Virtual power plant price prediction method and device, equipment and medium

Abstract

The invention discloses a price prediction method and device for a virtual power plant, equipment and a medium, and belongs to the technical field of electric power, and the method comprises the steps: obtaining an actual electricity price sequence of a previous price prediction period when a current price prediction period is reached; calculating a basic reward based on the predicted electricity price sequence of the previous price prediction period, the actual electricity price sequence and a preset market fluctuation index; obtaining the number of competitors participating in the same prediction task in the previous price prediction period, determining the competition intensity according to the number of the competitors, and dynamically selecting a reward calculation strategy matched with the competition intensity according to a preset mapping relationship between the competition intensity and a reward strategy; adjusting the basic reward based on the matched reward calculation strategy to obtain a comprehensive reward value; updating a price prediction model of the virtual power plant by using the comprehensive reward value; and generating a predicted electricity price sequence of a current price prediction period by using the updated price prediction model in combination with historical electricity price data of the electric power spot market and current market environment information.

Description

Technical Field

[0001] This application relates to the field of power technology, and in particular to a method, apparatus, equipment and medium for predicting prices of a virtual power plant. Background Technology

[0002] With the rapid development of new power systems, Virtual Power Plants (VPPs) have emerged as a distributed energy aggregation and coordinated control system based on advanced information and communication technologies and intelligent optimization algorithms. A VPP aggregates geographically dispersed and diverse distributed energy resources, including rooftop photovoltaics, small-scale wind power, user-side energy storage systems, and interruptible or adjustable loads (such as electric vehicle charging stations, central air conditioning, and flexible industrial production lines), into a single "power plant-like" entity with overall dispatch capabilities through a unified data acquisition, status awareness, and control interface. This entity then participates in the electricity spot market as a single market participant.

[0003] In the electricity spot market (such as the day-ahead market or real-time market), VPPs (Virtual Power Producers) must submit bids for electricity volume and price for future time periods (usually 96 15-minute points or 24-hour points on the following day) before the trading deadline. Market operators then conduct centralized clearing based on the bids from all market participants, combined with grid security constraints, to form time-of-use (TOU) node prices or regional marginal prices. A VPP's final revenue depends on whether its bid is accepted by the market (i.e., whether it wins the bid) and the difference between the settlement price corresponding to the won bid volume and its own generation / consumption costs. Therefore, accurately predicting future spot market prices is a prerequisite for VPPs to formulate effective bidding strategies. If the predicted price is significantly higher than the actual clearing price, the VPP may fail to win the bid due to an overly high price, losing revenue opportunities; if the prediction is too low, it may result in winning the bid at a low price or even operating at a loss.

[0004] In a VPP's intelligent forecasting system, the forecasting module plays a crucial role in predicting electricity spot market prices, and its output directly determines the effectiveness of subsequent dispatching and pricing strategies. To improve forecasting performance, reinforcement learning frameworks are commonly introduced, with the reward mechanism serving as a core element guiding model optimization and directly influencing the training direction and final accuracy of the forecasting model.

[0005] In related technologies, the reward mechanism commonly used in VPP typically adopts the form of "total reward = base reward - penalty term". The base reward reflects the economic gains under a specific strategy, while the penalty term quantifies the economic losses caused by power deviation, violation of scheduling rules, or assumption of market risks. However, this reward mechanism usually uses a fixed loss function (such as mean squared error or mean absolute error) when calculating the base reward and penalty term. When the market microstructure changes rapidly, the reward signal may mismatch with the actual optimization target under different market conditions. The prediction module may struggle to obtain accurate and effective feedback guidance, ultimately resulting in low price prediction accuracy and affecting the reliability of the overall VPP scheduling strategy.

[0006] Therefore, the price prediction method for virtual power plants in related technologies suffers from low prediction accuracy. Summary of the Invention

[0007] This application provides a method, apparatus, equipment, and medium for predicting prices in a virtual power plant, in order to improve the accuracy of price prediction in a virtual power plant.

[0008] In a first aspect, embodiments of this application provide a method for predicting the price of a virtual power plant, the method comprising: When the current price forecast period arrives, obtain the actual electricity price sequence of the previous price forecast period; The basic reward is calculated based on the predicted electricity price sequence of the previous price forecast period, the actual electricity price sequence, and the preset market volatility index. Obtain the number of competitors participating in the same prediction task of the previous price prediction period, determine the competition intensity based on the number of competitors, and dynamically select a reward calculation strategy that matches the competition intensity based on the preset mapping relationship between competition intensity and reward strategy. The base reward is adjusted based on the matching reward calculation strategy to obtain a comprehensive reward value; The price prediction model for the virtual power plant is updated using the comprehensive reward value; Using the updated price forecasting model, combined with historical electricity spot market price data and current market environment information, a forecasted electricity price sequence for the current price forecasting period is generated.

[0009] In some embodiments, the calculation of the basic reward based on the predicted electricity price sequence of the previous price forecast period, the actual electricity price sequence, and a preset market volatility index includes: Calculate the average absolute error based on the predicted electricity price sequence and the actual electricity price sequence; The base reward is determined by multiplying the mean absolute error by the preset market volatility index.

[0010] In some embodiments, determining the competition intensity based on the number of competitors and dynamically selecting a reward calculation strategy that matches the competition intensity according to a preset mapping relationship between competition intensity and reward strategy includes: When the number of competitors does not exceed a preset threshold, the competition intensity is determined to be low intensity, and a first reward calculation strategy matching the low intensity is selected based on the preset mapping relationship between competition intensity and reward strategy; the first reward calculation strategy is used to calculate rewards and penalties based on the relationship between the virtual power plant's own predicted electricity price sequence and the actual electricity price sequence. When the number of competitors exceeds the preset threshold, the competition intensity is determined to be high, and a second reward calculation strategy matching the high intensity is selected based on the preset mapping relationship between competition intensity and reward strategy. The second reward calculation strategy is used to penalize based on the relationship between virtual power plant prediction results and external market information.

[0011] In some embodiments, when the selected reward calculation strategy that matches the competition intensity is a first reward calculation strategy, the base reward is adjusted based on the matched reward calculation strategy to obtain a comprehensive reward value, including: Identify predefined key characteristic events within the previous price forecast period, including some or all of price spike events and trend consistency events; Based on the predicted and actual electricity prices for the corresponding time periods of the key feature events, the feature reward is calculated; Based on the predicted electricity price sequence and the actual electricity price sequence, a prediction error set is constructed, and the error risk value of the prediction error set is calculated based on a preset confidence level. The risk penalty is calculated based on the error risk value, the preset risk threshold, and the risk penalty coefficient. Based on the electricity spot market rules of the previous price forecast period and the violation data during the forecasting process, calculate the constraint penalties; Based on a preset first dynamic weighting coefficient, the basic reward, the feature reward, the risk penalty, and the constraint penalty are weighted and summed to obtain the comprehensive reward value.

