an additional layer that separates the functions of the above objectives into two complementary layers of blockchain. One layer is the Fast-Trading Layer (FTL), which is a lightweight, fast, and low-latency second-consortium-chain architecture designed for higher-frequency microtransactions; and the other layer is the Settlement & Audit Layer (SAL), which provides a more secure, immutable ledger to ensure that transactions can be audited, reported to regulatory authorities, and provide an avenue for resolving disputes. An AI-Enhanced Energy Prediction Engine (AI-EPE) works closely together with a Multi-Criteria Energy Matching Algorithm (MCEMA) for the purpose of pairing energy buyers and sellers based on their locations on a grid while using a cost-sensitive approach. Privacy-preserving metering using Zero Knowledge Proofs (ZKP) and a mixed cyber-fraud detection and incentive structure contributes towards the protection of participant data and the integrity of the system.

The three main contributions of this research include the following:

• The PBET-Chain dual-layer blockchain design that combines low latency, auditability, and regulatory compliance;
• An AI-enabled integrated multi-criteria matching methodology that takes into consideration short-term forecasts, grid limitations and trade partners' reputations;
• A layered security and privacy toolkit that includes a zero-knowledge proof system for meter verification combined with ML techniques for fraud detection and a token-based mechanism for reporting and rewarding good behavior;
• Implemented PBET-Chain in tandem with smart meter readings and carried out simulations of microgrids measuring the performance of PBET-Chain against that of single-layered blockchain systems.

The remainder of this paper consists of the following sections: In Section 2, the literature surrounding the topic, Section 3 provides a complete overview of PBET-Chain’s architecture and its associated threat model. Section 4 outlines the experimental infrastructure on which PBET-Chain’s performance was evaluated. Finally, Section 5 provides a conclusive statement about what was learned from this study and the possible directions for additional research.

Hierarchical levels establish the way that components interact with each other while performing their designated tasks in the architecture of the PBET-Chain, which is made up of five layers. These five layers' purpose is to provide the P2P energy exchange, or trading, necessary to securely, autonomously, and privately trade energy between both users and providers in an open and decentralized manner. In addition, the four layers provide a modular approach for scaling, interoperability, and robustness against uncertainties resulting from their operation. The next section will explain each of the five layers in detail.

3.1 Architectural overview

At the bottom level of the PBET-Chain is a smart meter, which provides the means for measuring DERs such as photovoltaic systems, batteries, and controllable loads. The smart meter continuously measures and records three key criteria: 1. real-time production and consumption, 2. voltage, and 3. frequency. A cryptographic hash of the information contained in each data packet is created using SHA-3 prior to the data packet being sent to the edge device. Edge devices perform specific preprocessing functions. Examples of preprocessing functions include anomaly filtering, timing synchronization, and local aggregation. The results are passed through a communication channel to allow edge devices to interact with one another virtually instantaneously with minimal delays.

The consortium blockchain layer consists of a secure and tamper-proof ledger for all trading transactions. The validation of blocks occurs using a DPoA method. Here, a limited number of trustworthy validators, typically utility operators and community hubs, come to consensus, achieving high throughput and low levels of energy consumption. Every block contains a hashed version of smart meter data, proof of the transaction performed, a trust score, and a timestamp, providing for transparency and auditability. Links between blocks throughout the blockchain are achieved using cryptographic hashes, making it virtually impossible to manipulate energy tokens or double spend them, as it is immutable.

The smart contract execution layer allows for the automation of market operation through a collection of smart contracts. The Energy Offer Contract (EOC) allows a prosumer to register the surplus energy they have available for sale. Energy Demand Contracts (EDC) allow consumers to request energy. The Matching and Settlement Contract (MSC) automatically pairs up buyers and sellers based on price, trust score, and availability. Smart contracts allow trades to occur autonomously, without the need for a centralized intermediary, because they carry out the terms and conditions of trading. A trade can only take place when the offered price meets the purchased price and when the prosumer meets the minimum trust requirement. Contracts control how token-based micro-payments and meter validation are performed, ensuring a fair, transparent, and tamper-resistant process.

PBET-Chain has an energy market optimizer layer as the intelligence behind its price, energy routing, and trust management system, as well as combining reinforcement learning (RL), graph optimizations, and behavioral analytics to continually adjust market prices, direct energy locally via RL algorithms, and assess trustworthiness based on evaluation metrics related to participant productivity.

