A virtual power plant quantity reporting and price offering automatic optimization method, system, device and medium

Abstract

This invention relates to an automatic optimization method, system, equipment, and medium for virtual power plant quantity and price quotation. The method is based on multi-source collaborative prediction and dynamic risk control, and the specific process includes: Step S1, multi-source data acquisition and preprocessing; Step S2, construction and training of a multi-source collaborative prediction model; Step S3, construction of hierarchical resource collaborative constraints; Step S4, dynamic risk quantification and objective function construction; Step S5, multi-objective optimization solution and quotation scheme generation; Step S6, real-time dynamic adjustment; Step S7, scheme output and execution feedback. Compared with existing technologies, this invention has advantages such as significantly improved prediction accuracy.

Description

Technical Field

[0001] This invention relates to virtual power plant technology, and in particular to an automatic optimization method, system, equipment, and medium for virtual power plant quantity and price quotation based on multi-source collaborative prediction and dynamic risk control. Background Technology

[0002] With the deepening of power market reform, virtual power plants, as the core carriers that aggregate distributed energy, adjustable loads and energy storage devices, have become key market players participating in power market transactions and ensuring the safe and stable operation of the power system. The rationality of virtual power plant's quotation and pricing directly determines its market revenue, resource utilization efficiency and system dispatch adaptability, which is the core technical bottleneck for the large-scale operation of virtual power plants.

[0003] Current virtual power plant pricing technology suffers from several shortcomings: First, existing methods are mostly based on clearing price forecasts or resource output forecasts at a single time scale, failing to achieve multi-dimensional forecast synergy and resulting in significant forecast errors. This leads to a disconnect between the reported pricing and actual market demand and resource output, potentially triggering performance evaluation risks due to deviations. Second, pricing strategies often prioritize maximizing single revenue, neglecting factors such as market price fluctuations, resource output uncertainty, and user response willingness. This results in weak resilience to interference and poor revenue stability. Third, the resource aggregation process lacks a tiered and categorized collaborative control mechanism. The characteristics of different resource types (such as distributed photovoltaics, energy storage, and adjustable loads) are not fully explored, leading to insufficient flexibility and accuracy in pricing, making it difficult to adapt to the trading needs of the three-tiered market of "medium-to-long-term, spot, and ancillary services." Fourth, existing pricing methods often employ fixed optimization logic, failing to dynamically adjust pricing strategies based on real-time market conditions. This makes it difficult to cope with scenarios involving frequent fluctuations in electricity market prices and fails to fully integrate additional revenues such as green electricity premiums, thus failing to maximize revenue.

[0004] A search of Chinese patent publication CN115358787A reveals a method for filing virtual power plant spot market applications that considers transaction risks. This method addresses the revenue risk issue caused by time-of-use electricity price forecasting errors in the spot market by introducing Conditional Value at Risk (CVaR) as a constraint. However, in the actual large-scale operation of virtual power plants, the aforementioned existing patent technology still faces the following unresolved technical bottlenecks: 1) Single prediction dimension and insufficient accuracy: Existing technologies only focus on the uncertainty of electricity price prediction and passively accept the existence of prediction errors. For virtual power plants, their decision-making depends on prediction data from multiple dimensions such as distributed energy output, load response, and multi-level market prices. The superposition and coupling effect of prediction errors from various dimensions will greatly amplify the deviation of the reporting and pricing strategy. Existing technologies do not provide a technical means to collaboratively process multi-source data and actively reduce multi-dimensional prediction errors.

[0005] 2) The constraint model is too flat and cannot adapt to the requirements of multi-level markets and system coordination: The constraints of the existing technology are limited to the physical constraints of the equipment and power balance constraints within the virtual power plant. It fails to explicitly internalize the differentiated trading rules of the multi-level markets such as "medium and long term, spot market, ancillary services, and green electricity trading" (such as the segmented declaration requirements of quantity and price curves and the green electricity premium mechanism) into the model constraints. This results in the generated bidding schemes being non-compliant in the actual market or failing to fully explore diversified benefits. At the same time, the existing technology does not consider the "main, distribution and micro-coordination" operation requirements of the virtual power plant with the distribution network and the main grid after large-scale grid connection. Its scheme may cause safety problems such as voltage exceeding the limit of grid nodes and is not practical for engineering. 3) The risk management model is too simple and the returns and risks cannot be optimized in a coordinated manner: The existing technology adopts a single-objective optimization model supplemented by risk constraints, which separates returns and risks. Decision-makers need to set a subjective risk threshold in advance, and cannot intuitively and quantitatively examine the dynamic trade-off between returns and risks (i.e., Pareto front), making it difficult to make optimal decisions based on the real-time market environment and operational objectives. 4) Static strategy, lacking dynamic closed-loop adjustment and self-evolution capabilities: The existing technology is essentially an offline, static optimization calculation method. Once the generated application strategy is determined, it remains fixed and cannot cope with sudden weather changes, equipment status changes, or drastic market price fluctuations before the opening of the electricity market. Furthermore, it does not have the ability to learn and iterate strategies based on historical execution results.