[0012] In some embodiments, when the selected reward calculation strategy that matches the competition intensity is a second reward calculation strategy, the base reward is adjusted based on the matched reward calculation strategy to obtain a comprehensive reward value, including: Obtain predicted and actual behavioral information of at least one competitor, and determine the penalty for competitor strategy differences based on the predicted and actual behavioral information. Obtain the market equilibrium strategy information of the electricity spot market in the previous price forecast period, and determine the market equilibrium deviation penalty based on the market equilibrium strategy information and the forecast electricity price sequence; Based on a preset second dynamic weighting coefficient, the basic reward, the opponent strategy difference penalty, and the market equilibrium deviation penalty are weighted and summed to obtain the comprehensive reward value.

[0013] In some embodiments, obtaining predicted and actual behavioral information of at least one competitor, and determining competitor strategy difference penalties based on the predicted and actual behavioral information, includes: Obtain the predicted and actual pricing strategy distributions of at least one competitor; where the strategy distribution is a probability distribution obtained by normalizing historical pricing data. Calculate the first distribution difference between the predicted pricing strategy distribution and the actual pricing strategy distribution for each competitor, and select the target distribution difference from each first distribution difference; The opponent strategy difference penalty is obtained by multiplying the target distribution difference degree by the preset opponent strategy difference coefficient.

[0014] In some embodiments, obtaining market equilibrium strategy information of the electricity spot market for the previous price forecast period, and determining a market equilibrium deviation penalty based on the market equilibrium strategy information and the forecasted electricity price sequence, includes: Obtain the market equilibrium pricing strategy of the electricity spot market for the previous price forecast period, and calculate the second distribution difference between the market equilibrium pricing strategy and the distribution formed after normalization of the predicted electricity price sequence; The market equilibrium deviation penalty is obtained by multiplying the second distribution difference degree by the preset market equilibrium deviation coefficient.

[0015] Secondly, embodiments of this application provide a price prediction device for a virtual power plant, the device comprising: The acquisition module is used to acquire the actual electricity price sequence of the previous price forecast period when the current price forecast period arrives. The calculation module is used to calculate the basic reward based on the predicted electricity price sequence of the previous price forecast period, the actual electricity price sequence, and the preset market volatility index. The selection module is used to obtain the number of competitors participating in the same prediction task of the previous price prediction period, determine the competition intensity based on the number of competitors, and dynamically select a reward calculation strategy that matches the competition intensity based on the preset mapping relationship between competition intensity and reward strategy. An adjustment module is used to adjust the base reward based on the matched reward calculation strategy to obtain a comprehensive reward value; The update module is used to update the price prediction model of the virtual power plant using the comprehensive reward value; The generation module is used to generate the predicted electricity price sequence for the current price prediction period by using the updated price prediction model, combined with historical electricity spot market price data and current market environment information.

[0016] Thirdly, embodiments of this application provide an electronic device, including: at least one processor, and a memory communicatively connected to the at least one processor, wherein: The memory stores a computer program that can be executed by at least one processor, which enables the at least one processor to perform the price prediction method for the virtual power plant described above.

[0017] Fourthly, embodiments of this application provide a storage medium in which, when a computer program in the storage medium is executed by a processor of an electronic device, the electronic device is able to execute the aforementioned price prediction method for a virtual power plant.

[0018] Fifthly, embodiments of this application provide a computer program product that, when executed by an electronic device, enables the electronic device to perform the aforementioned price prediction method for a virtual power plant.

[0019] In this embodiment, upon reaching the current price forecast period, the actual electricity price sequence of the previous price forecast period is obtained. Based on the predicted electricity price sequence, the actual electricity price sequence, and a preset market volatility index from the previous price forecast period, a basic reward is calculated. The number of competitors participating in the same forecast task from the previous price forecast period is obtained, and the competition intensity is determined based on the number of competitors. A reward calculation strategy matching the competition intensity is dynamically selected based on a preset mapping relationship between competition intensity and reward strategy. The basic reward is adjusted based on the matched reward calculation strategy to obtain a comprehensive reward value. The price forecast model of the virtual power plant is updated using the comprehensive reward value. Using the updated price forecast model, combined with historical electricity spot market price data and current market environment information, a predicted electricity price sequence for the current price forecast period is generated. This embodiment judges the competition intensity by monitoring the number of competitors, and then dynamically switches the reward calculation strategy based on the competition intensity. This allows the comprehensive reward value adjusted according to different reward calculation strategies to dynamically adapt to the competitive situation and volatility characteristics of the current market environment, effectively mitigating the reward-target mismatch problem caused by a fixed loss function. The forecast module can obtain accurate and effective feedback guidance, thereby effectively improving forecast accuracy.

[0020] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0021] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This application provides an illustration of an application scenario for a price prediction method for a virtual power plant. Figure 2 A schematic diagram of the architecture of a virtual power plant provided for an embodiment of this application; Figure 3 A flowchart illustrating a price prediction method for a virtual power plant provided in this application embodiment; Figure 4 This application provides a schematic diagram illustrating the selection of a reward calculation strategy. Figure 5 This application provides a schematic diagram of the calculation process for a comprehensive reward; Figure 6 This application provides an embodiment of another schematic diagram of the calculation process for comprehensive rewards; Figure 7 A schematic diagram of the structure of a price prediction device for a virtual power plant provided in an embodiment of this application; Figure 8 This is a schematic diagram of the hardware structure of an electronic device for implementing a price prediction method for a virtual power plant, provided as an embodiment of this application. Detailed Implementation

[0022] To improve the accuracy of price prediction for virtual power plants, embodiments of this application provide a method, apparatus, equipment, and medium for price prediction of virtual power plants.

[0023] The preferred embodiments of this application are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit this application. Furthermore, the embodiments and features in the embodiments of this application can be combined with each other without conflict.

[0024] It should be noted that the terms "first," "second," etc., used in the description of the embodiments of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices.

[0025] The acquisition, transmission, storage, and use of data in this application all comply with relevant national laws and regulations.

[0026] See Figure 1 , Figure 1 This is an application scenario diagram of a price prediction method for a virtual power plant provided in an embodiment of this application. The application scenario includes a terminal device 110 and a server 120.

