A terminal user electricity price multi-factor coupling prediction method and system
By processing multi-source data and calculating system operating costs, and combining retail package parameters, a series of end-user electricity prices is synthesized, which solves the problem that existing technologies fail to fully reflect changes in end-user electricity prices and achieves more accurate electricity price forecasting.
Patent Information
- Application Number
- CN202611254428.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-18
- Publication Date
- 2026-11-10
AI Technical Summary
Existing electricity price forecasting technologies fail to fully reflect the comprehensive changes in end-user electricity prices, ignore important variables in system operating costs, fail to construct a complete price transmission chain, and fail to simulate the strategic behavior of power generators, resulting in underestimation of forecast results.
By acquiring multi-source historical data, standardizing it, and using wholesale-side forecasting models and system operating cost parameters, the end-user electricity price is calculated. Then, by combining retail package parameters and intervening hourly, an end-user electricity price sequence is synthesized.
It enables multi-factor coupled forecasting of end-user electricity prices, reflecting the transmission relationship between wholesale prices, system operating costs, and retail packages, thus improving the accuracy and comprehensiveness of forecasts.
Smart Images

Figure CN123071003A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of electricity market price forecasting technology, and more specifically, to a multi-factor coupled forecasting method and system for end-user electricity prices. Background Technology
[0002] With the advancement of market-oriented reforms in the on-grid electricity price of new energy projects, the electricity generated by new energy projects will, in principle, all enter the electricity market, and the on-grid electricity price will be determined by market transactions. The randomness and volatility of new energy power generation and the time-of-use clearing mechanism of the electricity market create a structural contradiction, posing new challenges to electricity price forecasting technology.
[0003] Currently, existing electricity price forecasting techniques suffer from multiple shortcomings. Existing methods largely focus on generation-side node prices or day-ahead market clearing prices, failing to consider end-user electricity prices as a complete forecasting object. This results in an inability to comprehensively reflect changes in energy costs, including electricity trading prices, transmission and distribution prices, system operating costs, and government funds and surcharges. Regarding system operating costs, renewable energy price differential settlement fees, coal-fired power capacity compensation fees, pumped storage capacity compensation fees, and ancillary service fees have become important variables affecting end-user electricity prices, but existing methods rarely include them as independent variables in their models, and lack structured modeling of renewable energy price differential settlement fees. In terms of retail-side transmission, existing methods fail to construct a complete transmission chain from wholesale market prices to retail package prices and then to end-user settlement prices, lacking refined modeling of the differentiated price transmission mechanisms for different retail packages. Regarding strategic behavior simulation, existing methods generally assume power generators are price takers, ignoring the strategic bidding and price-raising behavior of power generators in an oligopolistic competitive landscape, leading to a systematic underestimation of wholesale electricity price forecasts.
[0004] Therefore, this application provides a multi-factor coupled prediction method and system for end-user electricity prices to solve one of the above-mentioned technical problems. Summary of the Invention
[0005] This application addresses the problems existing in the prior art by providing a multi-factor coupled prediction method and system for end-user electricity prices.
[0006] To achieve the above objectives, the technical solution adopted in this application is as follows: According to a specific embodiment of this application, in a first aspect, this application provides a multi-factor coupled prediction method for end-user electricity prices, including: Acquire standardized datasets generated from multi-source historical data, data for the period to be predicted, wholesale-side forecasting models, system operating cost parameters, and retail package parameters; based on the wholesale-side forecasting models, perform wholesale-side electricity price forecasting on the standardized datasets and data for the period to be predicted, obtaining day-ahead market price sequences, real-time market price sequences, and unified settlement point price sequences; calculate the per-kilowatt-hour allocation value of coal-fired power capacity compensation and pumped storage capacity compensation sequentially based on the system operating cost parameters, and calculate the per-kilowatt-hour allocation value of renewable energy price difference settlement fees based on the system operating cost parameters and the day-ahead market price sequences. The ancillary service rate in the system operating cost parameters is used as the ancillary service cost per kilowatt-hour allocation value to obtain a set of system operating cost item values. The allocation values per kilowatt-hour in the system operating cost item value set are added together to obtain the total allocation value of system operating costs per kilowatt-hour. Based on the retail package parameters and the unified settlement point price sequence, the retail-side settlement electricity price sequence is obtained according to the price transmission relationship corresponding to the retail package parameters. Based on the total allocation value of system operating costs per kilowatt-hour, transmission and distribution prices, government funds, and line loss fees, the retail-side settlement electricity price sequence is intervened hourly to obtain the end-user electricity price sequence.
[0007] According to a specific embodiment of this application, in a second aspect, this application provides a multi-factor coupled prediction system for end-user electricity prices, comprising: The wholesale-side electricity price forecasting module is used to forecast wholesale-side electricity prices based on a standardized dataset generated from multi-source historical data and data for the period to be predicted, according to a wholesale-side forecasting model, to obtain day-ahead market price series, real-time market price series, and unified settlement point price series. The system operating cost calculation module is used to calculate the per-kilowatt-hour allocation value of coal-fired power capacity compensation and pumped-storage capacity compensation sequentially based on system operating cost parameters, and to calculate the per-kilowatt-hour allocation value of renewable energy price difference settlement fees based on the system operating cost parameters and the day-ahead market price series, using the ancillary service rate in the system operating cost parameters as the ancillary service fee. The system operates on a cost per kilowatt-hour (kWh) allocation value is used to obtain a set of system operating cost item values. The total system operating cost per kWh allocation value is obtained by summing the kWh allocation values in this set. A retail package transmission module is used to obtain a retail-side settlement electricity price sequence based on retail package parameters and the unified settlement point price sequence, according to the price transmission relationship corresponding to the retail package parameters. An end-user electricity price synthesis module is used to intervene hourly in the retail-side settlement electricity price sequence based on the total system operating cost per kWh allocation value, transmission and distribution prices, government funds, and line loss costs, to obtain the end-user electricity price sequence.
[0008] Compared with the prior art, this application has the following advantages: This application provides a multi-factor coupled prediction method for end-user electricity prices. By acquiring standardized datasets, data for the period to be predicted, wholesale-side prediction models, system operating cost parameters, and retail package parameters, it predicts wholesale-side electricity prices to obtain day-ahead market price sequences, real-time market price sequences, and unified settlement point price sequences. Based on this, it sequentially calculates and summarizes the per-kilowatt-hour allocation values for coal-fired power capacity compensation, pumped-storage capacity compensation, renewable energy price difference settlement fees, and ancillary service fees to obtain the total per-kilowatt-hour allocation value of system operating costs, incorporating system operating costs into the prediction system in a structured, itemized manner. Simultaneously, based on retail package parameters and the unified settlement point price sequence, it obtains the retail-side settlement price sequence according to the corresponding price transmission relationships, realizing the price transmission from the wholesale market to the retail side. Furthermore, based on the total per-kilowatt-hour allocation value of system operating costs, transmission and distribution prices, government funds, and line loss fees, it intervenes hourly in the retail-side settlement price sequence to synthesize the end-user electricity price sequence. Attached Figure Description
[0009] Figure 1 A flowchart of a multi-factor coupled prediction method for end-user electricity prices; Figure 2 This is a flowchart of the processing of the hybrid prediction model; Figure 3 A flowchart illustrating the branches of a game theory prediction model; Figure 4 Flowchart for the safety-constrained economic scheduling optimization mode; Figure 5 A flowchart for adjusting grid prices at the high-proportion renewable energy transmission end; Figure 6 A flowchart for calculating the per-kilowatt-hour allocation value of new energy price difference settlement fees; Figure 7 Branch flowchart for price transmission in retail packages; Figure 8 A flowchart for predicting and quantifying risk; Figure 9 The functional structure block diagram of the multi-factor coupled prediction system for end-user electricity prices. Detailed Implementation
[0010] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0011] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the application. The singular forms “a,” “said,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.
[0012] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.
[0013] It should be understood that although the terms first, second, third, etc., may be used in the embodiments of this application, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, first may also be referred to as second without departing from the scope of the embodiments of this application, and similarly, second may also be referred to as first.
[0014] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”
[0015] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a product or system comprising a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a product or system. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the product or system that includes said element.
[0016] The present application will now be further described with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present application and should not be construed as limiting the scope of protection of the present application. It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present application.
[0017] The embodiments of this application will be described in detail below with reference to flowcharts and system structure diagrams.
[0018] In this embodiment of the application, the multi-factor coupled prediction method for end-user electricity prices is executed by a computer device.
[0019] Figure 1 The flowchart shows a multi-factor coupled prediction method for end-user electricity prices.