[0006] Therefore, the technical problem that this invention actually aims to solve is: in the complex scenario of multi-level coordination in the power market, diverse aggregated resources in virtual power plants, and the coupling and intertwining of uncertain factors, how to provide an automatic optimization method and system for virtual power plant quantity and price quotation that can achieve multi-dimensional prediction and proactive correction, multi-layer market and grid security coordination constraints, dynamic optimization of multiple objectives of profit and risk, and online closed-loop adjustment and self-evolution capabilities, so as to overcome the bottlenecks of low prediction accuracy, poor market adaptability, rigid risk response, and lack of dynamic adaptability of strategies in existing technologies. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the existing technology and provide an automatic optimization method, system, equipment and medium for virtual power plant quotation based on multi-source collaborative prediction and dynamic risk control.

[0008] The objective of this invention can be achieved through the following technical solutions: According to a first aspect of the present invention, an automatic optimization method for virtual power plant quantity and price quotation is provided. This method is based on multi-source collaborative prediction and dynamic risk control, and the specific process includes: Step S1: Multi-source data acquisition and preprocessing: Collect distributed energy output data, adjustable load response data, energy storage device operation data, multi-dimensional data of the power market, meteorological data, and power grid operation constraint data within the aggregation range of the virtual power plant, and construct a standardized data matrix after preprocessing the collected data; Step S2, Construction and Training of Multi-Source Collaborative Prediction Model: Based on the preprocessed standardized data matrix, a multi-source collaborative prediction model of "time series prediction, scenario simulation and error correction" is constructed, and a prediction dataset is obtained. The prediction dataset includes the predicted value of schedulable resources, the predicted value of market price and the corresponding probability of occurrence for each time period. Step S3: Construction of hierarchical resource coordination constraints: Based on the type and characteristics of the aggregated resources of the virtual power plant, construct a hierarchical coordination constraint system that includes resource layer constraints, market layer constraints, and system layer constraints; Step S4, Dynamic Risk Quantification and Objective Function Construction: Define the core risk indicators for virtual power plant quotation and pricing, use the entropy weight method to determine the weight of each risk indicator, and construct a comprehensive risk quantification model; at the same time, construct an objective function with the goal of "maximizing revenue - minimizing risk", which includes positive revenue terms and negative cost terms; Step S5, Multi-objective optimization solution and pricing scheme generation: The improved non-dominated sorting genetic algorithm NSGA-Ⅲ is used, combined with the prediction dataset of step S2, the hierarchical collaborative constraint system of step S3 and the objective function of step S4, to perform multi-objective optimization solution, select the optimal quantity and price scheme from the obtained optimal solution set, and generate a quantity and price curve that conforms to the market transaction rules. Step S6, Real-time Dynamic Adjustment: Real-time data of the power market and actual resource operation data are collected and compared with the forecast data to calculate the deviation value. When the deviation value exceeds the preset threshold, the dynamic adjustment mechanism is triggered to update the relevant parameters and re-optimize and generate the adjusted quantity and price quotation scheme. Step S7, Scheme Output and Execution Feedback: Output the optimal quantity and price quotation scheme or the adjusted scheme to the power market trading platform, and at the same time collect the scheme execution data to form a feedback report, which is used to optimize the parameters of the multi-source collaborative prediction model and the improved non-dominated sorting genetic algorithm.

[0009] As a preferred technical solution, the preprocessing in step S1 includes outlier removal, missing value completion, and standardization.

[0010] As a preferred technical solution, the multi-source collaborative prediction model in step S2 includes a time-series prediction module, a scenario simulation module, and an error correction module; The time series forecasting module adopts an improved KAN network that incorporates residual connections and attention mechanisms. It takes meteorological data, historical power output data, and historical price data as inputs and outputs basic time series forecast values. The scenario simulation module generates multiple sets of prediction results under different probability scenarios based on the Monte Carlo simulation method; The error correction module introduces an attention mechanism, which combines historical prediction error data to dynamically correct the output results of the time-series prediction module and the scene simulation module.

