Efficient Simulator for P2P Energy Trading: Customizable Bid Preferences for Trading Agents

Abstract

Given the accelerating global movement towards decarbonization, the importance of promoting renewable energy (RE) adoption and ensuring efficient transactions in energy markets is increasing worldwide. However, renewable energy sources, including photovoltaic (PV) systems, are subject to output fluctuations due to weather conditions, requiring large-scale backup power to balance supply and demand. This makes trading electricity from large-scale PV systems connected to the existing grid challenging. To address this, peer-to-peer (P2P) energy markets where individual prosumers can trade excess power within their local communities have been garnering attention. This study introduces a simulator for P2P energy trading, designed to account for the diverse behaviors and objectives of participants within a market mechanism. The simulator incorporates two risk aversion parameters: one related to transaction timing, expressed through order prices, and another related to forecast errors, managed by adjusting trade volumes. This allows participants to customize their trading strategies, resulting in more realistic analyses of trading outcomes. To explore the effects of these risk aversion settings, we conduct a case study with 120 participants, including both consumers and prosumers, using real data from household smart meters collected on sunny and cloudy days. Our analysis shows that participants with higher aversion to transaction timing tend to settle trades earlier, often resulting in unnecessary transactions due to forecast inaccuracies. Furthermore, trading outcomes are significantly influenced by weather conditions: sunny days typically benefit buyers through lower settlement prices, while cloudy days favor sellers who execute trades closer to their actual needs. These findings demonstrate the trade-off between early execution and forecast error losses, emphasizing the simulator’s ability to analyze trading outcomes while accounting for participant risk aversion preferences.

1. Introduction

With the global movement towards decarbonization, the number of distributed energy resources (DERs) installed worldwide is increasing year by year. For example, 167 GW of distributed photovoltaic (PV) systems were installed globally between 2019 and 2021. However, the combined peak output now exceeds the total peak consumption of France and Britain [1]. As more DERs are connected to the grid, maintaining voltage and frequency stability becomes increasingly challenging due to the intermittent nature of many renewable energy sources [2].

To address the challenges of integrating DERs into the grid, peer-to-peer (P2P) energy trading is expected to play a key role by allowing participants to autonomously manage supply and demand fluctuations within their communities [3]. Specifically, P2P energy trading enables participants to trade excess energy generated by DERs flexibly among themselves, helping to balance local supply and demand more effectively within microgrids [4]. The objective of this study is to introduce a market simulator for peer-to-peer (P2P) energy trading, accounting for the diverse risk preferences of market participants.

Several approaches have been proposed for implementing P2P energy trading, including methodologies such as game theory [5,6,7] and distributed optimization [8,9,10], as well as technological solutions like blockchain technology [11,12,13]. A market mechanism is also one of the approaches. Among various possible matching mechanisms [14], Continuous Double Auction (CDA), which matches orders based on price priority and time priority (details are explained in Section 2.2 later) is one of the commonly used methods [15]. Liaquat et al. [16] designed a day-ahead CDA-based P2P market structure that considers network losses and network utilization fees. On the other hand, Li et al. [17] showed that in P2P energy trading using CDA, profits could be effectively and fairly distributed among participants even without complex trading strategies or learning abilities.

To investigate the feasibility of P2P energy trading, various proof-of-concept projects have been conducted. From June 2019 to August 2020, an experiment at Toyota’s Higashi-Fuji Research Institute and the surrounding area verified a system that autonomously conducted power trading between houses, businesses, and electric vehicles using blockchain technology [18,19]. Another notable project, the Urawa-Misono demonstration, ran from August 2019 to March 2020 in Japan’s Urawa-Misono district. This project involved solar farms, convenience stores, and residences with battery storage to promote flexible and efficient energy trading using blockchain technology [20]. In the Swiss case, 35 households and 2 commercial entities participated in the real-world P2P energy market, demonstrating high engagement and diverse pricing preferences [21,22]. While physical implementation is essential for testing the feasibility of P2P energy trading, conducting these experiments or offering real services requires a significant amount of cost and time. Therefore, analyzing transaction results through simulations is crucial for evaluation before system implementation, which motivates us to develop an efficient and realistic simulator in this study.

Several studies have been conducted on P2P energy trading simulation to analyze its market mechanisms. Khorasany et al. proposed a decentralized P2P energy trading scheme for energy markets with a high penetration of distributed energy resources, which maximizes welfare while minimizing data exchange and achieving faster convergence [23]. Angaphiwatchawal et al. [24] introduced a multi-k double auction pricing mechanism that provides more flexibility in setting exchange prices. Pankiraj et al. [25] proposed a double-sided auction mechanism with a matching algorithm that maximizes monetary benefits while considering the physical constraints of the grid infrastructure. Xu et al. [26] presented two truthful double-auction mechanisms, VCG and LPBP, which incentivize truthful bidding and maximize social welfare, respectively. Additionally, Xu et al. [27] proposed an iterative uniform-price auction mechanism that effectively promotes fair trading prices, community savings, and local PV energy transactions in community microgrids. Sampath et al. [28] proposed a decentralized P2P energy trading market framework that guarantees grid constraint satisfaction, market equilibrium, and global optimality, while preserving prosumer privacy and autonomy. Finally, Chen et al. [29] studied the PISO model, which optimizes prosumer operations and trading strategies in Continuous Double Auction-based P2P markets, facilitating the integration of flexible resources into energy markets.

The aforementioned studies aim to adjust the market matching mechanisms themselves or optimize the trading behavior of participants to improve the overall efficiency of the trading market. However, when considering realistic systems, the purposes and behaviors of participants vary, making it difficult for platform providers to accommodate all the trading behaviors desired by participants. Therefore, a simulator that can adjust trading behavior for each participant is required. In this research, we develop a P2P energy trading simulator that captures the varied behaviors and goals of participants within a market mechanism. Our simulator includes two parameters for risk aversion: one addressing the timing of transactions, reflected in order pricing, and another concerning forecast errors, controlled by modifying trade volumes. These parameters allow participants to tailor their trading strategies, enabling a more realistic assessment of trading results. To evaluate the impact of these risk aversion settings, we conduct a case study involving 120 participants, a mix of consumers and prosumers, using actual household smart meter data recorded on both sunny and cloudy days. Our analysis reveals that participants with a higher aversion to timing uncertainties tend to finalize trades earlier, often resulting in unnecessary transactions due to inaccuracies in forecasts. Additionally, trading outcomes are heavily influenced by weather conditions: sunny days generally benefit buyers with lower settlement prices, whereas cloudy days tend to favor sellers who transact closer to their real-time needs. These insights highlight a trade-off between early execution and forecast error losses; participants who prioritize early transactions to mitigate the risk of missed opportunities may incur greater losses due to forecast inaccuracies, while those who wait for more precise information reduce forecast-related risks but increase the chance of unexecuted trades. This underscores the simulator’s effectiveness in analyzing trading outcomes while considering the risk aversion preferences of participants.