[0020] Reference Figure 1 This includes the following steps: S101. Obtain standardized datasets generated from multi-source historical data, data for the period to be predicted, wholesale-side prediction models, system operating cost parameters, and retail package parameters.
[0021] S102. Based on the wholesale-side forecasting model, conduct wholesale-side electricity price forecasting on the standardized dataset and the data for the period to be forecasted, and obtain the day-ahead market price series, the real-time market price series, and the unified settlement point price series.
[0022] S103a. Calculate the per-kilowatt-hour allocation value of coal-fired power capacity compensation and pumped storage capacity compensation in sequence according to the system operating cost parameters. Calculate the per-kilowatt-hour allocation value of new energy price difference settlement fees according to the system operating cost parameters and the day-ahead market price series. Use the ancillary service rate in the system operating cost parameters as the per-kilowatt-hour allocation value of ancillary service fees to obtain the set of system operating cost item values.
[0023] S103b. Based on the retail package parameters and the unified settlement point price sequence, the retail-side settlement electricity price sequence is obtained according to the price transmission relationship corresponding to the retail package parameters.
[0024] For example, S104a is executed after S103a has finished.
[0025] S104a. Add up the per-kilowatt-hour allocation values in the system operating cost item set to obtain the total per-kilowatt-hour allocation value of system operating costs.
[0026] For example, after S103b and S104a have both been executed, S105 is then executed.
[0027] S105. Based on the total allocated value of system operating costs per kilowatt-hour, transmission and distribution prices, government funds and line loss costs, the retail settlement electricity price sequence is intervened hourly to obtain the end-user electricity price sequence.
[0028] In this embodiment, after the wholesale electricity price forecast is completed, the system operating cost calculation branch uses the day-ahead market price series, and the retail package transmission branch uses the unified settlement point price series. Specifically, the two branches can be executed separately. The system operating cost calculation branch sequentially obtains the set of system operating cost item values and the total allocated value of system operating costs per kilowatt-hour, while the retail package transmission branch obtains the retail settlement electricity price series. Based on this, the end-user electricity price synthesis is performed after obtaining the total allocated value of system operating costs per kilowatt-hour and the retail settlement electricity price series.
[0029] In this application, historical data from multiple sources can be collected through a data interface.
[0030] As one specific implementation method, multi-source historical data includes electricity market operation data, renewable energy operation data, system operation cost data, user-side data, and grid operation boundary data. Specifically, electricity market operation data includes day-ahead market clearing prices, real-time market clearing prices, and medium- and long-term contract transaction prices; renewable energy operation data includes 96-point predicted output curves and actual output curves for wind farms and photovoltaic power plants, as well as renewable energy installed capacity and its proportion; system operation cost data includes coal-fired power capacity compensation fees, pumped storage capacity compensation fees, renewable energy price difference settlement fees, and ancillary service fees; user-side data includes time-of-use load curves for industrial and commercial users, electricity sales volume and proportion for various types of users, and retail package contract signing and settlement data; grid operation boundary data includes load forecast curves, transmission channel plans, and key section constraint information.
[0031] For example, computer devices use forward padding and backward padding to complete missing data.
[0032] For example, for each data column, the mean and standard deviation of the data column are calculated, data points outside the interval from the mean minus three standard deviations to the mean plus three standard deviations are truncated to the corresponding interval boundary, and data from different sources are unified to a 15-minute time granularity.
[0033] For example, the corresponding data is organized to obtain a dataset with a unified time base.
[0034] For example, a computer device extracts hour, minute, weekday, and month features from a timestamp and performs sine coding on the hour feature according to the following formula: ; in, The result of the hourly sine code. The hour value in the timestamp is used to construct the time feature in the standardized dataset; Computer equipment performs cosine encoding of the hour feature according to the following formula: ; in, The result of hourly cosine encoding. The hour value in the timestamp is used to construct the time feature in the standardized dataset.
[0035] For example, computer equipment can also extract historical prices as features from spot price series with lags of one, two, three, six, twelve, twenty-four, forty-eight, and ninety-six time steps. Based on this, after data processing and feature construction, a standardized dataset is obtained for use in wholesale electricity price forecasting and forecast risk quantification.
[0036] In this embodiment, the data for the period to be predicted includes the photovoltaic output, wind power output, load, and corresponding feature vectors used in the wholesale electricity price prediction for the period to be predicted. The wholesale-side prediction mode specifies the processing method for the wholesale electricity price, the system operating cost parameter is used to calculate the per-kilowatt-hour allocation of various system operating costs, and the retail package parameter is used to determine the retail package type and the corresponding price transmission relationship.
[0037] In this embodiment, the computer equipment processes the standardized dataset and the data for the period to be predicted according to the wholesale-side forecasting model. Specifically, after the wholesale-side electricity price forecast is completed, it outputs the day-ahead market price series, the real-time market price series, and the unified settlement point price series. Furthermore, the day-ahead market price series participates in the calculation of renewable energy price difference settlement fees, the unified settlement point price series participates in the transmission of retail packages, and the three price series are also output as components of the wholesale-side electricity price forecast result set.
[0038] As a specific implementation method, when the wholesale-side forecasting model adopts a simplified security-constrained economic scheduling physical model and obtains the day-ahead market price series, the computer equipment calculates the real-time market price at the corresponding moment according to the following formula: ; in, for The real-time market price at any given time, in yuan per megawatt-hour. for The current day's market price, in yuan per megawatt-hour. for The random perturbation term at time t, For each time point, the real-time market prices at each time point constitute a real-time market price sequence; For example, the random disturbance term can follow a normal distribution with a mean of zero and a standard deviation of three.
[0039] For example, the computer equipment calculates the uniform settlement price at the corresponding time using the following formula: ; in, for The unified settlement price at each time point, in yuan per megawatt-hour. for The current day's market price, in yuan per megawatt-hour. for The real-time market price at any given time, in yuan per megawatt-hour. At any given time, the unified settlement point price at each time constitutes a unified settlement point price sequence, which is used for the transmission of retail packages.
[0040] In some embodiments, the wholesale-side prediction model can be a random forest prediction model.
[0041] For example, the computer equipment uses photovoltaic output, wind power output, load, hour, day of the week, hourly sine coding, hourly cosine coding, and historical price lag features from a standardized dataset as historical feature vectors. It then trains a random forest electricity price prediction model using actual historical spot prices as labels. The feature vectors from the data for the period to be predicted are then input into the random forest electricity price prediction model to obtain the day-ahead market price series. Based on this, the day-ahead market price series obtained by the random forest prediction model, together with the real-time market price series and the unified settlement point price series, constitute the wholesale-side electricity price prediction result set.
[0042] In this embodiment of the application, the system operating cost breakdown set includes the per-kilowatt-hour apportionment value of coal-fired power capacity compensation, pumped storage capacity compensation, new energy price difference settlement fees, and ancillary service fees.
[0043] As one specific implementation method, the computer equipment calculates the per-kilowatt-hour allocation value of coal-fired power capacity compensation according to the following formula: ; in, This is the per-kilowatt-hour allocation value for coal-fired power capacity compensation. This refers to the standard electricity price per kilowatt-year for coal-fired power generation. This refers to the installed capacity of coal-fired power plants, expressed in tens of thousands of kilowatts (MW). The total monthly electricity consumption is used to construct the set of system operating cost items, with the per-kilowatt-hour allocation value of coal-fired power capacity compensation being the total monthly electricity consumption.
[0044] For example, the standard price for coal-fired power capacity can be set at 165 yuan per kilowatt-year.
[0045] For example, the computer equipment calculates the per-kilowatt-hour allocation value for pumped storage capacity compensation using the same cost-sharing relationship as coal-fired power capacity compensation: ; in, This is the per-kilowatt-hour allocation value for pumped storage capacity compensation. This refers to the electricity price standard for pumped storage capacity, expressed in yuan per kilowatt-year. This refers to the installed capacity of pumped storage hydroelectric power plants, expressed in tens of thousands of kilowatts (MW). The total monthly electricity consumption is used to construct the set of system operating cost items, with the per-kilowatt-hour allocation value of pumped storage capacity compensation being the total monthly electricity consumption.
[0046] For example, the per-kilowatt-hour allocation of the new energy price difference settlement fee is calculated based on the project type, the mechanism electricity price, the mechanism electricity ratio, the day-ahead market price series, the total new energy power generation, and the total monthly electricity consumption.
[0047] For example, the per-kilowatt-hour allocation of ancillary service fees uses the ancillary service rate in the system operating cost parameters.