[0027] In this embodiment, the terminal device 110 includes, but is not limited to, devices such as mobile phones, tablets, laptops, and desktop computers. The terminal device may have a client installed related to price prediction for the virtual power plant. This client can be software (such as a browser, price prediction software, etc.), or a webpage, mini-program, etc. The server 120 is the backend server corresponding to the software, webpage, mini-program, etc., or a server specifically used for price prediction of the virtual power plant; this application does not impose specific limitations. The server 120 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0028] It should be noted that the price prediction method of the virtual power plant in the various embodiments of this application can be executed by an electronic device, which can be a terminal device 110 or a server 120. That is, the method can be executed by the terminal device 110 or the server 120 alone, or by the terminal device 110 and the server 120 together.

[0029] In one alternative implementation, the terminal device 110 and the server 120 can communicate via a communication network.

[0030] In one alternative implementation, the communication network is a wired network or a wireless network.

[0031] It should be noted that, Figure 1 The examples shown are merely illustrative; in reality, the number of terminal devices and servers is unlimited and is not specifically limited in the embodiments of this application.

[0032] The following description, in conjunction with the accompanying drawings, illustrates some preferred embodiments of this application.

[0033] See Figure 2 , Figure 2 This application provides an embodiment of a virtual power plant architecture diagram, in which the virtual power plant mainly includes a prediction module, a resource aggregation and modeling module, an optimization decision-making module, a pricing strategy generation module, an implementation scheduling and control module, and a settlement and risk management module. The prediction module undertakes the key task of predicting the price of electricity spot market, and its output directly determines the effectiveness of subsequent scheduling and pricing strategies. The prediction module can realize renewable energy output prediction, adjustable load prediction, electricity spot market price prediction, and competitor pricing strategy prediction.

[0034] Specifically, in the electricity spot market (such as the day-ahead market or the real-time market), Virtual Power Plants (VPPs) must submit bids for electricity volume and price for future time periods (usually 96 15-minute points or 24-hour points on the following day) before the trading deadline. Market operators then conduct centralized clearing based on the bids from all market participants, combined with grid security constraints, to form time-of-use pricing or regional marginal pricing. The VPP's final revenue depends on whether its bid is accepted by the market (i.e., whether it wins the bid) and the difference between the settlement price corresponding to the won bid volume and its own generation / consumption costs. Therefore, accurately predicting electricity spot market prices and competitors' bidding strategies is crucial to maximizing VPP revenue. To improve prediction accuracy, related technologies often use reward mechanisms to guide model optimization.

[0035] However, in related technologies, the reward mechanism commonly used in VPP typically adopts the form of "total reward = basic reward - penalty term". The basic reward reflects the economic gains under a specific strategy, while the penalty term quantifies the economic losses caused by power deviation, violation of scheduling rules, or assumption of market risks. However, this reward mechanism usually uses a fixed loss function (such as mean squared error or mean absolute error) when calculating the basic reward and penalty term. When the market microstructure changes rapidly, the reward signal may mismatch with the actual optimization target under different market conditions. The prediction module may struggle to obtain accurate and effective feedback guidance, ultimately resulting in low price prediction accuracy and affecting the reliability of the overall VPP scheduling strategy.

[0036] Therefore, embodiments of this application provide a method, apparatus, equipment, and medium for predicting prices for virtual power plants, in order to improve the accuracy of price prediction for virtual power plants.

[0037] Before describing in detail the price prediction method for a virtual power plant provided in the embodiments of this application, its overall concept will be briefly explained first.

[0038] The core idea of ​​this application is to achieve on-demand online optimization of the price prediction model by introducing a dynamic reward mechanism. Specifically, upon reaching the current price prediction period, the actual electricity price sequence of the previous prediction period is first obtained and compared with the predicted electricity price sequence output by the model in the previous period. Combining a preset market volatility index and the competition intensity determined based on the number of competitors, a matching reward calculation strategy is dynamically selected to calculate a comprehensive reward value that reflects prediction accuracy, economic potential, and market adaptability. This comprehensive reward value serves as a feedback signal in reinforcement learning, used to update the parameters of the price prediction model in real time. Subsequently, using the updated model, combined with historical electricity spot market price data and current market environment information, the predicted electricity price sequence for the current prediction period is generated. Through this closed-loop mechanism of "previous period prediction → obtaining actual electricity price after the period ends → calculating reward and updating the model at the beginning of the current period → predicting the current period," the prediction model can respond promptly to the dynamic changes in the microstructure of the electricity spot market, effectively improving prediction accuracy and the reliability of virtual power plant dispatch decisions.

[0039] See Figure 3 , Figure 3 A flowchart of a price prediction method for a virtual power plant provided in an embodiment of this application is shown. The method includes the following steps.

[0040] In step 301, when the current price forecast period is reached, the actual electricity price sequence of the previous price forecast period is obtained.

[0041] The price forecast period can be every 2 hours, every 6 hours, or every 24 hours; this application does not limit this. When the price forecast period is reached, the virtual power plant can perform price forecasts according to the preset electricity price forecast time range set for participating in the electricity spot market. The time length and time granularity are determined by the trading rules of the target electricity spot market. For example, taking a price forecast period of 4 hours, a time length of 12 hours, and a time granularity of 1 hour as an example, the virtual power plant forecasts the electricity price for each hour of the next 12 hours every 4 hours.

[0042] In practice, when the current price forecast period arrives, the actual electricity price sequence of the previous price forecast period is obtained. In fact, after each price forecast ends, there is a set of predicted electricity price sequences and a set of actual electricity price sequences.

[0043] In step 302, the basic reward is calculated based on the predicted electricity price sequence, the actual electricity price sequence, and the preset market volatility index of the previous price forecast period.

[0044] Specifically, the base reward is used to quantify the economic relevance and market adaptability of the price forecast results of the virtual power plant (VPP) in the previous forecast period.

[0045] Optionally, the mean absolute error (MAE) is first calculated based on the predicted and actual electricity price sequences to measure the average level of prediction deviation. Then, the mean absolute error is multiplied by a preset market volatility index, and the result is either negative or normalized to determine the base reward.

[0046] For example, the basic reward can be represented as ;in, This indicates a market volatility index.

[0047] The VPP architecture may also include a market state awareness module, and the market volatility index can be obtained by monitoring and calculating the market state awareness module of the VPP.

[0048] In step 303, the number of competitors participating in the same prediction task of the previous price prediction period is obtained, the competition intensity is determined based on the number of competitors, and a reward calculation strategy that matches the competition intensity is dynamically selected based on the preset mapping relationship between competition intensity and reward strategy.