This paper is organized as follows: Section 2 presents the proposed simulator, detailing its components, including the energy trading market and bidding agents, along with the overall flow of the simulator. Section 3 describes the implementation of the simulator, covering the programming environment, key libraries and technologies used, and the step-by-step process flow of its operation. Section 4 outlines the case study conducted in this research, including the setup, scenarios, and parameters used, and explains the results of the simulation. Section 5 provides a comprehensive discussion, addressing real-world strategies and directions for future work. Finally, Section 6 concludes the paper with a discussion of the findings and potential directions for future work.

2. Problem Specification

As P2P energy trading continues to gain traction as a decentralized solution for balancing supply and demand in local energy markets, it is essential to understand the structure and key components that define these transactions. In a P2P energy market, trading participants, ranging from pure consumers to prosumers with DERs, engage in energy exchanges autonomously, without relying on central intermediaries. These participants use specialized trading agents that interact within a dynamic market environment, executing trades based on their predefined strategies and risk preferences. In this section, we describe the setup of the P2P energy market, focusing on how these agents facilitate energy transactions.

2.1. P2P Energy Trading Agent

The relationship between trading agents in the P2P energy trading market is illustrated in Figure 1. P2P energy trading agents are software systems that autonomously manage energy transactions on behalf of their owners. The behavior of these agents varies depending on the devices their owners possess. These agents universally aim to optimize their owners’ gains by executing transactions according to predefined trading strategies. The specifics of these strategies differ based on the type of participant: for example, pure consumers or prosumers with only solar power generation have their order amounts determined primarily by demand and supply forecasts. However, for participants equipped with storage batteries, the order amounts can be optimized to maximize self-consumption by solving an optimization problem.

In the proposed P2P energy trading system, alongside the general trading participants, there is also a Grid Agent that acts as the system operator. This agent places sell orders at the market’s highest price and buy orders at the market’s lowest price, functioning as the final trading counterpart when no matching orders are available in the P2P market. As a result, other participants conduct their transactions within the price range established by the Grid Agent’s buy and sell prices.

2.2. P2P Energy Trading Market

In the P2P energy trading market designed in this paper, electricity is traded in 30 min intervals, with 48 products available for trading each day as shown in Figure 2. Orders for these 30 min delivery periods are accepted starting N hours before the delivery time. Since orders for deliveries are accepted from N hours in advance, the number of markets available for trading simultaneously, including the market for immediate delivery, is $2N+1$. This means that during each trading session, participants can place orders in both the real-time market, where delivery is imminent, and in multiple future delivery markets.

In this study, the market simulator for P2P energy trading is constructed using PV-generated energy and grid power. Notably, our simulator allows trades to be specified by tagging power sources or more granular regions. In such cases, multiple markets for different power sources or individual regions can be opened simultaneously.

In the simulator, matching and execution in the P2P energy market are carried out using the Continuous Double Auction (CDA) mechanism. In this method, buy and sell orders from multiple participants are accepted simultaneously, and matching occurs based on the most favorable conditions. If a buy order is placed at a price equal to or higher than the lowest sell order on the order book, the trade is executed. Similarly, if a sell order is placed at a price equal to or lower than the highest buy order, the trade is executed. Orders that are not immediately matched remain on the order book. At all times, the price–time priority rule is applied: higher buy orders and lower sell orders take precedence. For orders at the same price, the one placed earlier in the market is prioritized. This price–time priority rule is applied to each electricity delivery product, and participants are notified when a trade is executed.

2.3. Customizable Bid Preferences

Regarding the determination of trading preferences, the proposed energy trading agent accounts for risk aversion to settlement timing by adjusting the order price for the target market, while risk aversion to forecast errors is reflected by adjusting the order quantity. Further details are provided in the following sections.

2.3.1. Risk Aversion to Settlement Timing

As a premise for determining order prices, it is assumed that when placing an order in the N-hour-ahead market, there is a chance of settlement, encouraging a more ambitious price setting at the beginning. Specifically, for buyers, the strategy involves initially aiming for settlement at the lowest possible price, while gradually increasing the bid price over time to maximize the chance of settlement by the end. Conversely, for sellers, the strategy is to aim for settlement at the highest possible price when the market opens, gradually lowering the offer price as time progresses to secure a contract.

In other words, if matching with general market participants cannot be achieved, trading with the Grid Agent will occur, resulting in an undesirable settlement for the trader. Therefore, the risk mentioned here represents the risk of not settling with general market participants, and the higher the risk aversion, the more the participant prioritizes settling with entities other than the Grid Agent.

The price determination formula that considers risk aversion to settlement timing is shown in (1):

(1)$P_x^{buy}=(P^{max}-P^{min})·{(max(0,1-{\displaystyle \frac{x}{C_{buy}}}))}^2+P^{min}$

where $P^{max}$ represents the highest market price, and $P^{min}$ represents the lowest market price. x is an integer indicating how many markets ahead the order is placed for the target market. $C_{buy}$ is a parameter that adjusts the risk aversion level, which is the degree to which a participant aims for a certain transaction even at the expense of their own profitability; the larger the $C_{buy}$, the higher the risk aversion. The resulting price changes are shown in Figure 3.

From the figure, it can be seen that for future markets, the price is higher when the value of $C_{buy}$ is large, and it adheres closely to the lowest market price when the value is small. Near the real-time market (x = 0), the change is gradual with a high value of $C_{buy}$, and sharp with a low value of $C_{buy}$. Regardless of the value of $C_{buy}$, the price ultimately reaches the highest market price $P^{max}$ in the real-time market to ensure a buy settlement is made.