[0048] For example, the system operating costs of computer equipment are summarized into a set of items according to the following formula: ; in, This represents the total allocated cost per kilowatt-hour for system operation. This is the per-kilowatt-hour allocation value for coal-fired power capacity compensation. This is the per-kilowatt-hour allocation value for pumped storage capacity compensation. This is the per-kilowatt-hour allocation value for the settlement cost of new energy price differences. The system operation cost per kilowatt-hour is used to synthesize the end-user electricity price, with the ancillary service cost per kilowatt-hour as the total system operation cost per kilowatt-hour.
[0049] In this embodiment of the application, the retail package parameters include the retail package type and the corresponding price parameters.
[0050] In this application, the computer equipment selects a price transmission relationship based on the retail package type, such as fixed price, price difference sharing, floating rate, or tiered pricing, and obtains the retail-side settlement electricity price hourly. Based on this, different retail package branches all use the retail-side settlement electricity price sequence as output, which is used to synthesize the end-user's electricity price.
[0051] For example, computer equipment converts the unit price of retail electricity using the following formula: ; in, for The retail settlement price of electricity is calculated in yuan per kilowatt-hour. for The retail settlement price of electricity is calculated in yuan per megawatt-hour. At any given time, the retail-side settlement price is used to synthesize the end-user electricity price.
[0052] For example, computer equipment synthesizes the end-user electricity price hourly according to the following formula: ; in, for The end-user electricity price at any given time. for Retail-side electricity price at any given time For transmission and distribution tariffs, This represents the total allocated cost per kilowatt-hour for system operation. As a government fund, The cost of line loss is calculated hourly for each cost item, which constitutes the end-user electricity price sequence.
[0053] In this embodiment, day-ahead market price series, real-time market price series, and unified settlement point price series are obtained through wholesale-side electricity price forecasting. The day-ahead market price series is used for system operating cost calculation, and the unified settlement point price series is used for retail package transmission. The total system operating cost per kilowatt-hour allocation and the retail-side settlement price series are jointly incorporated into the end-user electricity price synthesis, so that the end-user electricity price forecast can reflect the transmission relationship between wholesale-side prices, system operating costs, and retail packages, while retaining the real-time market price series as the wholesale-side forecast result.
[0054] Figure 2 This is a flowchart of the mixed prediction model processing.
[0055] Reference Figure 2 This includes the following steps: S201. The wholesale-side forecasting mode is a hybrid forecasting mode, which takes the historical photovoltaic output, historical wind power output and historical load in the standardized dataset as inputs, and uses a simplified safety-constrained economic dispatch physical model to obtain the historical baseline forecast value.
[0056] S202. Calculate the difference between the actual historical spot price and the historical benchmark forecast value in the standardized dataset to obtain the residual target value.
[0057] S203. Train a random forest residual prediction model using historical feature vectors from the standardized dataset as input and residual target values as labels.
[0058] S204a. Simplify the physical model of safe constraint economic dispatch by combining the photovoltaic output, wind power output and load input in the data of the period to be predicted, and obtain the baseline prediction value of the period to be predicted.
[0059] S204b: Input the feature vectors from the data of the period to be predicted into the random forest residual prediction model to obtain the residual prediction values.
[0060] S205. The baseline forecast value and the residual forecast value for the period to be predicted are added together hourly to obtain the day-ahead market price series.
[0061] In this embodiment, the simplified safety-constrained economic dispatch physical model uses photovoltaic power output, wind power output, and load as inputs.
[0062] As a specific implementation method, the simplified physical model for safety-constrained economic scheduling calculates the baseline prediction value according to the following formula: ; in, for The baseline forecast value at that time, in yuan per megawatt-hour. This is the benchmark price for coal-fired power generation, expressed in yuan per megawatt-hour. for Normalized system load at any given time for Normalized photovoltaic output at any given time for Normalized wind power output at any given time This is the load sensitivity coefficient to electricity prices. The effect of contributing to new energy sources on electricity price suppression. for The random noise term at time step, For each time period, the historical baseline forecast value is obtained when historical data is input, and the baseline forecast value for the forecast period is obtained when the data for the forecast period is input.
[0063] For example, the computer device subtracts the actual historical spot prices in the standardized dataset from the corresponding historical benchmark forecasts, and obtains the residual target value according to the following formula: ; in, For the first The target value of the residuals for each historical sample. For the first The actual historical spot price of a historical sample For the first Historical baseline predicted values for a historical sample The historical sample number is used as the training label for the random forest residual prediction model.
[0064] In this embodiment, the historical feature vector includes photovoltaic power output, wind power output, load, hour, weekday, hourly sine code, hourly cosine code, and historical price lag features. Specifically, the random forest residual prediction model is trained using the historical feature vector as input and the residual target value as the label, to learn the residual relationship between the actual historical spot price and the physical model's benchmark prediction value.
[0065] In this embodiment, during the prediction phase, the physical model processes the photovoltaic output, wind power output, and load for the period to be predicted, while the random forest residual prediction model processes the feature vector for the period to be predicted. Specifically, the two processes are executed separately, and after obtaining the baseline prediction value and residual prediction value for the period to be predicted, the computer equipment obtains the prices at each time point in the day-to-day market price sequence according to the following formula: ; in, The day-ahead market price is obtained from the mixed forecast. The base forecast value for the period to be predicted. The two are the residual forecast values, and they are added together at the same time. The hourly calculation results constitute the day-ahead market price series.
[0066] Furthermore, the wholesale-side forecast set also includes real-time market price series and unified settlement point price series.
[0067] In this embodiment, the hybrid prediction provides a baseline prediction through a physical model and uses a random forest residual prediction model to correct the systematic bias reflected in historical data, which is beneficial to take into account both physical processing relationships and historical residual relationships.
[0068] Figure 3 This is a flowchart of the branching process for the game prediction model.
[0069] Reference Figure 3 This includes the following steps: S301, responding to the wholesale-side forecasting mode as a game-theoretic forecasting mode, initializes multiple power generator agents with marginal costs and installed capacity.
[0070] S302. Determine the chosen game mode.
[0071] S303a. In response to the Cournot game model, construct an inverse demand function in which the market clearing price decreases as the total output of all generator agents increases. Calculate the optimal reaction function based on the marginal cost of each generator agent and the inverse demand function, and iterate until convergence to obtain the Nash equilibrium output. Obtain the equilibrium clearing price based on the Nash equilibrium output and the inverse demand function, and use the equilibrium clearing price to obtain the day-ahead market price series.
[0072] S303b: In response to the selection of the Stackelberg game model, the set of leader agents and follower agents is determined according to the installed capacity of each generator agent. The subgame perfect Nash equilibrium is solved by back induction for the set of leader agents and follower agents. The day-to-day market price series is obtained based on the subgame perfect Nash equilibrium.
[0073] S303c responds to the selection of a reinforcement learning game mode based on action value table iteration. An action value table is maintained for each power generator agent. The load level and new energy level are the states, and the discrete price increase rate set is the action. The action is selected according to the action value table and the preset exploration probability. The current action value is added to the product of the learning rate and the update difference to obtain the updated action value. The update difference is the difference obtained by subtracting the current action value from the sum of the products of the reward, the discount factor and the maximum action value of the subsequent state. The action value table is updated with the updated action value until convergence. The day-ahead market price series is obtained using the action value table.
[0074] In this embodiment, a computer device initializes multiple power generation agents. Specifically, each power generation agent has a corresponding marginal cost and installed capacity.
[0075] For example, a generator agent can also have a markup rate with an initial value of zero. Based on this, after initialization, the game proceeds according to the selected game mode, entering either a Cournot game, a Stackelberg game, or a reinforcement learning game branch based on action-value table iteration; these three branches are interchangeable.
[0076] In one feasible embodiment, the computer device performs Cournot game processing and constructs an inverse demand function according to the following formula: ; in, To clear prices in the market, This represents the total output of all power generation agents. For the required intercept parameter, The price sensitivity coefficient is used to calculate the market clearing price corresponding to total output using the inverse demand function.
[0077] For example, for the first For individual e-commerce agents, the computer equipment calculates the optimal reaction function according to the following formula: ; in, For the first The optimal response output for a single e-commerce agent. For the first The marginal cost of an e-commerce agent To exclude the first Total output of hair products from all hair product distributors other than the individual hair product distributor. For the required intercept parameter, Price sensitivity coefficient This is the generator agent's serial number; the optimal response output is used to update the output of the corresponding generator agent.
[0078] For example, the computer equipment iteratively calculates the optimal response output for each generator agent until the outputs of each generator agent converge, yielding the Nash equilibrium output.
[0079] For example, the computer equipment substitutes the Nash equilibrium output into the inverse demand function to obtain the equilibrium clearing price, and uses the equilibrium clearing price to obtain the day-ahead market price series.