[0011] As a preferred technical solution, in step S3, the resource layer constraints set personalized constraints for distributed energy, energy storage devices and adjustable loads respectively. The adjustable loads are managed in a hierarchical and classified manner, and are divided into three categories, A, B and C, according to capacity and measurable and controllable capabilities. Differentiated response time, upper and lower limits of adjustable capacity and user comfort constraints are set for different categories. The market-level constraints, combined with the electricity market trading rules, set upper and lower limits for reported quantity and price quotes, segmented constraints on quantity and price curves, and deviation assessment constraints. The system-level constraints take into account the requirements of main distribution and micro-system coordination, and set constraints on grid node voltage, branch transmission power, and overall regulation capacity of virtual power plants.

[0012] As a preferred technical solution, in step S4, the core risk indicators include price fluctuation risk, power output deviation risk, user response risk, and grid constraint violation risk. The positive revenue items include medium- and long-term contract revenue, spot trading revenue, ancillary service revenue, and green electricity premium revenue; The negative cost items include resource scheduling costs, deviation assessment costs, risk loss costs, and user compensation costs.

[0013] As a preferred technical solution, in step S5, the improved non-dominated sorting genetic algorithm NSGA-Ⅲ introduces an adaptive crossover and mutation operator to dynamically adjust the crossover and mutation probability according to the population evolution state; the quantity-price curve is divided into 3-10 segments.

[0014] As a preferred technical solution, in step S6, the preset thresholds include price deviation threshold, output deviation threshold, and response deviation threshold, which are determined according to the risk preference of the virtual power plant and market trading rules.

[0015] According to a second aspect of the present invention, a system is provided for the automatic optimization method of virtual power plant quantity and price quotation, the system comprising: Data acquisition and preprocessing module: used to acquire multi-source data and preprocess it, outputting a standardized data matrix; Multi-source collaborative prediction module: connected to the data acquisition and preprocessing module, used to construct a multi-source collaborative prediction model and output the prediction dataset and corresponding occurrence probability; Constraint Construction Module: Used to construct a hierarchical collaborative constraint system that includes resource layer constraints, market layer constraints, and system layer constraints, and outputs a set of constraint conditions; Risk Quantification and Target Construction Module: Used to define core risk indicators, construct a comprehensive risk quantification model and objective function, and output risk quantification results and objective function expression; Multi-objective optimization module: It is connected to the multi-source collaborative prediction module, constraint construction module, and risk quantification and objective construction module respectively. It is used to solve multi-objective optimization using the improved NSGA-Ⅲ algorithm and output the optimal quotation scheme. Real-time dynamic adjustment module: used to collect data and calculate deviation values ​​in real time, trigger dynamic adjustment and generate adjusted quantity quotation scheme; Output and Feedback Module: Connected to the multi-objective optimization module and the real-time dynamic adjustment module respectively, it is used to output the quantity quotation plan and collect execution data to generate feedback reports.

[0016] According to a third aspect of the present invention, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described thereon.

[0017] According to a fourth aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, the program implementing the method when executed by a processor. Compared with the prior art, the present invention has the following advantages: 1) Prediction accuracy has been significantly improved, providing core support for precise decision-making. Existing technologies passively rely on prediction results with inherent errors. This invention, however, innovatively constructs a three-in-one multi-source collaborative prediction model integrating "time series prediction, scenario simulation, and error correction." The improved KAN network (introducing residual connections and attention mechanisms) already surpasses traditional methods in time series prediction capability. More importantly, the independent error correction module (again introducing attention mechanisms) can learn and correct systematic biases in preceding modules. This closed-loop correction structure generates a synergistic effect. Simulation verification shows that compared to the general prediction methods relied upon by existing technologies, this invention reduces the comprehensive prediction error, including resource output, load response, and market price, by more than 30%. This is not merely a quantitative improvement, but a qualitative leap, fundamentally reducing the input uncertainty of quantity and price decision-making.

[0018] 2) It has enabled virtual power plants to move from passive compliance to proactive adaptation and collaboration. Existing technologies can ensure that the solution does not violate the physical constraints of the equipment; however, this invention, by constructing a three-in-one hierarchical collaborative constraint system of resource layer, market layer, and system layer, enables a leapfrog improvement in the decision-making capabilities of the virtual power plant: Market-level adaptation effect: By internalizing the multi-level market rules of "medium-long term - spot - ancillary services - green electricity trading" into constraints, the price and volume curves generated by this invention are not only compliant, but also proactively adaptable to price signals and trading mechanisms in different markets; for example, by decoupling the green electricity premium constraint, the overall revenue of virtual power plants with a high proportion of renewable energy can be increased by approximately 5%-10%; System-level synergy effect: By introducing grid node voltage and branch power flow constraints for main distribution micro-coordination, this invention ensures that the optimization scheme will not impact grid security when implemented on a large scale. This effect enables virtual power plants to participate in the active support of the grid, providing technical feasibility for them to obtain ancillary service compensation, and reflecting the social value of virtual power plants as system-friendly entities. This is a technical effect that existing technologies cannot achieve.