[0049] In practice, VPPs can obtain the number of competitors participating in the same forecasting task in the previous price forecasting cycle by calling the market coordination module or external interfaces. Here, the forecasting task can be the same electricity spot market bidding task, and competitors can be other VPPs, independent power generators, or load aggregators. The number of competitors reflects the intensity of competition in the current market environment.

[0050] Optionally, when the number of competitors does not exceed a preset threshold, such as 2, the competition intensity is determined to be low, and a first reward calculation strategy matching the low intensity is selected based on the preset mapping relationship between competition intensity and reward strategy; when the number of competitors exceeds the preset threshold, the competition intensity is determined to be high, and a second reward calculation strategy matching the high intensity is selected based on the preset mapping relationship between competition intensity and reward strategy.

[0051] The first reward calculation strategy is used to calculate rewards and penalties based on the relationship between the virtual power plant's own predicted electricity price series and the actual electricity price series. The second reward calculation strategy is used to penalize based on the relationship between the virtual power plant's prediction results and external market information. That is, it not only considers its own prediction error, but also further combines external market information (such as the average bid of competitors, the deviation of the market clearing price, or the level of group prediction consensus) to evaluate the prediction results. If the virtual power plant's prediction deviates significantly from the mainstream market expectation or leads to bidding failure, additional penalties will be imposed.

[0052] See Figure 4 This application provides a schematic diagram of reward calculation strategy selection. First, the market volatility index is obtained based on the market state perception module. It obtains the predicted electricity price sequence and actual electricity price sequence from the previous price forecast period based on the price forecast module, and then... Calculate the base reward and then determine the number of competitors in the market. If the number of competitors is no more than 2, select the first reward calculation strategy; if the number of competitors is more than 2, select the second reward calculation strategy.

[0053] In this way, the reward calculation strategy adapts to changing environments: when competition is mild, it encourages the prediction model to explore high-yield but high-risk prediction paths; when competition is intense, it guides the model to converge in a robust and collaborative direction, thus obtaining high-quality and robust price prediction results in different market ecosystems.

[0054] In step 304, the base reward is adjusted based on the matching reward calculation strategy to obtain the comprehensive reward value.

[0055] The following sections will detail two scenarios where the base reward is adjusted using different reward calculation strategies to obtain the overall reward value.

[0056] In the first scenario, when the selected reward calculation strategy, which matches the intensity of competition, is the primary reward calculation strategy, this application refines the basic reward by integrating multi-dimensional evaluation indicators. The specific implementation process includes the following steps: Step one can identify predefined key characteristic events within the previous price prediction period, including but not limited to price spikes and trend consistency events.

[0057] Among them, price spike events refer to electricity prices being significantly higher than normal levels at a certain moment, such as a sudden price increase above a threshold; trend consistency events refer to electricity prices showing a monotonous upward or downward trend over multiple consecutive periods.

[0058] For example, to identify price spikes, a standardized score can be calculated based on the mean and standard deviation of the actual electricity price series in the previous price forecast period. If the standardized score is greater than a preset score, it is considered that a price spike has been captured, and the time at this moment is called the spike moment. .

[0059] Step 2: Calculate the feature reward based on the predicted and actual electricity prices for the corresponding time periods of key feature events.

[0060] Specifically, The purpose of calculating feature rewards is to evaluate them from multiple scales.

[0061] The following example uses price spike events. The reward for price spike events (hereinafter referred to as spike capture) is determined by the mean squared error (MSE) between the predicted and actual values ​​at the spike moment and the spike capture reward coefficient PR. The corresponding predicted electricity price and actual electricity price ,calculate and The mean square error is then multiplied by the peak capture reward coefficient PR (empirical setpoint) to obtain the peak capture reward value, i.e. Since peak capture is one of the key features of the electricity market, peak capture rewards are considered as a feature reward. one.

[0062] Step 3: Based on the predicted electricity price series and the actual electricity price series, construct a prediction error set, calculate the error risk value of the prediction error set based on the preset confidence level, and calculate the risk penalty based on the error risk value, the preset risk threshold, and the risk penalty coefficient.

[0063] Specifically, the difference between each predicted electricity price in the predicted electricity price series and the corresponding actual electricity price in the actual electricity price series can be calculated sequentially to construct a prediction error set, i.e. Set up the prediction error set; then, based on a preset confidence level c (e.g., 95%), calculate each prediction error in the prediction error set. Given the probability of not exceeding each error threshold (s), calculate the minimum error threshold (s) that satisfies the condition, and determine the error threshold (s) as the error risk value VaR, i.e., VaR( ) = inf{s∈R:P( ≤s)≥c}, and then, based on the preset risk threshold Given the error risk value VaR, when VaR is determined to be greater than a preset risk threshold, the difference between VaR and the preset risk threshold is multiplied by the risk penalty coefficient RR to obtain the risk penalty value. When the error risk value VaR is determined to be no greater than the preset risk threshold, the risk penalty value can be set to a preset value, such as 0. That is, the risk penalty value can be expressed as: .

[0064] Step 4: Calculate the constraint penalties based on the electricity spot market rules of the previous price forecast period and the violation data during the forecasting process.

[0065] Among them, constraint and punishment This refers to the penalty imposed when a VPP's forecasts fail to meet market rule constraints, such as violating maximum / minimum output limits or exceeding declared limits. Specifically, when calculating the constraint penalty, it is necessary to check whether the VPP's forecasting process violates electricity market rules. If so, a penalty amount is given for each violation, based on the specific circumstances. The total penalty amount is the constraint penalty. .

[0066] Step 5: Based on the preset first dynamic weight coefficient, the basic reward, feature reward, risk penalty and constraint penalty are weighted and summed to obtain the comprehensive reward value.

[0067] In practice, after determining the basic reward, feature reward, risk penalty, and constraint penalty, a weighted sum can be calculated based on the weight coefficients corresponding to the basic reward, feature reward, risk penalty, and constraint penalty. For example, the basic reward value mentioned above can be used as the basis for the calculation. Feature reward value Risk penalty value Constraint penalty value The comprehensive reward value is obtained by multiplying each value by the corresponding coefficients α, β, γ, and δ. Where α, β, γ, and δ are dynamically adjustable in real time, satisfying the condition α + β + γ + δ = 1, that is: .

[0068] In this way, when the competition intensity is low, rewards and penalties can be calculated from multiple dimensions such as key feature events, risk penalties, and constraint penalties by using the relationship between the predicted electricity price sequence and the actual electricity price sequence based on the virtual power plant itself. This is beneficial to improving the accuracy and stability of the prediction system.