[0080] In another feasible embodiment, the computer device performs Stackelberg game processing. Specifically, the computer device determines the set of leader agents and follower agents according to the installed capacity of each generator agent, solves for the subgame perfect Nash equilibrium corresponding to the set of leader agents and follower agents through back induction, and obtains the day-ahead market price sequence based on the subgame perfect Nash equilibrium.
[0081] In another feasible embodiment, the computer device performs reinforcement learning game processing based on iterative action value table iteration. Specifically, the computer device maintains an action value table for each power generation agent, setting the combination of load level and new energy level as the state, and the discrete markup rate set as the action space.
[0082] For example, the computer device selects actions according to the action value table and preset exploration probabilities, and updates the action value according to the following formula: ; in, For state Select action The value of the action obtained after the update The value of the action before the update. For learning rate, For the first The reward that an e-commerce agent receives after performing certain actions. As a discount factor, For the subsequent state, This represents the maximum action value in subsequent states. The updated action value is used to update the action value table as a candidate action for the subsequent state.
[0083] For example, the computer equipment repeatedly selects actions and updates the action value table until the action value table converges, and uses the converged action value table to obtain the day-ahead market price series. Furthermore, the game prediction results also include real-time market price series and unified settlement point price series, which serve as wholesale-side prediction results to provide corresponding price data for subsequent cost calculations and retail package transmission.
[0084] In this embodiment, the three game branches respectively use simultaneous output decision-making, leader and follower sequential decision-making, or action value iteration to describe the strategic behavior of power generators, so that the wholesale electricity price forecast can adapt to the corresponding market decision-making relationship and reduce the forecast bias caused by treating all power generators as price takers.
[0085] Figure 4 The flowchart for the safety-constrained economic scheduling optimization mode.
[0086] Reference Figure 4 This includes the following steps: S401, responding to the wholesale-side forecasting mode as a safety-constrained economic dispatch optimization mode, defines the unit output variables of each unit in each time period.
[0087] S402a, Set a target to minimize total power generation cost.
[0088] S402b, set load balance constraints, unit output upper and lower limit constraints, ramping constraints and line power flow constraints.
[0089] S403. Solve the optimization model consisting of unit output variables, the objective of minimizing total power generation cost, load balance constraints, upper and lower limits of unit output constraints, ramping constraints, and line power flow constraints.
[0090] S404. Extract the dual variables of the load balance constraint in each time period as the nodal marginal electricity price, and obtain the day-ahead market price series based on the nodal marginal electricity price.
[0091] In this embodiment, the computer device defines a set of generating units and a set of time periods, and defines a non-negative, continuous unit output variable for each generating unit in each time period. Specifically, the unit output variable represents the output of the corresponding generating unit in the corresponding time period, in megawatts.
[0092] For example, the computer equipment sets the objective of minimizing the total power generation cost according to the following formula: ; in, For the assembly of units, For time period sets, For unit identification, For time period identification, For the first Taiwanese crew The unit output for a given time period, in megawatts. For the first The marginal cost of each generating unit and the objective of minimizing the total power generation cost are used to determine the unit output combination that satisfies the constraints.
[0093] In this embodiment, load balance constraints are used to constrain the balance between power generation and load in each time period; unit output upper and lower limit constraints are used to limit the allowable output range of each unit; ramping constraints are used to limit the changes in unit output between adjacent time periods; and line power flow constraints are used to limit line power flow. Specifically, the goal of minimizing total power generation cost and all constraints together constitute the safety-constrained economic dispatch optimization model.
[0094] Specifically, the total power generation cost minimization objective and various constraints can be set separately; after the objective and constraints are set, the computer equipment solves the optimization model composed of the unit output variables, the total power generation cost minimization objective, and various constraints.
[0095] As one specific implementation method, the computer equipment uses optimization modeling tools to establish an optimization model and employs a solver to solve the optimization model. Based on this, after the solution is completed, the computer equipment extracts the dual variables of the load balance constraints for each time period, uses these dual variables as the nodal marginal electricity price, and obtains the day-ahead market price series based on the nodal marginal electricity price. Furthermore, the wholesale-side forecast results also include real-time market price series and unified settlement point price series.
[0096] In this embodiment, the goal of minimizing total generation cost is solved together with load balance, unit output, ramping and line power flow constraints. The resulting nodal marginal electricity price can reflect the relationship between generation cost and corresponding grid operation constraints.
[0097] Figure 5 A flowchart for adjusting the grid price at the high-proportion renewable energy transmission end.
[0098] Reference Figure 5 This includes the following steps: S501a: In response to the predicted scenario of a high proportion of renewable energy transmission to the power grid, the external transmission ratio parameter and the external transmission pressure reduction coefficient are read, and the external transmission ratio parameter and the external transmission pressure reduction coefficient are multiplied to obtain the external transmission pressure reduction effect.
[0099] S501b: Read the normalized photovoltaic power output sequence and normalized wind power output sequence from the data of the period to be predicted, calculate the sum of photovoltaic power output in the normalized photovoltaic power output sequence and wind power output in the normalized wind power output sequence hourly, and optimize the value range of the sum based on the numerical reduction term and the product term in turn to obtain the new energy consumption pressure factor sequence.
[0100] S502. Subtract the external transmission pressure reduction effect and the new energy consumption pressure factor at the corresponding time from the prices of the day-ahead market price series to be corrected using the wholesale-side forecasting model to obtain the day-ahead market price series.
[0101] In this embodiment, the outbound ratio parameter is set based on historical outbound data.
[0102] For example, the computer equipment reads the outgoing transmission ratio parameter and the outgoing transmission pressure reduction coefficient, and calculates the outgoing transmission pressure reduction effect according to the following formula: ; in, For external pressure reduction effect, This is the parameter for the external transmission ratio. This is the external transmission pressure reduction coefficient, and the external transmission pressure reduction effect is used to correct the market price series before the correction date.
[0103] For example, the pressure reduction effect of power transmission and the pressure factor of new energy consumption can be calculated separately.
[0104] For example, the computer equipment reads the normalized photovoltaic power output sequence and the normalized wind power output sequence, and calculates the renewable energy consumption pressure factor at each time point according to the following formula: ; in, for The pressure factor of new energy consumption at all times In the normalized photovoltaic power output sequence Solar power output at all times In the normalized wind power output sequence Wind power output at all times At a given time, the sum of photovoltaic output and wind power output minus 0.5 is compared with zero. The larger value is multiplied by 30 to obtain the new energy consumption pressure factor. The new energy consumption pressure factor is used to adjust the market price before the correction date at the corresponding time.
[0105] For example, computer equipment adjusts the market price series for the day before the adjustment according to the following formula: ; in, For the revised The current market price at that moment. The forecast obtained using the wholesale-side forecasting model The market price is subject to revision at any time. For external pressure reduction effect, for The pressure factor of new energy consumption at all times At time t, the correction results at each time point constitute the day-ahead market price sequence.
[0106] In this embodiment, the pressure reduction effect of cross-regional transmission reflects the amount of correction to the price of the sending-end grid, and the new energy absorption pressure factor reflects the amount of price correction after the sum of photovoltaic output and wind power output exceeds the corresponding value. The two together participate in the correction, so that the day-ahead market price forecast can adapt to cross-regional transmission and high new energy output scenarios.
[0107] Figure 6 A flowchart for calculating the per-kilowatt-hour allocation value of the settlement fee for new energy price differences.
[0108] Reference Figure 6 This includes the following steps: S601a. Determine the mechanism electricity price and mechanism electricity ratio based on the item type in the system operation cost parameters.
[0109] S601b: Calculate the mean of the day-ahead market price series and convert the units to obtain the average market transaction price.
[0110] S602a. The larger of the difference between the mechanism-managed electricity price and the average market transaction price and zero is determined as the price difference compensation unit price.
[0111] S602b: Multiply the total renewable energy generation in the system operating cost parameters by the ratio of renewable energy generation to renewable energy generation to obtain the renewable energy generation.
[0112] S603. Multiply the price difference compensation unit price and the mechanism power volume and convert the units to obtain the total price difference settlement cost.
[0113] S604. Divide the total cost of price difference settlement by the total monthly electricity consumption in the system operating cost parameters and perform unit conversion to obtain the per-kilowatt-hour allocation value of the new energy price difference settlement cost.
[0114] In this embodiment, the determination of the mechanism parameters corresponding to the project type and the calculation of the average market transaction price can be performed separately. Specifically, the mechanism electricity price and the average market transaction price are used together to determine the price difference compensation unit price; the mechanism electricity ratio and the total renewable energy power generation are used together to determine the mechanism electricity volume. Based on this, after the price difference compensation unit price and the mechanism electricity volume are obtained, the computer equipment calculates the total price difference settlement cost, and then obtains the renewable energy price difference settlement cost per kilowatt-hour allocation value based on the total price difference settlement cost and the total monthly electricity consumption.