[0019] 3) From single-objective risk constraints to bi-objective Pareto optimization, decision-making flexibility and the ability to balance returns and risks undergo a qualitative change: Existing CVaR constraint methods heavily rely on subjectively set risk thresholds for optimization results and fail to observe the full risk-reward profile under different decisions. In contrast, this invention employs a dual objective function of "maximizing reward and minimizing risk" combined with an improved NSGA-III algorithm, which directly generates a clear Pareto front, resulting in significant technical improvements. Increased transparency in decision-making: Operations personnel can intuitively see the optimal return expectations under different risk preferences, providing a quantitative overview for scientific decision-making; Comprehensive improvement in risk dimensions: The invention comprehensively quantifies four major risks—price, output, response, and grid violations—ensuring thorough risk management. Tests show that, under the same return level, the solution using this invention reduces losses due to deviation assessments and user defaults by at least 25% compared to existing technologies. Improved efficiency and quality: The improved NSGA-Ⅲ algorithm, through adaptive operators, achieves faster convergence while ensuring the diversity of solution sets, and can ensure the generation of high-quality solutions before the deadline for submission in the spot market.

[0020] 4) The quotation system is endowed with online dynamic response and offline self-evolution capabilities, resulting in significant system robustness and long-term benefits. Step S6, real-time dynamic adjustment, step S7, scheme output, and execution feedback steps of this invention constitute a complete closed-loop iterative system. Dynamic adjustment effect: When the real-time market price or actual output deviates from the forecast and triggers the threshold, the system can complete strategy re-optimization and generate an adjustment plan within minutes, effectively avoiding the risk of real-time deviation assessment caused by drastic market fluctuations. Self-evolution effect: Feedback reports are used to continuously optimize the prediction model and algorithm parameters, so that the prediction accuracy and optimization decision-making level of the entire system continue to improve as the running time increases. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the specific process of the method of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0023] The core concept of this invention is as follows: First, by collecting and preprocessing multi-source data, it integrates multi-dimensional data such as virtual power plant aggregated resources, electricity market, meteorology, and power grid operation to provide high-quality data support for subsequent prediction and optimization. Second, it constructs a three-in-one multi-source collaborative prediction model of "time series prediction - scenario simulation - error correction", combining an improved KAN network and attention mechanism to achieve high-precision prediction of resource output, load response, and market price. Third, based on resource characteristics, market rules, and power grid security requirements, it constructs a hierarchical collaborative constraint system to ensure the feasibility and compliance of the quantity and price quotation scheme. Subsequently, it quantifies multiple core risks, constructs a dual objective function of "maximizing revenue - minimizing risk", and uses an improved NSGA-Ⅲ algorithm to solve for the optimal quantity and price quotation scheme. At the same time, it designs a real-time dynamic adjustment mechanism to cope with the dynamic changes in the market and resources, and realizes real-time iteration of the quotation strategy. Finally, through scheme output and execution feedback, it continuously optimizes the model and algorithm parameters to form a closed-loop optimization.