Next, the selling price determination is based on the following formula:

(2)$P_x^{sell}=P^{max}-(P^{max}-P^{min})·{(max(0,1-{\displaystyle \frac{x}{C_{sell}}}))}^2$

The basic parameters are the same as for purchases, with $C_{sell}$ serving as the coefficient to adjust the price variation. The resulting changes obtained from (2) are illustrated in Figure 4. The figure is a mirror image of the price changes for purchases, showing that with a small value of $C_{sell}$, the price change is gradual, while with a large value, the change is sharp. Regardless of the value of $C_{sell}$, the price ultimately reaches the lowest market price $P^{min}$ in the real-time market to ensure a sell settlement is made.

2.3.2. Risk Aversion to Forecast Errors

In this simulator, actual data from smart meters, such as observed generation or consumption data, are used to validate P2P energy trading. In the simulation, each agent is assumed to place bids based on actual generation or consumption data. If bidders are allowed to directly refer to actual data, it would correspond to a situation where future electricity consumption or generation could be perfectly predicted without any forecast errors. However, such a scenario is unrealistic, as it assumes the perfect foresight of future events. Therefore, while actual data are used as input in this simulator, it is assumed that prediction errors exist when these values are used for bidding purposes.

Specifically, prosumers who sell generation are assumed to have prediction errors in their generation forecasts, and consumers who buy electricity are assumed to have prediction errors in their consumption forecasts. It is further assumed that each participant has a risk aversion preference related to these prediction errors. Moreover, this simulator allows for the setting of parameters to account for such risk aversion.

Given the smart meter data, the estimated supply and demand are derived from the realized generation and consumption data, adjusted for prediction errors. For simplicity, the forecast errors considered in this study have a maximum error range that increases linearly over time, with errors randomly distributed within this range. It is also assumed that no error exists at the time of delivery. In the proposed approach, the order quantity varies based on the evaluation of forecast errors during order placement. If settlement occurs prematurely despite the presence of errors, this may result in unnecessary energy trades, highlighting the importance of accurate error estimation and risk management in bid strategies.

Despite this risk, if the price is favorable to participants, pursuing a settlement remains a viable option. Figure 5 illustrates the forecast error, showing how the forecast value at time t evolves during the prediction process. As time approaches t, the error range narrows, ultimately converging to an error-free value. If a settlement is established at time $t-5$, it may lead to unnecessary purchases of demand, necessitating a sell order at a later stage. However, if the purchase at time $t-5$ occurs at a low price, the transaction could still be advantageous overall. In this study, the risk aversion to forecast errors is expressed by adjusting the order amount based on the forecast error. The order amount at that time is expressed by (5).

(3)$\begin{array}{cc}\hfill coe{f}_0& ={\displaystyle \frac{\alpha P}{M}}·(M-x)\hfill \end{array}$

(4)$\begin{array}{cc}\hfill coef& =max(0,min(1,coe{f}_0))\hfill \end{array}$

(5)$\begin{array}{cc}\hfill A_{order}& =A_{pred}·coef\hfill \end{array}$

Here, M is a predefined value that sets the maximum range into the future that can be traded, x represents the number of the future delivery market to order, $\alpha$ is a parameter that adjusts the rate of change of the coefficient, and P represents the risk aversion level set by the participant (assumed to be between 1 and 10). $A_{pred}$ denotes the forecast amount including the error. The error amount is randomly generated within the range of the forecast error at the time of order placement explained in Figure 5.

The bid quantity $A_{order}$ is updated at each time step, and once a transaction is executed, a new bid must be placed to make any changes. Let the new bid quantity be $A_{order,new}$; then the actual bid amount is given by $A_{order,new}-A_{order}$. If a transaction is executed, the executed quantity is updated to $A_{order,new}$ at that time. Assuming that the sign of the bid quantity for buying is positive, a negative value of $A_{order,new}-A_{order}$ indicates a sell bid. Additionally, at time t, right before delivery, $A_{order,new}$ should match the actual demand.

In the case of a consumer, buy bids are placed to meet the required demand, but if the consumer engages in selling, this constitutes a transaction in the opposite direction of the intended purpose. The larger the total volume of such sell transactions, the more it can be considered excess trading. In this paper, we calculate the total amount of sell transactions executed by consumers (i.e., the total amount of energy sold). Specifically, the total sum of sell transactions executed by consumers is defined as the “redundant trading volume for buyers”.

On the other hand, for prosumers on the selling side, the natural direction of transactions should be selling. However, when the executed sales quantity is too large, they adjust by placing buy bids. Such a series of buy transactions by the seller can be considered to be redundant trades. In this study, we define “redundant trading volume for sellers” as the total amount of buy transactions executed by prosumers, who are primarily sellers of electricity.

2.4. Focus on Scenarios Without Energy Storage Systems

In this study, we focus on P2P energy trading scenarios, where prosumers manage supply and demand adjustments through interactions with the grid, without utilizing energy storage systems for charging and discharging. This approach reflects practical conditions in regions where energy storage adoption is limited due to regulatory and economic factors. For example, in Japan, regulatory constraints limit the use of stored electricity to self-consumption; electricity stored in batteries cannot be sold back to the grid or to other consumers without permission [30]. Moreover, the high initial investment required for energy storage installations often poses a barrier for many participants. By focusing on scenarios without energy storage, we aim to provide valuable insights into participant behaviors and market dynamics in settings where storage systems are less prevalent. Additionally, we aim to examine changes in trading volume driven by bidding strategies, so we concentrate on systems where supply and demand are influenced directly by immediate consumption and generation, without the time shifts caused by charging and discharging schedules of energy storage systems.

3. Implementation of the Simulator

In this section, we explain the implementation of the P2P energy trading simulator developed for this study. The simulator was designed to validate various bidding strategies and trading behaviors in a P2P energy market. This section details the development and execution environment, references previous studies that informed the simulator’s design, and presents a numerical experiment of a case study based on an actual smart meter.

3.1. Development Environment

The proposed simulator was developed using Python, a widely used programming language suitable for data processing and simulation. Key Python libraries used in the development include pandas for data manipulation, numpy for numerical calculations, and threading for managing parallel processing tasks. These libraries enabled the efficient handling of data, computational performance, and the implementation of complex simulation processes. The simulator runs in a virtualized environment using Docker [31], ensuring consistency and portability across different platforms. Participant data, such as demand and supply information, are stored and managed using a PostgreSQL database [32]. Additionally, bash shell scripts were employed to automate specific tasks and ensure smooth execution.