[0115] In this embodiment of the application, the project type is used to determine the mechanism electricity price and mechanism electricity ratio corresponding to the new energy project.
[0116] For example, a computer device calculates the mean of the day-to-day market price series and performs unit conversions using the following formula: ; in, The average market transaction price. This is the average of the current day's market price series, and the average market transaction price is used to calculate the price difference compensation unit price.
[0117] For example, the unit price for price difference compensation for computer equipment is determined according to the following formula: ; in, To compensate for the price difference in unit price, For mechanism-based electricity pricing, The larger of the difference between the market transaction price and the average market transaction price and zero is used as the price difference compensation unit price. The price difference compensation unit price is used to calculate the total cost of price difference settlement.
[0118] For example, the machine-generated electricity consumption and the unit price for price difference compensation can be calculated separately. The computer equipment determines the machine-generated electricity consumption according to the following formula: ; in, For the mechanism's power, This represents the total power generation from new energy sources. The percentage of electricity generated by the mechanism is used to calculate the total cost of the price difference settlement.
[0119] For example, the total cost of computer equipment is calculated using the following formula: ; in, Settle the total cost for the price difference. To compensate for the price difference in unit price, The value 1000 is used for unit conversion and the total cost of price difference settlement is used to calculate the per-kilowatt-hour allocation value of the price difference settlement cost for new energy.
[0120] For example, the computer equipment calculates the per-kilowatt-hour allocation value of the new energy price difference settlement fee according to the following formula: ; in, This is the per-kilowatt-hour allocation value for the settlement cost of new energy price differences. Settle the total cost for the price difference. The total monthly electricity consumption is represented by the value 10000 for unit conversion. The per-kilowatt-hour allocation of the new energy price difference settlement fee is used to form the set of system operation cost items.
[0121] In this embodiment, the mechanism electricity price and mechanism electricity ratio are determined by the project type, the average market transaction price is obtained by using the day-ahead market price series, and the price difference compensation unit price, mechanism electricity, total price difference settlement cost and per-kilowatt-hour apportionment value are calculated in sequence, so that the new energy price difference settlement cost can be included in the system operation cost calculation according to the mechanism price relationship and the corresponding electricity relationship.
[0122] Figure 7 A flowchart illustrating the branching process for pricing in retail packages.
[0123] Reference Figure 7 This includes the following steps: S701. Determine the retail package type based on the retail package parameters.
[0124] S702a: In response to the retail package type being a fixed-price category, the retail-side settlement electricity price at the corresponding time is calculated based on the calculation formula for the fixed-price category.
[0125] For example, the retail settlement electricity price at the corresponding time is calculated according to the following formula: ; in, for The retail settlement price of electricity at any given time, in yuan per megawatt-hour. This refers to the fixed time-of-use price in the retail package parameters, expressed in yuan per megawatt-hour. This is a fixed percentage in the retail package parameters. To unify the settlement point price series The unified settlement price at each time point, in yuan per megawatt-hour. At time t, the calculation results are used to apply capping constraints.
[0126] S703a, impose a capping constraint on the retail settlement electricity price at the corresponding time for fixed-price categories.
[0127] For example, a capping constraint is applied to the retail settlement electricity price at the corresponding time for the fixed-price class according to the following formula: ; Among them, the left side of the equation for Retail electricity price after applying capping constraints, in yuan per megawatt-hour, is shown on the right side of the equation. for Retail electricity settlement price before the capping constraint is applied. This is the capped price in the retail package parameters, expressed in yuan per megawatt-hour. At time t, the retail-side settlement electricity price after applying capping constraints at each time constitutes the retail-side settlement electricity price sequence.
[0128] S702b: In response to the retail package type being price difference sharing, the retail settlement electricity price at the corresponding time is calculated based on the calculation formula of the price difference sharing type.
[0129] For example, the retail settlement electricity price at the corresponding time is calculated according to the following formula: ; in, for The retail settlement price of electricity at any given time, in yuan per megawatt-hour. This is the base price for revenue sharing in the retail package parameters, expressed in yuan per megawatt-hour. To unify the settlement point price series The unified settlement price at each time point, in yuan per megawatt-hour. The user revenue sharing ratio in the retail package parameters, At time t, the calculation results are used to apply capping constraints.
[0130] S703b imposes a capping constraint on the retail settlement electricity price at the time corresponding to the price difference classification.
[0131] For example, a capping constraint is imposed on the retail settlement electricity price at the corresponding time point for each price difference classification, according to the following formula: ; Among them, the left side of the equation for Retail electricity price after applying capping constraints, in yuan per megawatt-hour, is shown on the right side of the equation. for Retail electricity settlement price before the capping constraint is applied. This is the capped price in the retail package parameters, expressed in yuan per megawatt-hour. At time t, the retail-side settlement electricity price after applying capping constraints at each time constitutes the retail-side settlement electricity price sequence.
[0132] S702c, in response to the retail package type being a floating rate type, calculates the retail settlement electricity price at the corresponding time based on the calculation formula for the floating rate type.
[0133] For example, the retail settlement electricity price at the corresponding time is calculated according to the following formula: ; in, for The retail settlement price of electricity at any given time, in yuan per megawatt-hour. To unify the settlement point price series The unified settlement price at each time point, in yuan per megawatt-hour. This refers to the electricity service fee per megawatt-hour (MWH) specified in the retail package parameters. At time t, the calculation results are used to apply capping constraints.
[0134] S703c imposes a capping constraint on the retail settlement electricity price at the corresponding time for floating rate categories.
[0135] For example, a capping constraint is imposed on the retail settlement electricity price at the corresponding time for floating rate classes according to the following formula: ; Among them, the left side of the equation for Retail electricity price after applying capping constraints, in yuan per megawatt-hour, is shown on the right side of the equation. for Retail electricity settlement price before the capping constraint is applied. This is the capped price in the retail package parameters, expressed in yuan per megawatt-hour. At time t, the retail-side settlement electricity price after applying capping constraints at each time constitutes the retail-side settlement electricity price sequence.
[0136] S702d, in response to the retail package type being tiered pricing, determines the tiered electricity price based on the user's monthly electricity consumption and the preset tiered parameter table in the retail package parameters, and sets the tiered electricity price as the retail-side settlement price at each time point, thus obtaining the retail-side settlement price sequence.
[0137] In this embodiment, the retail package type in the retail package parameters is used to select a fixed-price, price difference sharing, floating-rate, or tiered-price transmission branch. Specifically, the four branches are interchangeable, and the retail-side settlement electricity price sequence is used as the common output.
[0138] In one feasible embodiment, the retail package type is a fixed-price type.
[0139] For example, the fixed portion can be 40%. The fixed time-of-use price and the unified settlement point price are jointly included in the calculation of the retail settlement electricity price at the corresponding time according to the fixed portion. The capped price is used to limit the calculation results.
[0140] In another feasible embodiment, the retail package type is a price difference sharing category.
[0141] For example, the user revenue share can be between zero and one. The revenue share base price, the unified settlement point price, and the user revenue share all participate in the calculation of the retail settlement electricity price at the corresponding time. The capped price is used to limit the calculation results.
[0142] In another feasible embodiment, the retail package type is a floating rate type.
[0143] For example, the electricity service fee can range from zero yuan per megawatt-hour to twenty yuan per megawatt-hour. The electricity service fee and the unified settlement point price are used together to calculate the retail settlement electricity price at the corresponding time. The capped price is used to limit the calculation results.
[0144] In another feasible embodiment, the retail package type is a tiered pricing type. Specifically, the computer device reads the user's monthly electricity consumption from the retail package parameters, determines the tiered electricity price based on the user's monthly electricity consumption and a preset tiered parameter table, and uses the tiered electricity price as the retail-side settlement price at each time point to obtain a retail-side settlement price sequence.
[0145] In this embodiment, the corresponding price transmission branch is selected by the retail package type. The fixed price class, the price difference sharing class, and the floating rate class combine the unified settlement point price with the corresponding package parameters. The tiered price class determines the settlement price for all time periods according to the user's monthly electricity consumption, so that the wholesale price can be transmitted to the retail settlement price according to the differentiated retail package relationship.
[0146] Figure 8 A flowchart for quantifying predicted risks.
[0147] Reference Figure 8 This includes the following steps: S801a. Extract the benchmark photovoltaic power output sequence, benchmark wind power output sequence, and benchmark load sequence from the standardized dataset.
[0148] S801b, set the number of simulations, standard deviation of photovoltaic noise, standard deviation of wind power noise, and standard deviation of load noise.