The PBET-Chain utilizes various forms of cryptographic protection to enhance the confidentiality of energy trade transacted by individuals. The zero-knowledge meter proofs system allows a single user/consumer to validate that their usage/generation of energy was accurately tracked, but does not require the user to reveal actual values for both meter readings. Utility providers take additional steps to protect device anonymity by producing hashed versions of their identifier when communicating with the utility provider and associating them to eliminate the risk of user profiling. All interaction between all participants on the PBET-Chain is protected by end-to-end encryption, and access to sensitive information is restricted. This combination of methods provides safeguards for the integrity of the data, protection against impersonation attacks, mitigation of meter tampering, and maintenance of confidentiality and security for the operation of the decentralized energy trading markets.

3.2 Mathematical model

The smart meter's layer serves as the metering and sensing backbone of the PBET-Chain and is responsible for providing an honest view of energy measurements so that trustworthy energy transactions can occur between peers. The surplus for the seller $i$ is shown in Eq. (1).

${{s}_{i}}\left( t \right)=\left\{ 0,\,{{g}_{i}}\left( t \right)-{{c}_{i}}\left( t \right)-u_{i}^{reserve}\left( t \right) \right\}$                      (1)

Here, i denotes an index that represents an energy producer selling electricity through a P2P electricity trading network, and j denotes an index representing the consumer of electricity from the network. t denotes a discrete time index. s_i(t) represents the excess energy available for sale by prosumer i at time t, and g_i(t) represents the total amount of energy that prosumer i produces at time t. c_i(t) represents the quantity of energy consumed directly by prosumer i at time t and $u_{i}^{reserve}\left( t \right)$ refers to any quantity of energy that prosumer i has set aside for use in contingency situations or as a backup energy source. The overall balance between supply and demand in the entire system, and they ultimately have an effect on how prices, matches, and congestion management function in the PBET-Chain energy trading system as shown in Eq. (2).

$S\left( t \right)=\sum\limits_{i\in {{P}_{s}}}{{{s}_{i}}}\left( t \right),D\left( t \right)=\sum\limits_{j\in {{P}_{b}}}{{{d}_{j}}}\left( t \right)$                 (2)

Here, P_s denotes the collection of prosumers that take part in an energy exchange network. P_b denotes the collection of consumers that take part in an energy exchange network. S(t) denotes and represents the total quantity of energy supplied to the energy exchange marketplace at time t. ${{d}_{J}}\left( t \right)$ represents the amount of energy requested by buyer j.

The blockchain provides a basis for the distributed trust found in the exchange of data and from the trading participants. Consensus latency model for fast-time ledger (FTL) settlement as shown in Eq. (3).

${{\tau }_{FTL}}\left( t \right)=\tau _{0}^{FTL}+{{\beta }_{FTL}}\,{{n}_{tx}}\left( t \right)$                             (3)

Here, ${{\tau }_{FTL}}\left( t \right)$ denotes an average consensus latency and $\tau _{0}^{FTL}$ denotes the latency of the FTL mechanism when no transactions are present, capturing fixed overhead such as block proposal and validation delays. ${{\beta }_{FTL}}$ denotes scaling coefficient that represents the incremental latency added per transaction processed under FTL settlement. And ${{n}_{tx}}\left( t \right)$ denotes the number of individual energy trading transactions submitted for settlement at a time. The consensus latency model for slow aggregated ledger (SAL) settlement is computed in Eq. (4).

${{\tau }_{SAL}}\left( t \right)=\tau _{0}^{SAL}+{{\beta }_{SAL}}\,{{n}_{batch}}\left( t \right)$                        (4)

Here, ${{\tau }_{SAL\left( t \right)}}$ denotes average consensus latency for the SAL settlement on time t, used for processing groups of transactions. $\tau _{0}^{SAL}$ denotes the baseline ACL latency associated with SAL settlement, which includes the latencies associated with batching, aggregating, and delayed confirmations. ${{\beta }_{SAL}}$ denotes the ACL Latency Scaling Factor (LSF) for SAL settlement based on the impact that the size of the transaction batch has on the ACL of the settlement. And ${{n}_{batch\left( t \right)}}$ denotes the total number of transaction batches that were processed in the SAL ledger on time t. The transaction cost model is given as shown in Eq. (5).