3.2. Execution Environment

The simulator was executed on a MacBook Air (2022 model, 16GB RAM, M2 chip 8-core CPU, manufactured by Apple Inc., Cupertino, California, United States), which provided sufficient computational resources to run the simulations. Docker was used to create a consistent and isolated environment, ensuring the simulator ran effectively across different systems. The MacBook Air was chosen for its balance of performance and versatility for general-purpose use. In the case study discussed later (Section 4), it took approximately 8 s to run a one-day simulation.

3.3. Simulator Development Process

The development of the agent-based simulator was informed by previous studies, particularly the work by [33]. This reference provided critical insights into the construction of bidding agents and their interactions in a P2P trading system, serving as a foundation for the development of the simulator used in this study.

3.4. Flow of the Simulator

The flow of the proposed simulator is illustrated in Figure 6. Initially, before starting market transactions, participant data such as demand, supply, and order strategy information are loaded from the database.

As mentioned previously, this simulator allows participants to trade $2N+1$ products and place both sell and buy orders during each time period. Participants also have the option to withdraw previously submitted orders before placing new ones. This feature helps prevent unnecessary transactions or allows the cancellation of orders when solving optimization problems, and can be configured individually for each agent.

Once orders are placed, the bidding behavior of each participant is adjusted according to their characteristics. For instance, pure consumers or prosumers with only solar power generation have their order amounts determined by demand and supply predictions. However, participants with storage batteries can optimize their orders to maximize self-consumption by solving an optimization problem. To replicate realistic scenarios, prediction errors are introduced during the ordering process. In the case study, we analyze the impact of these prediction errors on trading outcomes. Next, the matching process is carried out. If a trade is executed, the results are recorded, and the relevant participants’ trading data are updated. Additional post-processing steps, such as updating the state of charge (SoC) of storage batteries, can also be performed as necessary.

This trading process is repeated for a predetermined duration, and once completed, all trading results are recorded, concluding the simulation.

In actual P2P energy trading using the CDA method, bids are prioritized by time, so it is rare for two bids to occur at exactly the same price and time. However, in the simulator, bids are placed at discrete time intervals, meaning that if bids are submitted at the same time, price will always be prioritized, making it difficult for time–priority matching to occur. To avoid this situation, the simulator first calculates only the bid prices (as shown in the “prepare orders” step in the flow), and the order in which bids are submitted is then assigned randomly. This allows time–priority matching to be applied even for bids placed at the same time.

4. Case Study

In the previous sections, we detailed the specification, development environment, and flow of the P2P energy trading simulator developed in this study. To validate the functionality and effectiveness of the simulator, we now conduct a case study to examine how different bidding strategies and behaviors affect overall trading outcomes under various market conditions. In particular, the analysis here focuses on how trading outcomes change depending on the level of risk aversion selected by the trading participants, both in terms of execution timing and prediction error, using the proposed simulator.

4.1. Simulation Setup

We first describe the specific setup used in the case study to analyze the P2P energy trading system. This setup involves defining the participants, their behaviors, and the environmental conditions under which the trades take place. The data used here are obtained from the smart meter data of 120 households across Chiba Prefecture, 80 of which are consumers (who only purchase electricity), and 40 are prosumers (who own PV systems and both purchase and sell electricity). The weather data are based on observations from the Choshi Local Meteorological Observatory. Sunny and cloudy days are selected from October 2021 data. The relationship between generation and demand on the sunny day is shown in Figure 7a, and that on the cloudy day is shown in Figure 7b. We select sunny and cloudy days for the data analyzed in this study because these weather conditions account for a significant portion of the annual weather in the target area. Upon investigating data provided by the Japan Meteorological Agency [34], based on average data observed from 1991 to 2020, we find that the area used in this study experiences approximately 102 days of precipitation annually. Note that “days of precipitation” refers to days when the daily precipitation amount is 1 mm or more. Therefore, about 72% of the year consists of sunny or cloudy weather. Furthermore, on rainy days, there is no electricity sales volume, making it impossible to examine differences in contracted electricity volumes resulting from the different bidding strategies employed in this study. For this reason, rainy days are excluded from the test data.

The Grid Agent places sell bids at JPY 29.62 and buy orders at JPY 10.57, values that are arbitrarily set but are consistent with those used in a previous experiment [35]. As a result, the former effectively becomes the market’s highest price and the latter the lowest, with transactions between consumers and prosumers occurring within this range.

For bidding strategies, each trading participant is randomly assigned a risk aversion level on a 10-point scale, covering both execution timing and prediction error. A value of 1 indicates low risk aversion, while a value of 10 indicates high risk aversion. For example, a participant with a risk aversion level of 10 for execution timing would place orders at a higher price in the future market, aiming for an early transaction. Similarly, a risk aversion level of 10 for prediction error would lead to cautious behavior, limiting the order amount until the exact energy requirement is known.

To verify the results, the risk aversion levels of all participants except for the selected consumer and prosumer are fixed at randomly assigned values. By varying the risk aversion levels of the two selected agents, we analyze how the trading outcomes change. Information on the energy demand of the selected consumer is shown in Figure 8a,b, while data on both the energy demand and supply of the prosumer are presented in Figure 9a,b.

For risk aversion concerning execution timing, participants are classified into three types: low risk aversion (prioritizing favorable prices), high risk aversion (prioritizing execution), and balanced. The risk aversion levels for these types are set to 1, 10, and 5, respectively. Similarly, for risk aversion regarding prediction error, participants are also divided into low risk aversion (prioritizing early execution), high risk aversion (avoiding unnecessary transactions), and balanced, with the same values of 1, 10, and 5. This results in a total of nine trading configurations. Figure 10 illustrates the settings used in this study, with the horizontal axis representing risk aversion for execution timing, and the vertical axis for prediction error. For instance, in the case of $A3\text{\_}P3$, where both risk aversion levels are high, the participant aims for an early transaction by placing buy orders at a higher price and limiting the order amount to account for prediction errors.