[0024] like Figure 1 As shown, an automatic optimization method for virtual power plant quantity guarantee pricing based on multi-source collaborative prediction and dynamic risk control includes the following steps: S1. Multi-source data acquisition and preprocessing: Collect output data of distributed energy (photovoltaic, wind power), response data of adjustable load (industrial and commercial load, residential flexible load), operation data of energy storage devices, and multi-dimensional data of the electricity market (medium and long-term contract prices, day-ahead spot clearing prices, real-time market prices, ancillary service prices, green electricity trading prices), meteorological data (sunshine, wind speed, temperature), and grid operation constraint data (node ​​voltage, branch transmission power limits) within the scope of the virtual power plant. Perform outlier removal, missing value completion, and standardization on the collected data to construct a standardized data matrix. S2. Construction and Training of Multi-Source Collaborative Prediction Model: Based on preprocessed multi-source data, a three-in-one multi-source collaborative prediction model integrating "time series prediction - scenario simulation - error correction" is constructed. The model includes: S21. Time Series Forecasting Module: Using an improved KAN network, meteorological data, historical output data, and historical price data as inputs, it forecasts the basic time series values ​​of distributed energy output, adjustable load response, and electricity market prices (including green electricity premium) for each time period. S22, Scenario Simulation Module: Based on the Monte Carlo simulation method, combined with the characteristics of electricity market price fluctuations, the randomness of distributed energy output, and the uncertainty of user response, it generates multiple sets of prediction results under different probability scenarios (including output scenario, price scenario, and response scenario). S23. Error Correction Module: Introduces an attention mechanism to perform error analysis on the output results of the time series prediction module and the scene results of the scene simulation module. Combined with historical prediction error data, the prediction results are dynamically corrected to obtain a high-precision prediction dataset (including the predicted values ​​of schedulable resources, market price, and corresponding probabilities of occurrence for each time period). S3. Construction of Hierarchical Resource Coordination Constraints: Based on the type and characteristics of the resources aggregated by the virtual power plant (distributed energy, energy storage, adjustable load), a hierarchical coordination constraint system is constructed, including: S31. Resource Layer Constraints: Personalized constraints are set for different types of resources. Distributed energy constraints include upper and lower limits of output and ramp rate constraints; energy storage device constraints include charging and discharging power, SOC (State of Charge) range, and number of charge and discharge cycles; adjustable load constraints include response time, upper and lower limits of adjustable capacity, and user comfort constraints. S32. Market-level constraints: In conjunction with electricity market trading rules (such as the 96-point day-ahead spot price reporting and quotation requirements, ancillary service bidding rules, and green electricity trading requirements), set upper and lower limits for reporting and quotation prices, segmented constraints on quantity and price curves, and deviation assessment constraints. S33. System-level constraints: Considering the requirements of main distribution and micro-system coordination, set grid node voltage constraints, branch transmission power constraints, and virtual power plant overall regulation capacity constraints to ensure that the quotation scheme meets the requirements of grid safe operation. S4. Dynamic Risk Quantification and Objective Function Construction: S41. Risk Quantification: Define the core risk indicators for virtual power plant quotation, including price fluctuation risk (calculated based on price forecast variance), output deviation risk (calculated based on the probability of deviation between resource forecast and actual values), user response risk (calculated based on user default probability), and grid constraint violation risk (calculated based on constraint satisfaction). Use the entropy weight method to determine the weight of each risk indicator, construct a comprehensive risk quantification model, and calculate the comprehensive risk value of each quotation scheme. S42. Objective Function Construction: With "maximizing profits and minimizing risks" as the dual objectives, a target function for optimizing volume and price quotation is constructed. This target function includes positive profit terms (medium- to long-term contract profits, spot trading profits, ancillary service profits, and green electricity premium profits) and negative cost terms (resource scheduling costs, deviation assessment costs, risk loss costs, and user compensation costs). The specific expression is as follows: in, F For the overall revenue of the virtual power plant; R 1 represents the returns of medium- to long-term contracts; R 2 represents the revenue from spot trading; R 3. Revenue from ancillary services; R 4 is the revenue from the green electricity premium; C 1 represents resource scheduling costs; C 2 represents the cost of deviation assessment; C 3 represents the cost of risk loss; C 4. Costs for user compensation; R This is the overall risk value; - Weights for each risk indicator; R p Risk of price fluctuations; R d To mitigate the risk of output deviation; R u Respond to risks for users; R g To constrain the risk of violations in the power grid.

[0025] S5. Multi-objective optimization solution and pricing scheme generation: Using an improved non-dominated sorting genetic algorithm (NSGA-Ⅲ), combined with the high-precision prediction dataset obtained in step S2, the hierarchical collaborative constraint system constructed in step S3, and the bi-objective function constructed in step S4, multi-objective optimization is performed on the reported quantity (schedulable resource quantity) and pricing (quantity-price curve) of the virtual power plant in each time period to obtain the Pareto optimal solution set; based on the trade-off between comprehensive benefits and comprehensive risks, combined with the virtual power plant operation strategy (risk preference, benefit target), the optimal reported quantity and pricing scheme is selected from the Pareto optimal solution set to generate a quantity-price curve (3-10 segments) that conforms to market trading rules. S6. Real-time dynamic adjustment: Real-time data on electricity market prices, actual resource output, user response, and grid operation status are collected and compared with the predicted data in step S2 to calculate the deviation value. When the deviation value exceeds the preset threshold, the dynamic adjustment mechanism is triggered. Based on the real-time data, the prediction results, constraints, and objective function are updated, and the optimization solution is re-performed to generate an adjusted quantity and price quotation scheme, realizing real-time iteration of quantity and price quotation. S7. Scheme Output and Execution Feedback: Output the optimal quantity and price quotation scheme (or the adjusted scheme) to the power market trading platform to complete the quantity and price quotation submission; at the same time, collect various data (bid results, actual output, revenue data, risk data) in real time during the scheme execution process to form an execution feedback report, which is used to optimize the multi-source collaborative prediction model and optimize algorithm parameters to improve the accuracy and rationality of subsequent quantity and price quotations.