${{C}_{tx}}\left( t \right)={{c}_{FTL}}\,n_{tx}^{FTL}\left( t \right)+{{c}_{SAL}}\,n_{tx}^{SAL}\left( t \right)$                   (5)

Here, ${{C}_{tx}}$ denotes the transaction processing cost incurred by the blockchain network at time T, ${{c}_{FTL}}$ denotes the unit cost to process one transaction in the FTL method, ${{c}_{SAL}}$ denotes the unit cost to process one transaction in the SAL method, $n_{tx}^{FTL}$ denotes the number of all transactions recorded on the FTL's ledger as of Time T, $n_{tx}^{SAL}$ denotes the number of all transactions recorded on the SAL's ledger as of time T. A latency penalty may be included in the system objective as shown in Eq. (6).

${{J}_{lat}}\left( t \right)=\lambda \cdot {{\tau }_{FTL}}\left( t \right)$                          (6)

Here, ${{J}_{lat\left( t \right)}}$ denotes the latency penalty incorporated into the total system objective function of the system as of time t. λ represents a weighting factor signifying the magnitude by which minimizing latency is relative to the importance of other optimization objectives like cost, trust, or energy usage.

Energy matching variables are represented as energy traded from seller i to buyer j. Multi-objective matching optimization with the scalarized objective function is given as Eq. (7).

$J(t){{\ }_{{{x}_{ij}}\left( t \right)}}=\sum\limits_{i,j}{(\alpha {{p}_{ij}}(t)+\beta {{l}_{ij}}-\gamma {{\rho }_{i}}(t)){{x}_{ij}}(t)}+{{n}_{trades}}{{C}_{tx}}$                    (7)

Here, the objective combines three competing goals: minimizing the total payment made by buyers, reducing network-related energy losses, and maximizing trust in the trading process. The weighting coefficients $\alpha $, $\beta $, and $\gamma $ allow system operators to tune the relative importance of economic efficiency, physical network performance, and participant trustworthiness. The term ${{n}_{trades}}{{C}_{tx}}$ accounts for blockchain transaction overhead, encouraging efficient batching and reduced ledger load. The first supply constraint is represented as Eq. (8).

$\sum\limits_{j}{{{x}_{ij}}}\left( t \right)\le {{s}_{i}}\left( t \right),\forall i\in {{P}_{s}}$                        (8)                        

Here, ${{P}_{s}}$ denotes all prosumers in the marketplace, ${{x}_{ij}}\left( t \right)$ denotes the volume of electricity sold by prosumer i to consumer j at time t, and ${{s}_{i}}\left( t \right)$ denotes Surplus electricity of prosumer i available at time t. Demand constraint is shown in Eq. (9).

$\sum\limits_{i}{{{x}_{ij}}}\left( t \right)\le {{d}_{j}}\left( t \right),\forall j\in {{P}_{b}}$                    (9)

Here, ${{P}_{b}}$ denoted as all consumers in the marketplace, ${{d}_{j}}\left( t \right)$ denotes the demand for electricity requested by consumer j at time t. The transmission capacity must be as Eq. (10).

${{x}_{ij}}\left( t \right)\le {{T}_{ij}}$                (10)

Here, ${{T}_{ij}}$ denoted as the maximum amount of electricity that can be transferred from prosumer i to consumer j due to limitations of the electrical power distribution network. ZKP enforcement must be as shown in Eq. (11).

${{x}_{ij}}\left( t \right)\le M\cdot {{\epsilon }_{ZK,i}}\left( t \right)$                       (11)

Here, ${{\epsilon }_{\left( ZK,i \right)}}\left( t \right)$ indicates whether a zero-knowledge proof has been validated at time t for seller i. M denotes a value intended to restrict energy trades through a ZKP-based methodology using a big-M approach. The budget constraint is considered as Eq. (12).

$\sum\limits_{i}{{{p}_{ij}}}\left( t \right){{x}_{ij}}\left( t \right)\le {{B}_{j}}\left( t \right)$                    (12)

Here, ${{B}_{j}}\left( t \right)$ refers to the largest potential monetary amount that buyer j has access to when acquiring energy at time t. It ${{p}_{ij}}\left( t \right)~$denotes the price (in dollars) of electricity prosumer i sells to consumer j at time t. This formulation results in a linear program (LP) or mixed-integer LP (MILP). As the next layer benefits from continual reinforcement learning (RL-based) and ongoing graph optimizations, it maintains market stability by curbing volatility and increasing fairness; it also removes the potential for fraudulent activity. The trust update rule is shown in Eq. (13).