[0149] S802. In each simulation, samples are taken from a normal distribution with a mean of zero according to the standard deviation of photovoltaic noise, wind power noise, and load noise at each time point to obtain the photovoltaic noise sequence, wind power noise sequence, and load noise sequence.
[0150] S803. Add the photovoltaic noise sequence to the reference photovoltaic output sequence, add the wind power noise sequence to the reference wind power output sequence, add the load noise sequence to the reference load sequence, and truncate the photovoltaic output, wind power output and load with added noise to the corresponding specified intervals to obtain the photovoltaic output sequence with noise, the wind power output sequence with noise and the load sequence with noise.
[0151] S804. Using the photovoltaic power output sequence with noise, the wind power output sequence with noise, and the load sequence with noise as the data for the period to be predicted, perform wholesale electricity price prediction, system operating cost calculation, retail package transmission, and end-user electricity price synthesis, and record the mean of the day-ahead market price sequence, the mean of the real-time market price sequence, the total system operating cost per kilowatt-hour, and the mean of the end-user electricity price sequence.
[0152] S805. Determine whether the number of simulations completed has reached the set number of simulations.
[0153] For example, if the set number of simulations is not reached, the noise sampling, noisy sequence generation, electricity price prediction, and index recording from S802 to S804 are repeated.
[0154] S806. In response to reaching the set number of simulations, the loop ends.
[0155] S807. Perform statistical analysis on the mean of the day-ahead market price series, the mean of the real-time market price series, the total system operating cost per kilowatt-hour, and the mean of the end-user electricity price series for each record, and obtain their respective mean, standard deviation, quantiles, and confidence intervals.
[0156] In this embodiment, the prediction risk quantification can employ Monte Carlo simulation. Specifically, the benchmark photovoltaic output sequence, benchmark wind power output sequence, and benchmark load sequence are derived from a standardized dataset. The number of simulations is used to limit the number of iterations, and the standard deviations of photovoltaic noise, wind power noise, and load noise are used to limit the distribution of the corresponding random disturbances.
[0157] For example, the baseline sequence extraction and simulation parameter setting can be performed separately. Once both the baseline sequence and simulation parameters are obtained, the computer equipment enters a loop according to the set number of simulations. In each loop, noise sampling, noisy sequence generation, complete electricity price prediction, and indicator recording are performed sequentially. Statistical analysis is then executed after the set number of simulations is reached.
[0158] In each simulation, the computer equipment samples the photovoltaic noise at each time point according to the following distribution: ; in, for Photovoltaic noise at any time This represents a normal distribution, with zero being the mean of the normal distribution. For the standard deviation of photovoltaic noise, At time t, the photovoltaic noise at each time t constitutes a photovoltaic noise sequence.
[0159] For example, the computer equipment samples wind power noise at various times according to the following distribution: ; in, for Constant wind power noise, This represents a normal distribution, with zero being the mean of the normal distribution. For wind power noise standard deviation, At time t, the wind power noise at each time t constitutes a wind power noise sequence.
[0160] For example, the computer equipment samples the load noise at each time point according to the following distribution: ; in, for Constant load noise, This represents a normal distribution, with zero being the mean of the normal distribution. The standard deviation of load noise, Let be the time interval, and the load noise at each time interval constitutes the load noise sequence.
[0161] For example, the three noise samples act independently on the corresponding reference sequence.
[0162] For example, a computer device generates noisy photovoltaic output according to the following formula: ; in, for Noisy photovoltaic output at all times for The benchmark photovoltaic output at any given time for Photovoltaic noise at any time A truncation function to cut off the sum of two numbers to a specified interval between zero and one. At time t, the photovoltaic output with noise at each time t constitutes a photovoltaic output sequence with noise.
[0163] For example, computer equipment generates noisy wind power output according to the following formula: ; in, for The constant noise-generating wind power output for The benchmark wind power output at any given time. for Constant wind power noise, A truncation function to cut off the sum of two numbers to a specified interval between zero and one. At time t, the wind power output with noise at each time t constitutes a sequence of wind power output with noise.
[0164] For example, a computer device generates a noisy load according to the following formula: ; in, for Constant noise load, for The baseline load at that time. for Constant load noise, A truncation function to cut off the sum to a specified interval between 0.2 and 1. At time t, the noisy loads at each time t constitute a noisy load sequence.
[0165] For example, the computer equipment uses the noisy photovoltaic power output sequence, the noisy wind power output sequence, and the noisy load sequence as the data for the forecast period in this simulation. It then sequentially performs wholesale electricity price forecasting, system operating cost calculation, retail package transmission, and end-user electricity price synthesis. The computer records the day-ahead market price sequence mean, real-time market price sequence mean, total system operating cost per kilowatt-hour allocation, and end-user electricity price sequence mean of this simulation.
[0166] Based on this, after completing the recording of the indicators for this simulation, the computer equipment updates the number of simulations completed. If the number of simulations completed is less than the set number of simulations, noise sampling, noisy sequence generation, complete electricity price prediction, and indicator recording are re-executed; if the number of simulations completed reaches the set number of simulations, the loop ends and statistical analysis is performed on the results of each recording.
[0167] For example, for any recorded metric, the computer device calculates the mean using the following formula: ; in, As an indicator The corresponding mean, For the number of simulations, For the first The index values recorded in this simulation. The simulated serial number is used as the average value, which is then used as the indicator. One of the results of the statistical analysis.
[0168] For example, the standard deviation of computer devices is calculated using the following formula: ; in, As an indicator The corresponding standard deviation For the number of simulations, For the first The index values recorded in this simulation. As an indicator The corresponding mean, The simulation number is used to characterize the distribution of each simulation result relative to the mean.
[0169] Furthermore, the computer equipment determines the corresponding quantiles based on the distribution of each recorded result, and uses the corresponding lower and upper quantiles to determine the confidence intervals. Specifically, the mean, standard deviation, quantiles, and confidence intervals of the day-ahead market price series mean, the real-time market price series mean, the total system operating cost per kilowatt-hour allocated to electricity, and the end-user electricity price series mean are obtained respectively.
[0170] In this embodiment of the application, by applying random perturbations to photovoltaic power output, wind power output, and load and repeatedly executing the complete electricity price prediction chain, statistical distribution information corresponding to wholesale prices, system operating costs, and end-user electricity prices can be obtained, so that the impact of changes in new energy output and load on electricity price prediction results can be expressed in terms of mean, standard deviation, quantiles, and confidence intervals.
[0171] Figure 9 The functional structure block diagram of the multi-factor coupled prediction system for end-user electricity prices.
[0172] Reference Figure 9 The multi-factor coupled prediction system for end-user electricity prices includes a wholesale electricity price prediction module 901, a system operating cost calculation module 902, a retail package transmission module 903, and an end-user electricity price synthesis module 904.
[0173] The wholesale electricity price forecasting module 901 is used to forecast the wholesale electricity price based on the standardized dataset generated from multi-source historical data and the data of the period to be predicted, according to the wholesale forecasting model, to obtain the day-ahead market price series, the real-time market price series and the unified settlement point price series. The system operating cost calculation module 902 is used to calculate the per-kilowatt-hour allocation value of coal-fired power capacity compensation and pumped storage capacity compensation in sequence according to the system operating cost parameters, calculate the per-kilowatt-hour allocation value of new energy price difference settlement fees according to the system operating cost parameters and the day-ahead market price series, take the ancillary service fee rate in the system operating cost parameters as the per-kilowatt-hour allocation value of ancillary service fees, obtain the set of system operating cost item values, and add up the per-kilowatt-hour allocation values in the set of system operating cost item values to obtain the total per-kilowatt-hour allocation value of system operating costs; The retail package transmission module 903 is used to obtain the retail-side settlement electricity price sequence according to the retail package parameters and the unified settlement point price sequence, and according to the price transmission relationship corresponding to the retail package parameters. The end-user electricity price synthesis module 904 is used to intervene in the retail-side settlement electricity price sequence hourly based on the total allocated value of system operating costs per kilowatt-hour, transmission and distribution prices, government funds and line loss costs, to obtain the end-user electricity price sequence.