[0069] See Figure 5 , Figure 5 This application provides a schematic diagram of the calculation process for a comprehensive reward. First, based on the price prediction module, the predicted electricity price sequence and the actual electricity price sequence of the previous price prediction period are obtained to determine the basic reward. Then, market rule constraints are obtained. Based on the predicted electricity price sequence and the actual electricity price sequence, price peak events are captured and the error risk value of the prediction error is calculated. Constraint penalties are calculated based on the market rule constraints. Then, peak capture rewards are calculated based on the captured price peak events. In addition, the rewards corresponding to each key feature are combined to calculate the feature reward. The risk penalty value is calculated based on the error risk value. Finally, after determining the basic reward, feature reward, risk penalty, and constraint penalty, the comprehensive reward value can be obtained by weighted summation according to the weight coefficients corresponding to the basic reward, feature reward, risk penalty, and constraint penalty.

[0070] For example, the VPP's price prediction module provides electricity price predictions for time 0, time 1, and time 2, respectively. [0]=4、 [1]=4、 [2]=5, corresponding to the actual electricity price [0]=7、 [1]=6、 [2]=6, calculate the average absolute error of these three sets of values, i.e. The market volatility index is then provided by the market status perception module. ; final calculation .

[0071] Suppose we obtain the actual electricity price sequences at times 0, 1, and 2: Price[0]={6.9,7,7.1}, Price[1]={5.8,6,5.9}, and Price[2]={6.1,6,5.9}. We calculate the ZScore values ​​(i.e., Z-scores) of the three as 0, 0.387, and 0, respectively. We set a threshold of 0.2. If the ZScore value is greater than the threshold, it is considered that a price peak event has been captured. The time corresponding to 0.387 (>0.2) is called the peak time, i.e., tpeak=1 time.

[0072] Find the corresponding price prediction value at the peak moment (time 1) mentioned above. [1]=4 and actual value [1]=6, and the mean square error of both can be obtained. Set the peak capture reward coefficient PR=-500, and calculate the result. Assuming that key market characteristics only consider peak capture, i.e. .

[0073] The three groups in the above steps are respectively and The corresponding prediction error is obtained by subtraction. [0]=3、 [1]=2、 [2]=1, the probability that the three sets of error values ​​do not exceed a certain threshold s must satisfy the confidence level c=66.67%. The lower bound of this threshold is the VaR (error risk value) of the error, that is...

[0074] Set risk threshold If the VaR value is determined to be greater than 1.5, the difference between the two is calculated to be 0.5; the risk penalty coefficient RR is set to 800, and finally the calculation is performed.

[0075] Suppose that the VPP violates an electricity market rule during its forecasting process, and a penalty of 300 is imposed. .

[0076] Finally, setting α=0.5, β=0.2, γ=0.2, and δ=0.1, the calculation is performed. Ultimately, this value becomes the standard for rewarding or punishing; a positive value represents a reward, a negative value represents a punishment, and a larger value indicates a greater severity.

[0077] The second scenario: When the selected reward calculation strategy, which matches the intensity of competition, is the second reward calculation strategy, this application further refines the basic reward by incorporating external market information. The specific implementation process includes the following steps: Step one involves obtaining predicted and actual behavioral information of at least one competitor, and determining the penalty for the competitor's strategy difference based on the predicted and actual behavioral information.

[0078] In specific implementation, the VPP architecture may also include a competitor strategy awareness module, which can obtain the predicted and actual pricing strategy distributions of at least one competitor. The strategy distribution is a probability distribution obtained after normalizing historical pricing data. The first distribution difference degree between the predicted and actual pricing strategy distributions for each competitor is calculated, and a target distribution difference degree is selected from each first distribution difference degree. The target distribution difference degree is multiplied by a preset competitor strategy difference coefficient to obtain the competitor strategy difference penalty.

[0079] The first distribution difference can be calculated using KL divergence (Kullback-Leibler Divergence), or JS divergence or Wasserstein distance, etc. This application does not limit the specific method used.

[0080] Taking the calculation of the first distribution difference using KL divergence as an example, the KL divergence between the predicted and actual pricing strategy distributions for each competitor can be calculated. The largest KL divergence value is then selected from the KL divergences. This largest KL divergence value is multiplied by the preset competitor strategy difference coefficient OD to obtain the competitor strategy difference penalty. Specifically, the competitor strategy difference penalty can be expressed as: ; in, .

[0081] Step 2: Obtain the market equilibrium strategy information of the electricity spot market in the previous price forecast period, and determine the market equilibrium deviation penalty based on the market equilibrium strategy information and the forecast electricity price series.

[0082] In practice, the VPP's pricing strategy generation module can be used to obtain the market equilibrium pricing strategy of the electricity spot market in the previous price forecast period (for example, by solving the Nash equilibrium constraint condition of multiple competitors' strategies), and calculate the second distribution difference between the market equilibrium pricing strategy and the distribution formed after normalization of the predicted electricity price sequence; multiply the second distribution difference by the preset market equilibrium deviation coefficient to obtain the market equilibrium deviation penalty.

[0083] The second distribution difference can be calculated using KL divergence (Kullback-Leibler Divergence), or JS divergence or Wasserstein distance, etc. This application does not limit the specific method used.

[0084] Taking the calculation of the second distribution divergence using KL divergence as an example, the KL divergence between the market equilibrium pricing strategy and the distribution formed after the normalization of the predicted electricity price series can be calculated. Multiplying the KL divergence value by the preset market equilibrium deviation coefficient MI yields the market equilibrium deviation penalty, which can be specifically expressed as: ; in, .

[0085] Step 3: Based on the preset second dynamic weighting coefficient, the basic reward, the opponent strategy difference penalty, and the market equilibrium deviation penalty are weighted and summed to obtain the comprehensive reward value.

[0086] In practice, after determining the base reward, the opponent strategy difference penalty, and the market equilibrium deviation penalty, a weighted sum can be calculated based on the weight coefficients corresponding to the base reward, the opponent strategy difference penalty, and the market equilibrium deviation penalty. For example, the base reward value mentioned above... Penalty for differences in opponent's strategy Market equilibrium deviation penalty The comprehensive reward value is obtained by multiplying each value by the corresponding coefficients α, β, and γ. Where α, β, and γ are dynamically adjustable in real time, satisfying the condition α + β + γ = 1, that is: .