${{\rho }_{i}}\left( t+1 \right)=clip\,\left( {{\rho }_{i}}\left( t \right)+\eta \Delta {{\rho }_{i}}\left( t \right),\,0,\,1 \right)$                         (13)

Here, ${{\rho }_{i}}\left( t \right)$ is the trust or reputation score of seller i at time t. This score will be between 0 and 1.$~{{\rho }_{i}}\left( t~+~1 \right)$ is the updated trust score for seller i, at a future time slot. $clip\left( \cdot ,0,1 \right)$ is a bounding function that limits the trust score to be between 0 and 1. η is the trust learning rate that controls the responsiveness of trust updates due to seller actions. $\Delta {{\rho }_{i}}\left( t \right)~$is the incremental increase or decrease in trust for seller i, based on actions taken by seller i, at time t.

$\Delta {{\rho }_{i}}\left( t \right)={{w}_{1}}{{1}_{{on~time}}}+{{w}_{2}}{{\epsilon }_{ZK,i}}\left( t \right)-{{w}_{3}}{{1}_{anomaly}}$                        (14)

Here$,~{{w}_{1}},~{{w}_{2}},~and~{{w}_{3}}$ are weighting parameters used to denote how much weighting to give to positive and negative seller actions. ${{1}_{on ~ time}}~$is the indicator function and is 1 if the seller delivered and settled the energy on time and 0 if they did not. ${{\epsilon }_{\left( ZK,i \right)}}\left( t \right)$ is an indicator function that is the outcome of zero-knowledge proof validation for seller i at time t. ${{1}_{anomaly}}$ is the indicator function and is 1 if there is an anomalous or fraudulent action detected from seller i and 0 otherwise. ${{m}_{i}}\left( t \right)$ is the actual meter reading (metered consumption or generated) of seller i at time t. The smart meter records the reading and generates a random nonce; a cryptographic commitment is computed using Eq. (15).

${{h}_{i}}\left( t \right)=H\left( {{m}_{i}}\left( t \right)\parallel {{r}_{i}}\left( t \right) \right)$                         (15)

Here, Hash(⋅) denotes the cryptographic hash function SHA-3, used so that no two data sets can produce the same hash value. ${{h}_{i\left( t \right)}}$ denotes the hashed commitment of the meter reading broadcast to the blockchain. R denotes the range of valid meter readings, defined by the system or regulatory authorities. The seller improves in zero knowledge without revealing$~{{m}_{i}}\left( t \right)$ as shown in Eq. (16).

${{m}_{i}}(t)\in RandH({{m}_{i}}(t)\parallel {{r}_{i}}(t))={{h}_{i}}(t)$                       (16)

This is implemented using a range-proof ZKP protocol as shown in Eq. (17).

${{\pi }_{i}}\left( t \right)=ZKProve\left( {{m}_{i}}\left( t \right),{{r}_{i}}\left( t \right),{{h}_{i}}\left( t \right) \right)$                      (17)

Then the blockchain validators check the committed reading and approve using Eq. (18).

$Verify\left( {{\pi }_{i}}\left( t \right),{{h}_{i}}\left( t \right) \right)\to {{\epsilon }_{ZK,i}}\left( t \right)\in \left\{ 0,1 \right\}$                      (18)

Here, ${{\epsilon }_{ZK,i}}\left( t \right)$ zero shows the proof rejected, and ${{\epsilon }_{ZK,i}}\left( t \right)$ one shows the proof accepted. The result is stored on-chain and used in the trust update model. The dynamic pricing model is given as Eq. (19).

${{p}_{ij}}\left( t \right)={{p}^{base}}\left( t \right)+\kappa \,Con{{g}_{ij}}\left( t \right)-\delta \,{{\rho }_{i}}\left( t \right)+\phi \,\left( \frac{D\left( t \right)-S\left( t \right)}{S\left( t \right)+\varepsilon } \right)$                         (19)

Here, ${{p}^{base}}$ denote the base price, $\kappa $ denote the congestion price scaling coefficient, $\delta $ denote the trust-based price adjustment coefficient, and $\phi $ denote the surge pricing sensitivity coefficient. S(t) is the total energy supply in the market, $\varepsilon $ let the small constant, and D(t) is the total energy demand in the market. The reinforcement learning reward is computed using the various factors as shown in Eq. (20).