As previously mentioned in Section 2.3.2, the simulation in this study utilizes actual data from smart meter readings to provide a realistic foundation for analyzing P2P energy trading behavior. Initially, we assume ideal conditions with no forecast errors to establish a baseline scenario, wherein participants would theoretically know precise generation and consumption values in advance—an assumption that is unrealistic in practical applications. To address this, we introduce forecast errors as random deviations within a range of +30% to −30% to simulate more realistic prediction uncertainties. For bidding calculations, any negative values in the net difference between generation and demand are set to zero to ensure that surplus quantities remain non-negative.

Additionally, while the bid sequence itself may vary due to the random nature of forecast error generation, it should be noted that the trading outcomes are consistently determined by each participant’s specific bidding strategy as defined in our model. This ensures that, despite the inherent variability of forecast errors, the results remain primarily attributable to each predefined strategy rather than incidental randomness. Although multiple randomized scenarios could offer additional insights, our setup is effective for illustrating strategy-based outcomes within a controlled framework. In this way, the simulator allows us to examine the effects of forecast error on trading outcomes, illustrating the dynamics influenced by participant behavior.

4.2. Execution Results

The experiment is conducted using the settings described above. Figure 11a,b display box plots showing the trends in transaction prices for all agents, organized by time period. The horizontal axis represents time, and the vertical axis represents the transaction price per kWh. In this system, buy agents update their bid prices from lower to higher over time, while sell agents adjust their prices from higher to lower. This mechanism reflects the balance between supply and demand. When demand exceeds supply, sellers tend to secure transactions earlier, while when supply exceeds demand, buyers tend to complete transactions faster. As a result, prices rise when demand is high and fall when supply is abundant. This behavior is reflected in the different weather conditions observed. On sunny days, when supply exceeds demand, prices tend to decrease. Conversely, on cloudy days, since demand does not drop significantly but supply decreases, prices tend to rise.

4.3. Individual Agent Behavior

We will analyze how the trading behavior of an arbitrarily selected individual prosumer changes according to their level of risk aversion. As previously mentioned, risk aversion is classified into nine categories. This analysis focuses on the prosumer’s trading amounts on a sunny day, using market results from 10:00 to 14:00, when P2P energy trading is most active. Figure 12a shows the prosumer’s selling amounts on a sunny day. The horizontal axis represents the time relative to the actual power delivery, indicating how far in advance the transaction was executed, and the vertical axis shows the amount of energy traded at that time. From this figure, it can be observed that the bidding strategies of $A1\text{\_}P3$, $A2\text{\_}P3$, and $A3\text{\_}P3$, which aim for early execution, result in transactions being completed at an early stage (between 4:00 and 5:30). The transaction volume is highest for $A1$, which targets larger amounts despite higher prediction errors, and decreases progressively for $A2$ and $A3$. For other bidding strategies, transactions can be observed starting from 3:30, with a higher number of transactions occurring in the real-time market right before delivery. Additionally, Figure 12b, which shows the prosumer’s purchases on a sunny day, reveals that $A1\text{\_}P3$ and $A2\text{\_}P3$ engage in more purchasing. This is due to buyback transactions aimed at compensating for excess energy sold, highlighting the impact of significant prediction errors on these trades.

Next, we investigate how the revenue and the error in the final required amount of energy change when each trading agent selects different bidding strategies. The results for the prosumer on the sunny day are shown in Figure 13a, and the results for the cloudy day are shown in Figure 13b. In these figures, the vertical axis represents the total daily revenue, and the horizontal axis represents the redundant trading volume for sellers, as described in Section 2.3.2, caused by prediction errors. The bidding strategies shown in Figure 10 are plotted as scatter points.

From Figure 13a, it can be observed that $A1\text{\_}P3$ and $A2\text{\_}P3$ achieve the highest revenue despite having the largest error totals (the more negative the value, the higher the amount sold, which is a desirable outcome). This is because, on days like the sunny day when a large amount of energy is traded, if transactions are not matched among general trading participants, sellers may have to sell to the Grid Agent at the market’s lowest price. However, if transactions are completed early, it leads to more favorable outcomes for the sellers. Additionally, the fact that revenue remains relatively unchanged for $A1\text{\_}P3$ and $A2\text{\_}P3$ despite an error total of about 4 kWh is due to the benefits of early transactions offsetting the negative impacts of errors. For other bidding strategies, although there is no change in the error total, $A3\text{\_}P3$, which aims for early transactions, shows the next highest revenue. There is a trend of decreasing revenue as the transaction timing approaches the moment of power delivery.

The results for the prosumer on the cloudy day, as shown in Figure 13b, are discussed. From this figure, it is evident that the trend is the opposite of that seen in Figure 13a. Additionally, it can be observed that revenue improves as the error amount decreases (on cloudy days, the amount of energy sold is lower, and the amount paid exceeds the profit). This is because when the amount of energy traded in the P2P energy market is low, the impact of errors on revenue becomes more significant. In situations where sellers have an advantage, delaying the transaction until closer to the power delivery timing can maximize revenue.

The results for the consumer reveal the opposite trend compared to the prosumer results. Specifically, on sunny days (Figure 14a), the impact of prediction errors is significant, and the earlier the transaction is executed due to the bidding strategy, the worse the revenue becomes. This is because, on sunny days, there is a large amount of energy in the trading market, leading to a tendency for prices to drop. Therefore, it is preferable to execute transactions closer to the actual time when the energy is needed, rather than at an earlier stage. In contrast, on cloudy days (Figure 14b), it becomes clear that the early execution of transactions is important. This indicates that when trading volumes are low, as on cloudy days, securing transactions early is crucial.

5. Discussion

In the preceding sections, we demonstrated the construction of our P2P energy trading simulator and examined, via case study, how varying levels of risk aversion influence trading outcomes and behaviors of individual prosumers. Building on these insights, this section discusses potential extensions of our study’s framework and outlines directions for future research, with each topic addressed in the following subsections.

5.1. Extended Simulations and Enhanced Product Tagging

In this study, we focused on a single-day simulation and limited product tagging to analyze short-term participant behavior and trading dynamics. Future studies could extend the simulator to multi-day simulations to capture inter-day trading patterns and the cumulative effects of participant strategies over longer periods. Additionally, a broader range of product tags, such as specific energy source types or time-differentiated tags, could allow for a more granular analysis of trading preferences and behavior. These extensions would enable a more comprehensive understanding of P2P energy trading, including sophisticated market mechanisms that may emerge under varying conditions.