[0087] In this way, when the competition intensity is high, penalties can be imposed based on the relationship between the virtual power plant forecast results and external market information. That is, not only should the forecast error be considered, but the forecast results should also be evaluated by combining external market information (such as the distribution of competitors' pricing strategies and market equilibrium pricing strategies). This has the ability to be applied to diversified and highly competitive markets, which is conducive to improving the accuracy and stability of the forecast system.

[0088] See Figure 6 , Figure 6 This application provides another schematic diagram of the calculation process for the comprehensive reward. First, after determining the basic reward based on the predicted electricity price sequence and actual electricity price sequence of the previous price prediction period obtained by the price prediction module, the predicted bidding strategy distribution and actual bidding strategy distribution of at least one competitor are obtained based on the competitor strategy perception module. The KL divergence between the predicted bidding strategy distribution and the actual bidding strategy distribution corresponding to each competitor is calculated. Then, the largest KL divergence value is selected from each KL divergence value, and the largest KL divergence value is multiplied by the preset competitor strategy difference coefficient OD to obtain the competitor strategy difference penalty. Next, the market equilibrium bidding strategy of the electricity spot market of the previous price prediction period is obtained based on the bidding strategy generation module. The KL divergence between the market equilibrium bidding strategy and the distribution formed after the normalization of the predicted electricity price sequence is calculated. Then, the largest KL divergence value is multiplied by the preset market equilibrium deviation coefficient to obtain the market equilibrium deviation penalty. Finally, after determining the basic reward, competitor strategy difference penalty and market equilibrium deviation penalty, the comprehensive reward value can be obtained by weighted summation according to the weight coefficients corresponding to the basic reward, competitor strategy difference penalty and market equilibrium deviation penalty.

[0089] For example, still using For example, assuming there is a competitor A in the electricity market, the competitor strategy perception module of the VPP provides predictions of A's bidding behavior at time 0, time 1, and time 2. (0) = 5 (1) = 5 (2) = 5, while A's actual bidding behavior is (0) = 5.5 (1) = 4.5 (2)=4, calculate the KL divergence between the predicted and actual values ​​of A. .

[0090] When there are n competitors in the market, calculate the KL divergence values ​​for each of the n competitors; take the maximum value among these n KL divergence values, assuming the maximum value is still the KL divergence value of A, 0.92; set the coefficient OD = 1000, and finally calculate the competitor strategy difference penalty as follows: .

[0091] The VPP's pricing strategy generation module provides market equilibrium pricing strategies at time 0, time 1, and time 2, respectively. (0) = 4.1 (1) = 4.2 (2) = 4.8, while still using the electricity price forecast values ​​at time 0, time 1, and time 2. (0) = 4 (1) = 4 (2) = 5, calculate the KL divergence value between the two. .

[0092] With a set coefficient MI = 2000, the market equilibrium deviation penalty is calculated as follows: .

[0093] Finally, setting α=0.5, β=0.3, and γ=0.2, the calculations were performed. Ultimately, this value becomes the standard for rewarding or punishing; a positive value represents a reward, a negative value represents a punishment, and a larger value indicates a greater severity.

[0094] In step 305, the price prediction model for the virtual power plant is updated using the comprehensive reward value.

[0095] Specifically, the price prediction model can employ a reinforcement learning-based architecture, where model parameters are optimized by maximizing long-term cumulative rewards. At the beginning of each price prediction cycle, the comprehensive reward value is used as a feedback signal for the corresponding decision-making behavior in that cycle and input into the model's training module. Subsequently, gradient ascent or policy update algorithms are used to adjust the model's internal parameters (e.g., neural network weights) based on this reward value, making the model more inclined to generate electricity price predictions that yield high rewards in future predictions. This process realizes an online adaptive learning mechanism that combines data-driven and goal-oriented approaches.

[0096] In step 306, the updated price forecasting model is used to generate a forecasted electricity price sequence for the current price forecasting period by combining historical electricity spot market price data and current market environment information.

[0097] Specifically, historical electricity spot market price data includes, but is not limited to: day-ahead market clearing prices, real-time market prices, load curves, renewable energy output, and other time-series data from the past several days / weeks; the current market environment information includes: current supply-demand ratio, fuel prices, weather forecasts, holiday markers, system reserve capacity, and contextual features such as the market volatility index output by the market status perception module. After preprocessing (e.g., normalization, sliding window slicing, feature embedding), the above multi-source heterogeneous data is fed as input into the updated price prediction model. The model outputs a price prediction sequence for a future time window (e.g., the next 12 hours, with a granularity of 1 hour), i.e., the prediction result for the current price prediction period.

[0098] This application embodiment introduces a dynamic reward mechanism to achieve on-demand online optimization of the price prediction model. Specifically, upon reaching the current price prediction cycle, the actual electricity price sequence of the previous prediction cycle is first obtained and compared with the predicted electricity price sequence output by the model in the previous cycle. Combining a preset market volatility index and the competition intensity determined based on the number of competitors, a matching reward calculation strategy is dynamically selected to calculate a comprehensive reward value that reflects prediction accuracy, economic potential, and market adaptability. This comprehensive reward value serves as a feedback signal in reinforcement learning, used to update the parameters of the price prediction model in real time. Subsequently, using the updated model, combined with historical electricity spot market price data and current market environment information, a predicted electricity price sequence for the current prediction cycle is generated. Through this closed-loop mechanism of "previous cycle prediction → obtaining actual electricity price after cycle ends → calculating reward and updating model at the beginning of the current cycle → predicting the current cycle," the prediction model can respond promptly to dynamic changes in the microstructure of the electricity spot market, effectively improving prediction accuracy and the reliability of virtual power plant dispatch decisions.

[0099] Based on the same technical concept, this application also provides a price prediction device for a virtual power plant. The principle of the price prediction device for a virtual power plant is similar to that of the price prediction method for a virtual power plant described above. Therefore, the implementation of the price prediction device for a virtual power plant can refer to the implementation of the price prediction method for a virtual power plant, and the repeated parts will not be described again.

[0100] Figure 7 A schematic diagram of the structure of a virtual power plant price prediction device 700 provided in this application embodiment includes: The acquisition module 701 is used to acquire the actual electricity price sequence of the previous price forecast period when the current price forecast period arrives. Calculation module 702 is used to calculate the basic reward based on the predicted electricity price sequence of the previous price prediction period, the actual electricity price sequence, and the preset market volatility index; The selection module 703 is used to obtain the number of competitors participating in the same prediction task of the previous price prediction period, determine the competition intensity based on the number of competitors, and dynamically select a reward calculation strategy that matches the competition intensity based on the preset mapping relationship between competition intensity and reward strategy. The adjustment module 704 is used to adjust the basic reward based on the matching reward calculation strategy to obtain a comprehensive reward value; Update module 705 is used to update the price prediction model of the virtual power plant using the comprehensive reward value; The generation module 706 is used to generate the predicted electricity price sequence for the current price prediction period by using the updated price prediction model, combined with historical electricity spot market price data and current market environment information.