[0174] For example, the system also includes a prediction risk quantification module 905, which is used to: extract a benchmark photovoltaic output sequence, a benchmark wind power output sequence, and a benchmark load sequence from a standardized dataset; set the number of simulations, the standard deviation of photovoltaic noise, the standard deviation of wind power noise, and the standard deviation of load noise; in each simulation, sample from a normal distribution with a mean of zero according to the standard deviations of photovoltaic noise, wind power noise, and load noise at each time point to obtain the photovoltaic noise sequence, wind power noise sequence, and load noise sequence; add the photovoltaic noise sequence to the benchmark photovoltaic output sequence; add the wind power noise sequence to the benchmark wind power output sequence; add the load noise sequence to the benchmark load sequence; and truncate the photovoltaic output, wind power output, and load with added noise to the corresponding specified intervals to obtain... The simulation process involves obtaining noisy photovoltaic (PV) output sequences, noisy wind power output sequences, and noisy load sequences. These sequences are used as data for the forecast period. Wholesale electricity price forecasting, system operating cost calculation, retail package transmission, and end-user electricity price synthesis are performed to record the mean of the day-ahead market price sequence, the mean of the real-time market price sequence, the total system operating cost per kilowatt-hour, and the mean of the end-user electricity price sequence. After completing the corresponding number of simulation cycles, statistical analysis is performed on the mean of the day-ahead market price sequence, the mean of the real-time market price sequence, the total system operating cost per kilowatt-hour, and the mean of the end-user electricity price sequence recorded for each cycle, yielding their respective means, standard deviations, quantiles, and confidence intervals.
[0175] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.
[0176] Although the operations are described in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order or serial order shown, or requiring all of the operations shown to obtain the desired result. In certain environments, multitasking and parallel processing may be advantageous.
[0177] The methods and systems of this application can be implemented using standard programming techniques, utilizing rule-based logic or other logic to implement various method steps. It should also be noted that the terms "system" and "module" as used herein and in the claims are intended to include implementations using one or more lines of software code and / or hardware implementations and / or devices for receiving input.
[0178] Any step, operation, or procedure described herein may be performed or implemented using one or more hardware or software modules, either alone or in combination with other devices. In one embodiment, the software module is implemented using a computer program product comprising a computer-readable medium containing computer program code, which is executable by a computer processor to perform any or all of the described steps, operations, or procedures.
[0179] The foregoing description of implementations of this application has been provided for illustrative and descriptive purposes. The foregoing description is not exhaustive and is not intended to limit this application to the exact forms disclosed. Various modifications and variations may exist in accordance with the foregoing teachings, or may arise from practice of this application. These embodiments were chosen and described to illustrate the principles of this application and its practical application, enabling those skilled in the art to utilize this application in various implementations and modifications to suit the specific purpose of the concept.
[0180] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.
[0181] It can be further understood that, unless otherwise specified, "connection" includes both direct connections where no other components exist between the two parties and indirect connections where other components exist between them.
[0182] It is further understood that although the operations are described in a specific order in the accompanying drawings in the embodiments of this application, this should not be construed as requiring these operations to be performed in the specific order or serial order shown, or requiring all the operations shown to be performed to obtain the desired result. In certain environments, multitasking and parallel processing may be advantageous.
[0183] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the field of this application that are not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.
[0184] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.
[0185] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
[0186] Finally, it should be noted that the above content is only used to illustrate the technical solution of this application, and is not intended to limit the scope of protection of this application. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of this application shall not depart from the substance and scope of the technical solution of this application.
Claims
1. A multi-factor coupled prediction method for end-user electricity prices, characterized in that, include: Acquire standardized datasets generated from multi-source historical data, data for the period to be predicted, wholesale-side prediction models, system operating cost parameters, and retail package parameters; Based on the wholesale-side forecasting model, wholesale-side electricity prices are predicted on the standardized dataset and the data for the period to be predicted, resulting in day-ahead market price series, real-time market price series, and unified settlement point price series. Based on the system operating cost parameters, the per-kilowatt-hour allocation value of coal-fired power capacity compensation and pumped storage capacity compensation is calculated sequentially. Based on the system operating cost parameters and the day-ahead market price sequence, the per-kilowatt-hour allocation value of new energy price difference settlement fee is calculated. The ancillary service fee rate in the system operating cost parameters is used as the per-kilowatt-hour allocation value of ancillary service fee, thus obtaining the set of system operating cost item values. The total allocated value of system operating costs per kilowatt-hour is obtained by adding up the allocated values of each kilowatt-hour in the set of system operating cost items. Based on the retail package parameters and the unified settlement point price sequence, the retail-side settlement electricity price sequence is obtained according to the price transmission relationship corresponding to the retail package parameters; Based on the total allocated value of system operating costs per kilowatt-hour, transmission and distribution prices, government funds, and line loss costs, the retail-side settlement electricity price sequence is intervened hourly to obtain the end-user electricity price sequence.
2. The multi-factor coupled prediction method for end-user electricity prices according to claim 1, characterized in that, The step of forecasting wholesale electricity prices based on the standardized dataset and the data for the period to be predicted, according to the wholesale-side forecasting model, yields a day-ahead market price series, a real-time market price series, and a unified settlement point price series, including: In response to the wholesale-side forecasting mode being a hybrid forecasting mode, the historical photovoltaic output, historical wind power output, and historical load in the standardized dataset are used as inputs, and a simplified security-constrained economic dispatch physical model is adopted to obtain the historical baseline forecast value. The difference between the actual historical spot price in the standardized dataset and the historical benchmark prediction value is calculated to obtain the residual target value. The historical feature vector in the standardized dataset is used as input, and the residual target value is used as the label to train a random forest residual prediction model. The photovoltaic output, wind power output and load in the data of the period to be predicted are input into the simplified safety-constrained economic dispatch physical model to obtain the baseline prediction value of the period to be predicted. The feature vectors in the data of the period to be predicted are input into the random forest residual prediction model to obtain the residual prediction value; The baseline forecast value for the period to be predicted is added to the residual forecast value hourly to obtain the day-ahead market price series.
3. The multi-factor coupled prediction method for end-user electricity prices according to claim 1, characterized in that, The step of forecasting wholesale electricity prices based on the standardized dataset and the data for the period to be predicted, according to the wholesale-side forecasting model, yields a day-ahead market price series, a real-time market price series, and a unified settlement point price series, including: In response to the fact that the wholesale-side forecasting mode is a game-theoretic forecasting mode, multiple power generation agents with marginal costs and installed capacity are initialized. In response to the selection of the Cournot game model, an inverse demand function is constructed in which the market clearing price decreases as the total output of all generator agents increases; the optimal reaction function is calculated based on the marginal cost of each generator agent and the inverse demand function, and iterated until convergence to obtain the Nash equilibrium output; the equilibrium clearing price is obtained based on the Nash equilibrium output and the inverse demand function, and the day-ahead market price sequence is obtained using the equilibrium clearing price; or In response to the selection of the Stackelberg game model, the leader agent and follower agent sets are determined according to the installed capacity of each power generator agent; for the leader agent and follower agent sets, the subgame perfect Nash equilibrium is solved by back induction, and the day-ahead market price sequence is obtained based on the subgame perfect Nash equilibrium; or In response to the selection of a reinforcement learning game mode based on action value table iteration, an action value table is maintained for each of the power generation agents; with load level and new energy level as states and discrete price markup rate set as actions, actions are selected according to the action value table and preset exploration probabilities, and the current action value is added to the product of the learning rate and the update difference to obtain the updated action value. The action value table is updated with the updated action value until convergence, and the day-ahead market price sequence is obtained using the action value table; wherein, the update difference is the difference obtained by subtracting the current action value from the sum of the products of the return and the discount factor and the maximum action value of the subsequent state.
4. The multi-factor coupled prediction method for end-user electricity prices according to claim 1, characterized in that, The step of forecasting wholesale electricity prices based on the standardized dataset and the data for the period to be predicted, according to the wholesale-side forecasting model, yields a day-ahead market price series, a real-time market price series, and a unified settlement point price series, including: In response to the wholesale-side forecasting mode being a safety-constrained economic dispatch optimization mode, the unit output variables of each unit in each time period are defined, and the total power generation cost minimization objective, load balance constraint, unit output upper and lower limit constraint, ramp constraint and line power flow constraint are set. The optimization model composed of the unit output variables, the total power generation cost minimization objective, the load balance constraint, the unit output upper and lower limit constraint, the ramp constraint and the line power flow constraint is solved. The dual variables of the load balance constraint in each time period are extracted as the nodal marginal electricity price, and the day-ahead market price series is obtained based on the nodal marginal electricity price.