5.2. Future Work

Our current study focuses on P2P energy trading markets, primarily analyzing management aspects and participant behaviors. However, we recognize the importance of addressing technical and market-oriented challenges for the practical implementation and efficiency of these markets. Future research will be directed towards several key areas, including the integration of energy storage systems, market liquidity and stability, advanced grid management, and network topology solutions.

5.2.1. Integration of Energy Storage Systems

Incorporating energy storage systems presents a promising opportunity to enhance market efficiency and facilitate the integration of renewable energy sources. In this study, we do not include battery storage mainly because we want to analyze how changes in order amounts in response to demand and supply forecast errors vary depending on different bidding strategies. Introducing optimal control for charging and discharging timings would introduce additional complexities, making it difficult to evaluate the impact of bidding strategies alone.

Future work will focus on developing evaluation methods that can accommodate the optimal control of storage systems while still allowing us to assess how bidding strategies influence market behaviors. By addressing this challenge, we aim to provide a more nuanced understanding of the interplay between storage optimization and bidding strategies in P2P energy trading. Additionally, we plan to develop strategies for the optimal use of storage systems in P2P energy trading, focusing on their influence on market stability and participant behaviors.

5.2.2. Market Liquidity and Stability

Price stability is crucial for effective P2P energy trading. In our current model, price boundaries set by a Grid Agent mitigate price fluctuations, yet market outcomes vary with weather conditions, especially between consumers and prosumers. Future research will aim to improve market liquidity and stability by introducing new roles, such as market makers [36,37], to achieve more consistent trading results. We will also analyze the effect of price-setting mechanisms and the addition of liquidity providers on long-term market sustainability.

5.2.3. Advanced Grid Management

The accurate management and tracking of power flows between prosumers are essential for the effective operation of P2P energy markets. Future research will focus on exploring smart grid technologies and advanced monitoring systems that can manage and track these power flows precisely. By researching these technologies, we aim to understand how they can support decentralized trading without necessitating a complete overhaul of existing grid infrastructure. This includes examining how real-time data analytics, IoT devices, and automation can enhance grid responsiveness and reliability, thereby facilitating seamless energy exchanges between participants.

5.2.4. Network Topology Solutions

The physical and virtual network structures supporting P2P energy trading play a significant role in its scalability and accessibility. We plan to explore alternative network designs, such as microgrids or virtual networks, that can support P2P trading without requiring direct physical connections between all participants. Additionally, we will investigate the use of blockchain and other distributed ledger technologies to facilitate secure and efficient energy transactions. By leveraging these technologies, we aim to address challenges related to trust, transparency, and transaction costs, thereby promoting a more robust and user-friendly P2P energy market.

Through addressing these challenges, we aim to contribute to the development of a resilient, efficient, and accessible P2P energy market that promotes prosumer participation and supports the widespread adoption of renewable energy sources.

6. Conclusions

This study introduced a P2P energy trading simulator based on market mechanisms, allowing individual participants to adjust their trading behavior. The simulator was designed to account for the diverse objectives and behaviors of market participants by employing the CDA mechanism, in contrast to existing methods that often rely on platform operators to set a single settlement price. These conventional approaches tend to overlook the heterogeneity among participants as previously mentioned.

To model participant behavior, two types of risk aversion were incorporated. The first relates to transaction timing, expressed through variations in bid prices. The second concerns forecast errors, represented by limiting the trading volume until the actual time when energy is needed. By adjusting these risk aversion parameters for each participant, this study analyzed their effect on trading outcomes.

In the case study, 120 participants, consisting of 80 consumers and 40 prosumers, were analyzed using data from two distinct days: one sunny and one cloudy. To provide a deeper analysis, one consumer and one prosumer were selected to examine how different bidding strategies influenced trading behavior and revenue.

In terms of trading patterns, it was observed that participants with higher risk aversion to transaction timing tended to secure deals earlier. As a result, they were more susceptible to forecast errors, leading to unnecessary trades.

Regarding trading outcomes, significant differences were noted between consumers and prosumers depending on the weather. On sunny days, with a larger volume of energy traded, transaction prices tended to fall, benefiting buyers. Therefore, even in the presence of forecast errors, securing agreements early generally led to higher revenues. On the other hand, on cloudy days, lower trading volumes favored sellers. Finalizing deals closer to the time of energy delivery reduced the impact of forecast errors and typically improved revenue. Interestingly, the trends for consumers were opposite to those for prosumers.

These findings demonstrate that the proposed simulator is effective in analyzing trading outcomes while accounting for the diverse behaviors of market participants.

Author Contributions

Conceptualization, Y.T., Y.S., K.F., Y.Y. and K.T.; methodology, Y.T. and Y.Y.; software, Y.T.; validation, Y.T.; formal analysis, Y.T.; investigation, Y.T. and Y.Y.; data curation, Y.T.; writing—original draft preparation, Y.T.; writing—review and editing, Y.T. and Y.Y.; visualization, Y.T.; supervision, Y.Y. and K.T.; project administration, Y.Y. and K.T.; funding acquisition, Y.Y. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement

Data is unavailable due to privacy and ethical restrictions.

Conflicts of Interest

The authors declare no conflicts of interest. This study belongs to the authors and does not represent the views of the companies to which the first and second authors are affiliated.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures

Figure 1. Relationship between P2P energy trading agents, a Grid Agent, and the P2P energy trading market.

Figure 2. Trading products and example of multiple markets using tags. The diagram illustrates the structure of the P2P energy trading market in the simulator, where electricity is traded in 30 min intervals, with 48 products traded each day. Each product corresponds to a distinct time slot, starting at either 00 or 30 min (e.g., Product 1 at 00:00 and Product 2 at 00:30). In this example, two types of electricity, identified by Tag1 and Tag2, are traded separately within the same time slot, each managed through its own order book. The order books for Tag1 and Tag2 independently display buy and sell orders organized by price and quantity, with buy orders shown in red and sell orders in green, illustrating the separate handling of different electricity types. The matching of buy and sell orders is conducted using the CDA mechanism, where trades are matched based on price and time priority, promoting efficient and competitive trading. As each trading period advances, the available products shift forward to the next time slot, allowing participants to continuously bid on upcoming delivery periods. Each product may also have a suitable bidding window specific to its tag and time slot, reflecting variations in supply, demand, and market conditions.