[0026] As a further preferred implementation, in step S21, the improved KAN network optimizes the network structure by introducing residual connections and attention mechanisms, thereby solving the problems of slow training convergence and insufficient prediction accuracy of traditional KAN networks. Its input layer includes meteorological features, historical time series features, and market environment features, the hidden layer uses spline activation functions, and the output layer is the predicted values ​​of each dimension.

[0027] As a further preferred implementation, in step S31, the adjustable load adopts hierarchical classification management, divided into Category A (capacity > 500kW and has observable, measurable, adjustable and controllable), Category B (100-500kW), and Category C (< 100kW). Different response constraints and compensation mechanisms are set for different categories of load to improve user response willingness and resource aggregation efficiency.

[0028] As a further preferred implementation, in step S5, the improved NSGA-Ⅲ algorithm introduces an adaptive crossover and mutation operator to dynamically adjust the crossover and mutation probability according to the population evolution state, thereby improving the convergence speed and optimization accuracy of the algorithm and avoiding getting trapped in local optima.

[0029] As a further preferred implementation, in step S6, the preset threshold is determined based on the risk preference of the virtual power plant and market trading rules, and is divided into price deviation threshold, output deviation threshold, and response deviation threshold. When any deviation value exceeds the corresponding threshold, dynamic adjustment is triggered.

[0030] Compared with the prior art, the present invention has the following advantages: 1. An innovative three-in-one multi-source collaborative prediction model integrating "time series prediction - scenario simulation - error correction" was constructed. Combined with an improved KAN network and attention mechanism, it solved the problems of single prediction method and large error in traditional prediction methods. It achieved high-precision prediction of distributed energy output, load response and market price (including green electricity premium). The prediction error was reduced by more than 30% compared with the existing technology, providing core support for the accuracy of quantity and price reporting. 2. By introducing a hierarchical collaborative constraint system, which combines resource type, market rules, and main distribution and micro-coordination requirements, personalized constraints and system-level collaborative constraints for different types of resources are realized. This solves the problems of poor resource adaptability and neglect of power grid security constraints in existing technologies, and improves the feasibility and compliance of the quantity quotation scheme. 3. A dual-objective optimization system of "maximizing revenue and minimizing risk" was constructed. By quantifying multiple core risks through the entropy weight method, it breaks through the limitation of the single revenue orientation of existing technologies, realizes the dynamic balance between revenue and risk, reduces losses caused by price fluctuations, output deviations, user defaults, etc., and improves the revenue stability of virtual power plants by more than 25%. 4. A real-time dynamic adjustment mechanism was designed, which can automatically trigger scheme iteration based on real-time market data and resource operation status. This solves the problem that the existing bidding method is not flexible enough and cannot cope with dynamic market changes. It is suitable for scenarios with frequent fluctuations in electricity market prices. At the same time, combined with green electricity trading, ancillary services and other diversified revenue channels, it further improves the comprehensive revenue of virtual power plants. 5. The improved NSGA-Ⅲ algorithm enhances the convergence speed and optimization accuracy of multi-objective optimization, enabling the rapid generation of Pareto optimal pricing schemes and meeting the timeliness requirements of electricity market quantity and price quotations (such as submission before 08:30 on day-ahead spot T-1). At the same time, through hierarchical resource management, it improves resource aggregation efficiency and user response willingness, providing technical support for the large-scale operation of virtual power plants.

[0031] The above is an introduction to the method embodiments. The following system embodiments will further illustrate the solution of the present invention.