5. The multi-factor coupled prediction method for end-user electricity prices according to claim 1, characterized in that, The step of forecasting wholesale electricity prices based on the standardized dataset and the data for the period to be predicted, according to the wholesale-side forecasting model, yields a day-ahead market price series, a real-time market price series, and a unified settlement point price series, including: In response to the predicted scenario of a high proportion of renewable energy transmission to the power grid, the transmission ratio parameter and the transmission pressure reduction coefficient are read, and the transmission ratio parameter and the transmission pressure reduction coefficient are multiplied to obtain the transmission pressure reduction effect. Read the normalized photovoltaic power output sequence and normalized wind power output sequence from the data of the period to be predicted, calculate the sum of photovoltaic power output in the normalized photovoltaic power output sequence and wind power output in the normalized wind power output sequence hourly, and optimize the value range of the sum based on the numerical reduction term and the product term in turn to obtain the new energy consumption pressure factor sequence. The day-ahead market price series is obtained by subtracting the external transmission pressure reduction effect and the corresponding new energy absorption pressure factor in the new energy absorption pressure factor series from the prices at each time point of the day-ahead market price series obtained using the wholesale-side forecasting model.
6. The multi-factor coupled prediction method for end-user electricity prices according to claim 1, characterized in that, The calculation of the per-kilowatt-hour allocation value of the renewable energy price difference settlement fee based on the system operating cost parameters and the day-ahead market price sequence includes: The mechanism electricity price and mechanism electricity ratio are determined based on the item type in the system operating cost parameters. Calculate the mean of the previous day's market price series and convert the units to obtain the average market transaction price; The larger of the difference between the mechanism-based electricity price and the average market transaction price and zero is determined as the price difference compensation unit price; The total renewable energy generation in the system operating cost parameters is multiplied by the ratio of the organic power generation to obtain the organic power generation. Multiply the price difference compensation unit price and the mechanism power consumption and convert the units to obtain the total price difference settlement cost; The total cost of the price difference settlement is divided by the total monthly electricity consumption in the system operating cost parameters and the units are converted to obtain the per-kilowatt-hour allocation value of the new energy price difference settlement cost.
7. The multi-factor coupled prediction method for end-user electricity prices according to claim 1, characterized in that, The step of obtaining the retail-side settlement electricity price sequence based on the retail package parameters and the unified settlement point price sequence, according to the price transmission relationship corresponding to the retail package parameters, includes: In response to the retail package type being a fixed price in the retail package parameters, the retail settlement electricity price at the corresponding time is calculated according to the following formula: ; in, for Retail-side electricity price at any given time The fixed time-sharing price in the retail package parameters, This refers to a fixed percentage of the retail package parameters. For the unified settlement point price series The unified settlement point price at any given time; The following formula applies a capping constraint to the retail settlement electricity price at the corresponding time point: ; in, for Retail-side electricity price at any given time This is the capped price in the retail package parameters; The retail-side settlement electricity price sequence is obtained based on the retail-side settlement electricity price at each time point; or In response to the retail package type being classified as price difference sharing in the retail package parameters, the retail settlement electricity price at the corresponding time is calculated according to the following formula: ; in, for Retail-side electricity price at any given time The base price for revenue sharing in the retail package parameters. For the unified settlement point price series The unified settlement point price at all times, The user revenue sharing ratio in the retail package parameters; The following formula applies a capping constraint to the retail settlement electricity price at the corresponding time point: ; in, for Retail-side electricity price at any given time This is the capped price in the retail package parameters; The retail-side settlement electricity price sequence is obtained based on the retail-side settlement electricity price at each time point; or In response to the retail package type being a floating rate type in the aforementioned retail package parameters, the retail settlement electricity price at the corresponding time is calculated according to the following formula: ; in, for Retail-side electricity price at any given time For the unified settlement point price series The unified settlement point price at all times, This refers to the electricity service fee per kilowatt-hour in the retail package parameters; The following formula applies a capping constraint to the retail settlement electricity price at the corresponding time point: ; in, for Retail-side electricity price at any given time This is the capped price in the retail package parameters; The retail-side settlement electricity price sequence is obtained based on the retail-side settlement electricity price at each time point; or In response to the retail package type being a tiered pricing type in the retail package parameters, the tiered electricity price is determined based on the user's monthly electricity consumption and the preset tiered parameter table in the retail package parameters. The tiered electricity price is then determined as the retail-side settlement electricity price at each time point, resulting in the retail-side settlement electricity price sequence.
8. The multi-factor coupled prediction method for end-user electricity prices according to claim 1, characterized in that, The method further includes: Extract the benchmark photovoltaic power output sequence, benchmark wind power output sequence, and benchmark load sequence from the standardized dataset; Set the number of simulations, the standard deviation of photovoltaic noise, the standard deviation of wind power noise, and the standard deviation of load noise; In each simulation, for each time point, samples are taken from a normal distribution with a mean of zero according to the standard deviation of photovoltaic noise, the standard deviation of wind power noise, and the standard deviation of load noise to obtain photovoltaic noise sequence, wind power noise sequence, and load noise sequence. The photovoltaic noise sequence is then added to the reference photovoltaic output sequence. The wind power noise sequence is added to the reference wind power output sequence, the load noise sequence is added to the reference load sequence, and the photovoltaic output, wind power output and load with added noise are respectively truncated to the corresponding specified intervals to obtain the photovoltaic output sequence with noise, the wind power output sequence with noise and the load sequence with noise. The noisy photovoltaic power output sequence, the noisy wind power output sequence, and the noisy load sequence are used as the data for the forecast period. Wholesale electricity price forecasting, system operating cost calculation, retail package transmission, and end-user electricity price synthesis are performed to record the mean of the day-ahead market price sequence, the mean of the real-time market price sequence, the total system operating cost per kilowatt-hour, and the mean of the end-user electricity price sequence. After completing the number of simulation cycles, statistical analysis is performed on the mean of the day-ahead market price series, the mean of the real-time market price series, the total system operating cost per kilowatt-hour, and the mean of the end-user electricity price series recorded in each cycle, to obtain their respective mean, standard deviation, quantiles, and confidence intervals.
9. A multi-factor coupled prediction system for end-user electricity prices, characterized in that, include: The wholesale electricity price forecasting module is used to forecast wholesale electricity prices based on a standardized dataset generated from multi-source historical data and data for the period to be forecasted, according to the wholesale forecasting model, to obtain day-ahead market price series, real-time market price series and unified settlement point price series; The system operating cost calculation module is used to calculate the per-kilowatt-hour allocation value of coal-fired power capacity compensation and pumped storage capacity compensation in sequence according to the system operating cost parameters, calculate the per-kilowatt-hour allocation value of new energy price difference settlement fees according to the system operating cost parameters and the day-ahead market price series, take the ancillary service fee rate in the system operating cost parameters as the per-kilowatt-hour allocation value of ancillary service fees, obtain the system operating cost item value set, and add up the per-kilowatt-hour allocation values in the system operating cost item value set to obtain the total per-kilowatt-hour allocation value of system operating costs; The retail package transmission module is used to obtain the retail-side settlement electricity price sequence according to the retail package parameters and the unified settlement point price sequence, and according to the price transmission relationship corresponding to the retail package parameters. The end-user electricity price synthesis module is used to intervene hourly in the retail-side settlement electricity price sequence based on the total allocated value of system operating costs per kilowatt-hour, transmission and distribution prices, government funds, and line loss costs, to obtain the end-user electricity price sequence.
10. The multi-factor coupled prediction system for end-user electricity prices according to claim 9, characterized in that, The system further includes a risk prediction quantification module, which is used for: Extract the benchmark photovoltaic power output sequence, benchmark wind power output sequence, and benchmark load sequence from the standardized dataset; Set the number of simulations, the standard deviation of photovoltaic noise, the standard deviation of wind power noise, and the standard deviation of load noise; In each simulation, for each time point, samples are taken from a normal distribution with a mean of zero according to the standard deviation of photovoltaic noise, the standard deviation of wind power noise, and the standard deviation of load noise to obtain photovoltaic noise sequence, wind power noise sequence, and load noise sequence. The photovoltaic noise sequence is then added to the reference photovoltaic output sequence. The wind power noise sequence is added to the reference wind power output sequence, the load noise sequence is added to the reference load sequence, and the photovoltaic output, wind power output and load with added noise are respectively truncated to the corresponding specified intervals to obtain the photovoltaic output sequence with noise, the wind power output sequence with noise and the load sequence with noise. The noisy photovoltaic power output sequence, the noisy wind power output sequence, and the noisy load sequence are used as the data for the forecast period. Wholesale electricity price forecasting, system operating cost calculation, retail package transmission, and end-user electricity price synthesis are performed to record the mean of the day-ahead market price sequence, the mean of the real-time market price sequence, the total system operating cost per kilowatt-hour, and the mean of the end-user electricity price sequence. After completing the number of simulation cycles, statistical analysis is performed on the mean of the day-ahead market price series, the mean of the real-time market price series, the total system operating cost per kilowatt-hour, and the mean of the end-user electricity price series recorded in each cycle, to obtain their respective mean, standard deviation, quantiles, and confidence intervals.