$R\left( t \right)={{\omega }_{1}}UG\left( t \right)-{{\omega }_{2}}APV\left( t \right)+{{\omega }_{3}}NR\left( t \right)-{{\omega }_{4}}VP\left( t \right)$                     (20)

Here, ${{\omega }_{1}},{{\omega }_{2}},{{\omega }_{3}},{{\omega }_{4}}$ denotes the weighing coefficients balancing the objectives, $UG\left( t \right)$ denotes the ratio of successfully traded energy, APV(t) denotes the average deviation of prices from the mean price across buyers, $NR\left( t \right)~$denotes the net economic revenue generated by the market over time. $VP\left( t \right)~$denotes the penalty term for constraint violations, instability, and unfair behavior. The RL agent's maximums are given as Eq. (21).

$\sum\limits_{t}{{{\gamma }^{t}}}R\left( t \right)$                    (21)

Here, $\gamma $ denotes the discount factor, R(t) is the immediate reward received by the RL agent. The fairness penalty is the quadratic penalty measuring unmet demand to promote the energy distribution, computed using Eq. (22).

$FPen\left( t \right)=\zeta \sum\limits_{j}{{{\left( \left( 0,{{d}_{j}}\left( t \right)-\sum\limits_{i}{{{x}_{ij}}}\left( t \right) \right)\  \right)}^{2}}}$                        (22)

Here, $\zeta $ denotes the fairness weighting coefficient, ${{d}_{j}}\left( t \right)$ denotes an energy demand requested by the buyer, and denotes an energy allocated from the seller to the buyer.

The framework of PBET-Chain uses mathematical formulations to develop models to support decentralized P2P energy trading, blockchain-based settlement, estimation of trust, and price optimization. As part of the PBET-Chain architecture, this section outlines the symbols used to illustrate the PBET-Chain equations that govern the behavior of energy price flows through the network, how transactions are verified on the blockchain, and how the PBET-Chain determines what will be sold by the seller to buyers in the market. The PBET-Chain framework for trading distributed energy incorporates all elements necessary for providing a secure, efficient, and accountable marketplace for users and buyers to exchange energy. Each of the five functional layers supports and integrates with the others. Smart meters provide accurate data; the blockchain establishes secure transactions; the smart contract enables automated trades; the AI-driven optimizer improves trading efficiency; and the privacy & security components maintain confidentiality and safeguard the integrity of the marketplace. PBET-Chain is built on the foundation of being able to grow, evolve, and achieve maximum transparency while supporting a decentralized P2P trading experience for all users.

PBET-Chain is an innovative framework built on a blockchain to facilitate P2P energy trading. It uses reinforcement learning to develop dynamic pricing structures, find trusted traders, and verify electricity usage. Experimental testing of PBET-Chain on a modified IEEE 34 bus microgrid. It indicates PBET-Chain's dual-layer blockchain architecture and intelligent market coordination enable significant improvements in efficiency and economic performance. The findings of this study show that PBET-Chain achieved 92.4% market efficiency and 88.7% renewable energy use efficiency, outperforming traditional blockchain and RL-based trading methodologies. The DPoA used in PBET-Chain ensures low latency time for processing transactions, supports high transaction throughput, and facilitates fast contract execution. This allows for near-real-time transaction processing in the decentralized energy trading market. The use of RL-driven pricing agents has demonstrated that PBET-Chain's pricing structure is stable while enabling an increase in supply and demand proximity and that the use of an integrated trust engine increased the reliability of prosumer identification by 94.8%, thus reducing fraudulent participation and increasing fairness by 17% when compared to random matching methods.

The continued development will include research into resilience to adversarial attacks as well as future development opportunities related to predictive analysis and anomaly detection using enhanced machine learning models. Energy is being traded and exchanged between microgrids. There is potential for microgrids to work together by connecting through an integrated energy network that can be connected to a national marketplace. Also allowing the application of privacy-preserving federated learning techniques to assist with decentralized pricing intelligence. Future research will enable PBET-Chain to be developed from a successful high-quality research prototype into a comprehensive, market-ready solution for decentralized energy distribution.