Figure 3. Changes in buy price for each parameter [Forumla omitted. See PDF.].

Figure 4. Changes in sell price for each parameter [Forumla omitted. See PDF.].

Figure 5. Relationship between prediction error and settlement timing.

Figure 6. P2P energy trading simulator process flow.

Figure 7. Net demand of all participants.

Figure 8. Selected consumer’s net demand.

Figure 9. Selected prosumer’s net demand.

Figure 10. Risk aversion settings matrix: mapping low to high levels of prediction error and settlement timing.

Figure 11. Box plot of trading results.

Figure 12. Selected prosumer’s energy traded amounts by time gap between actual and settled trading timings. (a) Selling Amount on the Selected Sunny Day. (b) Buying Amount on the Selected Sunny Day.

Figure 13. Relationship between total cost and error power in energy trading: results for the selected prosumer.

Figure 14. Relationship between total cost and error power in energy trading: results for the selected consumer.

References

1. IEA. Executive Summary—Unlocking the Potential of Distributed Energy Resources—Analysis—IEA. Available online: https://www.iea.org/reports/unlocking-the-potential-of-distributed-energy-resources/executive-summary (accessed on 30 September 2024).

2. IEA. Net Zero by 2050. A Roadmap for the Global Energy Sector. 2021; Available online: https://www.iea.org/reports/net-zero-by-2050 (accessed on 30 September 2024).

3. Islam, S.N. A Review of Peer-to-Peer Energy Trading Markets: Enabling Models and Technologies. Energies; 2024; 17, 1702. [DOI: https://dx.doi.org/10.3390/en17071702]

4. Zhang, C.; Wu, J.; Zhou, Y.; Cheng, M.; Long, C. Peer-to-Peer energy trading in a Microgrid. Appl. Energy; 2018; 220, pp. 1-12. [DOI: https://dx.doi.org/10.1016/j.apenergy.2018.03.010]

5. Devi, N.N.; Thokchom, S.; Singh, T.D.; Panda, G.; Naayagi, R.T. Multi-Stage Bargaining of Smart Grid Energy Trading Based on Cooperative Game Theory. Energies; 2023; 16, 4278. [DOI: https://dx.doi.org/10.3390/en16114278]

6. Izanlo, A.; Sheikholeslami, A.; Gholamian, S.A.; Kazemi, M.V.; Hosseini, S.N. A combination of MILP and game theory methods for P2P energy trading by considering network constraints. Appl. Energy; 2024; 374, 123916. [DOI: https://dx.doi.org/10.1016/j.apenergy.2024.123916]

7. Lee, W.P.; Han, D.; Won, D. Grid-Oriented Coordination Strategy of Prosumers Using Game-theoretic Peer-to-Peer Trading Framework in Energy Community. Appl. Energy; 2022; 326, 119980. [DOI: https://dx.doi.org/10.1016/j.apenergy.2022.119980]

8. Aminlou, A.; Mohammadi-Ivatloo, B.; Zare, K.; Razzaghi, R.; Anvari-Moghaddam, A. Activating demand side flexibility market in a fully decentralized P2P transactive energy trading framework using ADMM algorithm. Sustain. Cities Soc.; 2024; 100, 105021. [DOI: https://dx.doi.org/10.1016/j.scs.2023.105021]

9. Han, B.; Zahraoui, Y.; Mubin, M.; Mekhilef, S.; Korõtko, T.; Alshammari, O. Distributed optimal storage strategy in the ADMM-based peer-to-peer energy trading considering degradation cost. J. Energy Storage; 2024; 96, 112651. [DOI: https://dx.doi.org/10.1016/j.est.2024.112651]

10. Wang, L.; Wang, Z.; Li, Z.; Yang, M.; Cheng, X. Distributed optimization for network-constrained peer-to-peer energy trading among multiple microgrids under uncertainty. Int. J. Electr. Power Energy Syst.; 2023; 149, 109065. [DOI: https://dx.doi.org/10.1016/j.ijepes.2023.109065]

11. Saini, V.K.; Purohit, C.S.; Kumar, R.; Al-Sumaiti, A.S. Proof of Work Consensus Based Peer to Peer Energy Trading in the Indian Residential Community. Energies; 2023; 16, 1253. [DOI: https://dx.doi.org/10.3390/en16031253]

12. Boumaiza, A. A Blockchain-Based Scalability Solution with Microgrids Peer-to-Peer Trade. Energies; 2024; 17, 915. [DOI: https://dx.doi.org/10.3390/en17040915]

13. Merrad, Y.; Habaebi, M.H.; Islam, M.R.; Gunawan, T.S.; Elsheikh, E.A.; Suliman, F.M.; Mesri, M. Machine Learning-Blockchain Based Autonomic Peer-to-Peer Energy Trading System. Appl. Sci.; 2022; 12, 3507. [DOI: https://dx.doi.org/10.3390/app12073507]

14. Kim, H.J.; Chung, Y.S.; Kim, S.J.; Kim, H.T.; Jin, Y.G.; Yoon, Y.T. Pricing mechanisms for peer-to-peer energy trading: Towards an integrated understanding of energy and network service pricing mechanisms. Renew. Sustain. Energy Rev.; 2023; 183, 113435. [DOI: https://dx.doi.org/10.1016/j.rser.2023.113435]

15. Capper, T.; Gorbatcheva, A.; Mustafa, M.A.; Bahloul, M.; Schwidtal, J.M.; Chitchyan, R.; Andoni, M.; Robu, V.; Montakhabi, M.; Scott, I.J. et al. Peer-to-peer, community self-consumption, and transactive energy: A systematic literature review of local energy market models. Renew. Sustain. Energy Rev.; 2022; 162, 112403. [DOI: https://dx.doi.org/10.1016/j.rser.2022.112403]

16. Liaquat, S.; Hussain, T.; Celik, B.; Fourney, R.; Hansen, T.M. Day-ahead continuous double auction-based peer-to-peer energy trading platform incorporating trading losses and network utilisation fee. IET Smart Grid; 2023; 6, pp. 312-329. [DOI: https://dx.doi.org/10.1049/stg2.12103]