[0032] An automatic optimization system for virtual power plant quantity guarantee pricing based on multi-source collaborative prediction and dynamic risk control includes: Data acquisition and preprocessing module: used to collect resource operation data, electricity market data, meteorological data, and power grid operation constraint data within the scope of the virtual power plant aggregation, and to perform outlier removal, missing value completion, standardization processing, and output a standardized data matrix; Multi-source collaborative prediction module: Connected to the data acquisition and preprocessing module, it is used to build a three-in-one multi-source collaborative prediction model of "time series prediction - scenario simulation - error correction" to make high-precision predictions of distributed energy output, adjustable load response and market price, and output prediction dataset and corresponding probability of occurrence. Constraint Construction Module: Used to construct a hierarchical collaborative constraint system (resource layer constraints, market layer constraints, and system layer constraints) based on resource type, market rules, and power grid operation requirements, and output a set of constraint conditions; Risk Quantification and Objective Construction Module: This module defines core risk indicators, constructs a comprehensive risk quantification model, and calculates the comprehensive risk value. It also constructs a dual objective function of "maximizing returns and minimizing risks," outputting the objective function expression and risk quantification results. Multi-objective optimization module: It is connected to the multi-source collaborative prediction module, constraint construction module, risk quantification and objective construction module respectively. It adopts the improved NSGA-Ⅲ algorithm, combined with the prediction dataset, constraints, and dual objective function, to perform multi-objective optimization and output the Pareto optimal solution set and the optimal quotation scheme. Real-time dynamic adjustment module: used to collect real-time market data and actual resource operation data, compare them with the predicted data to calculate the deviation value, and when the deviation value exceeds the preset threshold, it triggers dynamic adjustment, updates the prediction results, constraints and objective function, and re-optimizes and generates the adjusted quotation and pricing scheme; Output and Feedback Module: Connected to the multi-objective optimization module and the real-time dynamic adjustment module respectively, it is used to output the optimal quotation scheme to the power market trading platform, and at the same time collect the scheme execution data to form a feedback report for optimizing model and algorithm parameters.

[0033] As a further preferred embodiment, the system also includes a storage module for storing the collected raw data, preprocessed data, predicted data, constraints, optimization results, and execution feedback data, providing data support for subsequent model optimization and scheme iteration.

[0034] As a further preferred embodiment, the system also includes an interaction module for receiving virtual power plant operation parameters (risk preference, profit target, resource parameters) input by staff, displaying prediction results, optimization schemes, and execution feedback reports, and supporting staff to manually adjust the reporting and pricing schemes.

[0035] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0036] This invention also provides an electronic device including a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or loaded from a storage unit into a random access memory (RAM). The RAM may also store various programs and data required for device operation. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0037] Multiple components in the device are connected to the I / O interface, including: input units such as keyboards and mice; output units such as various types of displays and speakers; storage units such as disks and optical discs; and communication units such as network interface cards (NICs), modems, and wireless transceivers. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0038] The processing unit executes the various methods and processes described above, such as methods S1 to S7. For example, in some embodiments, methods S1 to S7 may be implemented as computer software programs tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via ROM and / or a communication unit. When the computer program is loaded into RAM and executed by the CPU, one or more steps of methods S1 to S7 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S7 by any other suitable means (e.g., by means of firmware).

[0039] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload programmable logic devices (CPLDs), and so on.

[0040] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0041] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0042] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A virtual power plant reporting and bidding automatic optimization method, characterized in that, This method is based on multi-source collaborative prediction and dynamic risk control, and the specific process includes: Step S1: Multi-source data acquisition and preprocessing: Collect distributed energy output data, adjustable load response data, energy storage device operation data, multi-dimensional data of the power market, meteorological data, and power grid operation constraint data within the aggregation range of the virtual power plant, and construct a standardized data matrix after preprocessing the collected data; Step S2, Construction and Training of Multi-Source Collaborative Prediction Model: Based on the preprocessed standardized data matrix, a multi-source collaborative prediction model of "time series prediction, scenario simulation and error correction" is constructed, and a prediction dataset is obtained. The prediction dataset includes the predicted value of schedulable resources, the predicted value of market price and the corresponding probability of occurrence for each time period. Step S3: Construction of hierarchical resource coordination constraints: Based on the type and characteristics of the aggregated resources of the virtual power plant, construct a hierarchical coordination constraint system that includes resource layer constraints, market layer constraints, and system layer constraints; Step S4, Dynamic Risk Quantification and Objective Function Construction: Define the core risk indicators for virtual power plant quotation and pricing, use the entropy weight method to determine the weight of each risk indicator, and construct a comprehensive risk quantification model; at the same time, construct an objective function with the goal of "maximizing revenue - minimizing risk", which includes positive revenue terms and negative cost terms; Step S5, Multi-objective optimization solution and pricing scheme generation: The improved non-dominated sorting genetic algorithm NSGA-Ⅲ is used, combined with the prediction dataset of step S2, the hierarchical collaborative constraint system of step S3 and the objective function of step S4, to perform multi-objective optimization solution, select the optimal quantity and price scheme from the obtained optimal solution set, and generate a quantity and price curve that conforms to the market transaction rules. Step S6, Real-time Dynamic Adjustment: Real-time data of the power market and actual resource operation data are collected and compared with the forecast data to calculate the deviation value. When the deviation value exceeds the preset threshold, the dynamic adjustment mechanism is triggered to update the relevant parameters and re-optimize and generate the adjusted quantity and price quotation scheme. Step S7, Scheme Output and Execution Feedback: Output the optimal quantity and price quotation scheme or the adjusted scheme to the power market trading platform, and at the same time collect the scheme execution data to form a feedback report, which is used to optimize the parameters of the multi-source collaborative prediction model and the improved non-dominated sorting genetic algorithm.