[0101] In some embodiments, the computing module 702 is specifically used for: Calculate the average absolute error based on the predicted electricity price sequence and the actual electricity price sequence; The base reward is determined by multiplying the mean absolute error by the preset market volatility index.

[0102] In some embodiments, the selection module 703 is specifically used for: When the number of competitors does not exceed a preset threshold, the competition intensity is determined to be low intensity, and a first reward calculation strategy matching the low intensity is selected based on the preset mapping relationship between competition intensity and reward strategy; the first reward calculation strategy is used to calculate rewards and penalties based on the relationship between the virtual power plant's own predicted electricity price sequence and the actual electricity price sequence. When the number of competitors exceeds the preset threshold, the competition intensity is determined to be high, and a second reward calculation strategy matching the high intensity is selected based on the preset mapping relationship between competition intensity and reward strategy. The second reward calculation strategy is used to penalize based on the relationship between virtual power plant prediction results and external market information.

[0103] In some embodiments, when the selected reward calculation strategy that matches the competition intensity is the first reward calculation strategy, the adjustment module 704 is specifically used for: Identify predefined key characteristic events within the previous price forecast period, including some or all of price spike events and trend consistency events; Based on the predicted and actual electricity prices for the corresponding time periods of the key feature events, the feature reward is calculated; Based on the predicted electricity price sequence and the actual electricity price sequence, a prediction error set is constructed, and the error risk value of the prediction error set is calculated based on a preset confidence level. The risk penalty is calculated based on the error risk value, the preset risk threshold, and the risk penalty coefficient. Based on the electricity spot market rules of the previous price forecast period and the violation data during the forecasting process, calculate the constraint penalties; Based on a preset first dynamic weighting coefficient, the basic reward, the feature reward, the risk penalty, and the constraint penalty are weighted and summed to obtain the comprehensive reward value.

[0104] In some embodiments, when the selected reward calculation strategy that matches the competition intensity is the second reward calculation strategy, the adjustment module 704 is specifically used for: Obtain predicted and actual behavioral information of at least one competitor, and determine the penalty for competitor strategy differences based on the predicted and actual behavioral information. Obtain the market equilibrium strategy information of the electricity spot market in the previous price forecast period, and determine the market equilibrium deviation penalty based on the market equilibrium strategy information and the forecast electricity price sequence; Based on a preset second dynamic weighting coefficient, the basic reward, the opponent strategy difference penalty, and the market equilibrium deviation penalty are weighted and summed to obtain the comprehensive reward value.

[0105] In some embodiments, the adjustment module 704 is specifically used for: Obtain the predicted and actual pricing strategy distributions of at least one competitor; where the strategy distribution is a probability distribution obtained by normalizing historical pricing data. Calculate the first distribution difference between the predicted pricing strategy distribution and the actual pricing strategy distribution for each competitor, and select the target distribution difference from each first distribution difference; The opponent strategy difference penalty is obtained by multiplying the target distribution difference degree by the preset opponent strategy difference coefficient.

[0106] In some embodiments, the adjustment module 704 is specifically used for: Obtain the market equilibrium pricing strategy of the electricity spot market for the previous price forecast period, and calculate the second distribution difference between the market equilibrium pricing strategy and the distribution formed after normalization of the predicted electricity price sequence; The market equilibrium deviation penalty is obtained by multiplying the second distribution difference degree by the preset market equilibrium deviation coefficient.

[0107] The module division in this application embodiment is illustrative and only represents one logical functional division. In actual implementation, other division methods are possible. Furthermore, the functional modules in each embodiment of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. Coupling between modules can be achieved through interfaces, typically electrical communication interfaces, but mechanical interfaces or other types of interfaces are also possible. Therefore, modules described as separate components may or may not be physically separate; they can be located in one place or distributed across different locations on the same or different devices. The integrated modules described above can be implemented in hardware or as software functional modules.

[0108] Having introduced the price prediction method and apparatus for virtual power plants according to exemplary embodiments of this application, we will now introduce an electronic device according to another exemplary embodiment of this application.

[0109] The following reference Figure 8 To describe an electronic device 130 implemented according to this embodiment of the present application. Figure 8 The electronic device 130 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0110] like Figure 8 As shown, the electronic device 130 is presented in the form of a general electronic device. The components of the electronic device 130 may include, but are not limited to: at least one processor 131, at least one memory 132, and a bus 133 connecting different system components (including memory 132 and processor 131).

[0111] Bus 133 represents one or more of several bus structures, including a memory bus or memory controller, peripheral bus, processor, or local bus using any of the various bus structures.

[0112] The memory 132 may include a readable medium in the form of volatile memory, such as random access memory (RAM) 1321 and / or cache memory 1322, and may further include read-only memory (ROM) 1323.

[0113] The memory 132 may also include a program / utility 1325 having a set (at least one) of program modules 1324, including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0114] Electronic device 130 can also communicate with one or more external devices 134 (e.g., keyboard, pointing device, etc.), and with one or more devices that enable a user to interact with electronic device 130, and / or with any device that enables electronic device 130 to communicate with one or more other electronic devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 135. Furthermore, electronic device 130 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 136. As shown, network adapter 136 communicates with other modules used in electronic device 130 via bus 133. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 130, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0115] In an exemplary embodiment, a storage medium is also provided, which, when executed by a processor of an electronic device, enables the electronic device to perform the aforementioned price prediction method for a virtual power plant. Optionally, the storage medium may be a non-transitory computer-readable storage medium, such as a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.

[0116] In an exemplary embodiment, the electronic device of this application may include at least one processor and a memory communicatively connected to the at least one processor, wherein the memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, it enables the at least one processor to perform the steps of any virtual power plant price prediction method provided in the embodiments of this application.

[0117] In an exemplary embodiment, a computer program product is also provided, which, when executed by an electronic device, enables the electronic device to implement any of the exemplary methods provided in this application.

[0118] Furthermore, computer program products may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, RAM, ROM, erasable programmable read-only memory (EPROM), flash memory, optical fiber, compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0119] The program product for price prediction of virtual power plants in the embodiments of this application can be a CD-ROM and include program code, and can run on a computing device. However, the program product of this application is not limited to this. In this document, the readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0120] A readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying readable program code. This propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0121] The program code contained on the readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, radio frequency (RF), or any suitable combination thereof.