17. Li, Z.; Ma, T. Peer-to-peer electricity trading in grid-connected residential communities with household distributed photovoltaic. Appl. Energy; 2020; 278, 115670. [DOI: https://dx.doi.org/10.1016/j.apenergy.2020.115670]

18. Takeda, Y.; Nakai, Y.; Senoo, T.; Tanaka, K. Designing a User-Centric P2P Energy Trading Platform: A Case Study—Higashi-Fuji Demonstration. Energies; 2021; 14, 7289. [DOI: https://dx.doi.org/10.3390/en14217289]

19. Matsuda, Y.; Yamazaki, Y.; Oki, H.; Takeda, Y.; Sagawa, D.; Tanaka, K. Demonstration of blockchain based peer to peer energy trading system with real-life used PHEV and HEMS charge control. Energies; 2021; 14, 7484. [DOI: https://dx.doi.org/10.3390/en14227484]

20. Kontani, R.; Tanaka, K.; Yamada, Y. Feasibility conditions for demonstrative peer-to-peer energy market. Energies; 2021; 14, 7418. [DOI: https://dx.doi.org/10.3390/en14217418]

21. Ableitner, L.; Tiefenbeck, V.; Meeuw, A.; Wörner, A.; Fleisch, E.; Wortmann, F. User behavior in a real-world peer-to-peer electricity market. Appl. Energy; 2020; 270, 115061. [DOI: https://dx.doi.org/10.1016/j.apenergy.2020.115061]

22. Wörner, A.M.; Ableitner, L.; Meeuw, A.; Wortmann, F.; Tiefenbeck, V. Peer-to-Peer Energy Trading in the Real World: Market Design and Evaluation of the User Value Proposition. Proceedings of the Fortieth International Conference on Information Systems; Munich, Germany, 15–18 December 2019; [DOI: https://dx.doi.org/10.3929/ethz-b-000395174]

23. Khorasany, M.; Mishra, Y.; Ledwich, G. A Decentralized Bilateral Energy Trading System for Peer-to-Peer Electricity Markets. IEEE Trans. Ind. Electron.; 2020; 67, pp. 4646-4657. [DOI: https://dx.doi.org/10.1109/TIE.2019.2931229]

24. Angaphiwatchawal, P.; Sompoh, C.; Chaitusaney, S. A Multi-k double auction pricing mechanism for peer-to-peer energy trading market of prosumers. Proceedings of the ECTI-CON 2021—18th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology; Chiang Mai, Thailand, 19–22 May 2021; pp. 473-476. [DOI: https://dx.doi.org/10.1109/ECTI-CON51831.2021.9454879]

25. Pankiraj, J.S.; Yassine, A.; Choudhury, S. Double-Sided Auction Mechanism for Peer-to-Peer Energy Trading Markets. Proceedings of the 2021 IEEE International Conference on Progress in Informatics and Computing, PIC 2021; Shanghai, China, 17–19 December 2021; pp. 443-452. [DOI: https://dx.doi.org/10.1109/PIC53636.2021.9687026]

26. Xu, S.; Zhao, Y.; Li, Y.; Cui, K.; Cui, S.; Huang, C. Truthful Double-auction Mechanisms for Peer-to-peer Energy Trading in a Local Market. Proceedings of the 2021 IEEE Sustainable Power and Energy Conference: Energy Transition for Carbon Neutrality, iSPEC 2021; Nanjing, China, 23–25 December 2021; pp. 2458-2463. [DOI: https://dx.doi.org/10.1109/iSPEC53008.2021.9735918]

27. Xu, S.; Zhao, Y.; Li, Y.; Zhou, Y. An iterative uniform-price auction mechanism for peer-to-peer energy trading in a community microgrid. Appl. Energy; 2021; 298, 117088. [DOI: https://dx.doi.org/10.1016/j.apenergy.2021.117088]

28. Sampath, L.P.M.I.; Paudel, A.; Nguyen, H.D.; Foo, E.Y.; Gooi, H.B. Peer-to-Peer Energy Trading Enabled Optimal Decentralized Operation of Smart Distribution Grids. IEEE Trans. Smart Grid; 2022; 13, pp. 654-666. [DOI: https://dx.doi.org/10.1109/TSG.2021.3110889]

29. Chen, K.; Lin, J.; Song, Y. Trading strategy optimization for a prosumer in continuous double auction-based peer-to-peer market: A prediction-integration model. Appl. Energy; 2019; 242, pp. 1121-1133. [DOI: https://dx.doi.org/10.1016/j.apenergy.2019.03.094]

30. Ministry of Economy, Trade and Industry. Reverse Flow Aggregation Utilization as a Regulation Power. Available online: https://www.meti.go.jp/shingikai/energy_environment/energy_resource/pdf/010_07_00.pdf (accessed on 30 September 2024).

31. Docker: Accelerated Container Application Development. Available online: https://www.docker.com/ (accessed on 30 September 2024).

32. PostgreSQL: The World’s Most Advanced Open Source Database. Available online: https://www.postgresql.org/ (accessed on 30 September 2024).

33. Takeda, Y.; Tanaka, K. Bidding Agent Model for P2P Energy Trading. IEEJ Trans. Ind. Appl.; 2020; 140, pp. 738-745. [DOI: https://dx.doi.org/10.1541/ieejias.140.738]

34. Japanese Meteorological Agency. The Average Values of Sunny Days and Rainy Days. Available online: https://www.data.jma.go.jp/cpd/longfcst/kaisetsu/tenkinissuu/tenkinissuu.html (accessed on 30 September 2024).

35. Sumitomo Mitsui Trust Bank, Ltd. Utilization of Storage Batteries for Large-Scale Introduction of Renewable Energy in Local Areas. Available online: https://www.smtb.jp/-/media/tb/about/sustainability/news/pdf/20240620.pdf (accessed on 30 September 2024).

36. Kuno, S.; Tanaka, K.; Yamada, Y. Effectiveness and Feasibility of Market Makers for P2P Electricity Trading. Energies; 2022; 15, 4218. [DOI: https://dx.doi.org/10.3390/en15124218]

37. Kanamura, T.; Bunn, D.W. Market making and electricity price formation in Japan. Energy Econ.; 2022; 107, 105765. [DOI: https://dx.doi.org/10.1016/j.eneco.2021.105765]