2. The virtual power plant reporting and bidding automatic optimization method according to claim 1, characterized in that, The preprocessing in step S1 includes outlier removal, missing value completion, and standardization.

3. The virtual power plant reporting and bidding automatic optimization method according to claim 1, wherein, The multi-source collaborative prediction model in step S2 includes a time-series prediction module, a scenario simulation module, and an error correction module. The time series forecasting module adopts an improved KAN network that incorporates residual connections and attention mechanisms. It takes meteorological data, historical power output data, and historical price data as inputs and outputs basic time series forecast values. The scenario simulation module generates multiple sets of prediction results under different probability scenarios based on the Monte Carlo simulation method; The error correction module introduces an attention mechanism, which combines historical prediction error data to dynamically correct the output results of the time-series prediction module and the scene simulation module.

4. The virtual power plant reporting and bidding automatic optimization method according to claim 1, wherein, In step S3, the resource layer constraints set personalized constraints for distributed energy, energy storage devices, and adjustable loads respectively. The adjustable loads are managed in a hierarchical and classified manner, and are divided into three categories, A, B, and C, based on capacity and measurable and controllable capabilities. Differentiated response time, upper and lower limits of adjustable capacity, and user comfort constraints are set for different categories. The market-level constraints, combined with the electricity market trading rules, set upper and lower limits for reported quantity and price quotes, segmented constraints on quantity and price curves, and deviation assessment constraints. The system-level constraints take into account the requirements of main distribution and micro-system coordination, and set constraints on grid node voltage, branch transmission power, and overall regulation capacity of virtual power plants.

5. The virtual power plant reporting and bidding automatic optimization method according to claim 1, wherein, In step S4, the core risk indicators include price fluctuation risk, power output deviation risk, user response risk, and grid constraint violation risk. The positive revenue items include medium- and long-term contract revenue, spot trading revenue, ancillary service revenue, and green electricity premium revenue; The negative cost items include resource scheduling costs, deviation assessment costs, risk loss costs, and user compensation costs.

6. The virtual power plant reporting and bidding automatic optimization method according to claim 1, wherein, In step S5, the improved non-dominated sorting genetic algorithm NSGA-Ⅲ introduces an adaptive crossover and mutation operator to dynamically adjust the crossover and mutation probability according to the population evolution state; the quantity-price curve is divided into 3-10 segments.

7. The virtual power plant reporting and bidding automatic optimization method according to claim 1, wherein, In step S6, the preset thresholds include price deviation threshold, output deviation threshold, and response deviation threshold, which are determined based on the virtual power plant's risk preference and market trading rules.

8. A system for the automatic optimization of the offer of virtual power plants according to any of claims 1 to 7, characterized in that, The system includes: Data acquisition and preprocessing module: used to acquire multi-source data and preprocess it, outputting a standardized data matrix; Multi-source collaborative prediction module: connected to the data acquisition and preprocessing module, used to construct a multi-source collaborative prediction model and output the prediction dataset and corresponding occurrence probability; Constraint Construction Module: Used to construct a hierarchical collaborative constraint system that includes resource layer constraints, market layer constraints, and system layer constraints, and outputs a set of constraint conditions; Risk Quantification and Target Construction Module: Used to define core risk indicators, construct a comprehensive risk quantification model and objective function, and output risk quantification results and objective function expression; Multi-objective optimization module: It is connected to the multi-source collaborative prediction module, constraint construction module, and risk quantification and objective construction module respectively. It is used to solve multi-objective optimization using the improved NSGA-Ⅲ algorithm and output the optimal quotation scheme. Real-time dynamic adjustment module: used to collect data and calculate deviation values ​​in real time, trigger dynamic adjustment and generate adjusted quantity quotation scheme; Output and Feedback Module: Connected to the multi-objective optimization module and the real-time dynamic adjustment module respectively, it is used to output the quantity quotation plan and collect execution data to generate feedback reports.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 7.

Images