[0122] Program code for performing the operations of this application can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, such as a Local Area Network (LAN) or a Wide Area Network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0123] It should be noted that although several units or sub-units of the device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this application, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units.

[0124] Furthermore, although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0125] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0126] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0127] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0128] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0129] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0130] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A price prediction method for a virtual power plant, characterized in that, The method includes: When the current price forecast period arrives, obtain the actual electricity price sequence of the previous price forecast period; The basic reward is calculated based on the predicted electricity price sequence of the previous price forecast period, the actual electricity price sequence, and the preset market volatility index. Obtain the number of competitors participating in the same prediction task of the previous price prediction period, determine the competition intensity based on the number of competitors, and dynamically select a reward calculation strategy that matches the competition intensity based on the preset mapping relationship between competition intensity and reward strategy. The base reward is adjusted based on the matching reward calculation strategy to obtain a comprehensive reward value; The price prediction model for the virtual power plant is updated using the comprehensive reward value; Using the updated price forecasting model, combined with historical electricity spot market price data and current market environment information, a forecasted electricity price sequence for the current price forecasting period is generated.

2. The method as described in claim 1, characterized in that, The basic reward is calculated based on the predicted electricity price sequence from the previous price forecast period, the actual electricity price sequence, and the preset market volatility index, including: Calculate the average absolute error based on the predicted electricity price sequence and the actual electricity price sequence; The base reward is determined by multiplying the mean absolute error by the preset market volatility index.

3. The method as described in claim 1, characterized in that, The step of determining the competition intensity based on the number of competitors, and dynamically selecting a reward calculation strategy that matches the competition intensity according to a preset mapping relationship between competition intensity and reward strategy, includes: When the number of competitors does not exceed a preset threshold, the competition intensity is determined to be low intensity, and a first reward calculation strategy matching the low intensity is selected based on the preset mapping relationship between competition intensity and reward strategy; the first reward calculation strategy is used to calculate rewards and penalties based on the relationship between the virtual power plant's own predicted electricity price sequence and the actual electricity price sequence. When the number of competitors exceeds the preset threshold, the competition intensity is determined to be high, and a second reward calculation strategy matching the high intensity is selected based on the preset mapping relationship between competition intensity and reward strategy. The second reward calculation strategy is used to penalize based on the relationship between virtual power plant prediction results and external market information.

4. The method as described in claim 3, characterized in that, When the selected reward calculation strategy that matches the competition intensity is the first reward calculation strategy, the base reward is adjusted based on the matched reward calculation strategy to obtain a comprehensive reward value, including: Identify predefined key characteristic events within the previous price forecast period, including some or all of price spike events and trend consistency events; Based on the predicted and actual electricity prices for the corresponding time periods of the key feature events, the feature reward is calculated; Based on the predicted electricity price sequence and the actual electricity price sequence, a prediction error set is constructed, and the error risk value of the prediction error set is calculated based on a preset confidence level. The risk penalty is calculated based on the error risk value, the preset risk threshold, and the risk penalty coefficient. Based on the electricity spot market rules of the previous price forecast period and the violation data during the forecasting process, calculate the constraint penalties; Based on a preset first dynamic weighting coefficient, the basic reward, the feature reward, the risk penalty, and the constraint penalty are weighted and summed to obtain the comprehensive reward value.

5. The method as described in claim 3, characterized in that, When the selected reward calculation strategy that matches the competition intensity is the second reward calculation strategy, the base reward is adjusted based on the matched reward calculation strategy to obtain a comprehensive reward value, including: Obtain predicted and actual behavioral information of at least one competitor, and determine the penalty for competitor strategy differences based on the predicted and actual behavioral information. Obtain the market equilibrium strategy information of the electricity spot market in the previous price forecast period, and determine the market equilibrium deviation penalty based on the market equilibrium strategy information and the forecast electricity price sequence; Based on a preset second dynamic weighting coefficient, the basic reward, the opponent strategy difference penalty, and the market equilibrium deviation penalty are weighted and summed to obtain the comprehensive reward value.

6. The method as described in claim 5, characterized in that, The step of obtaining predicted and actual behavioral information of at least one competitor, and determining the competitor's strategy difference penalty based on the predicted and actual behavioral information, includes: Obtain the predicted and actual pricing strategy distributions of at least one competitor; where the strategy distribution is a probability distribution obtained by normalizing historical pricing data. Calculate the first distribution difference between the predicted pricing strategy distribution and the actual pricing strategy distribution for each competitor, and select the target distribution difference from each first distribution difference; The opponent strategy difference penalty is obtained by multiplying the target distribution difference degree by the preset opponent strategy difference coefficient.

7. The method as described in claim 5, characterized in that, The step of obtaining market equilibrium strategy information for the electricity spot market in the previous price forecast period, and determining market equilibrium deviation penalties based on the market equilibrium strategy information and the forecasted electricity price series, includes: Obtain the market equilibrium pricing strategy of the electricity spot market for the previous price forecast period, and calculate the second distribution difference between the market equilibrium pricing strategy and the distribution formed after normalization of the predicted electricity price sequence; The market equilibrium deviation penalty is obtained by multiplying the second distribution difference degree by the preset market equilibrium deviation coefficient.

8. A price prediction device for a virtual power plant, characterized in that, The device includes: The acquisition module is used to acquire the actual electricity price sequence of the previous price forecast period when the current price forecast period arrives. The calculation module is used to calculate the basic reward based on the predicted electricity price sequence of the previous price forecast period, the actual electricity price sequence, and the preset market volatility index. The selection module is used to obtain the number of competitors participating in the same prediction task of the previous price prediction period, determine the competition intensity based on the number of competitors, and dynamically select a reward calculation strategy that matches the competition intensity based on the preset mapping relationship between competition intensity and reward strategy. An adjustment module is used to adjust the base reward based on the matched reward calculation strategy to obtain a comprehensive reward value; The update module is used to update the price prediction model of the virtual power plant using the comprehensive reward value; The generation module is used to generate the predicted electricity price sequence for the current price prediction period by using the updated price prediction model, combined with historical electricity spot market price data and current market environment information.

9. An electronic device, characterized in that, include: At least one processor, and a memory communicatively connected to said at least one processor, wherein: The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the method as described in any one of claims 1-7.

10. A storage medium, characterized in that, When the computer program in the storage medium is executed by the processor of the electronic device, the electronic device is able to perform the method as described in any one of claims 1-7.

Images