Output and cost data processing method of virtual power plant considering air conditioning load

Abstract

The invention provides an output and cost data processing method of a virtual power plant considering an air-conditioning load, and the method comprises the steps: determining the output power of each unit in the virtual power plant in the next day, and the predicted clearing electricity price and cost; inputting the output power of each unit, the clearing electricity price and the cost into a preset day-ahead stage transaction prediction model established based on an information gap decision theory; and obtaining a bid winning result and a corresponding income of the virtual power plant. And in a real-time scheduling stage of the next day, determining an output real-time deviation value of each unit according to the bid winning result and the actual output of each unit. And determining the adjustment plan and the lowest adjustment cost of each unit according to the minimum deviation adjustment total cost. According to the technical scheme, a more flexible and efficient virtual power plant two-stage transaction optimization model is provided, and the objective of the invention is to realize optimal configuration of power resources and maximization of benefits of the virtual power plant through accurate quotation and output plan adjustment.

Description

Technical Field

[0001] The present invention relates to the technical field of power plants, and in particular to a method for processing output and cost data of a virtual power plant taking air conditioning load into account. Background Art

[0002] In existing technologies, virtual power plant trading models, when dealing with electricity price fluctuations and renewable energy output uncertainty, rigidly execute transactions based on day-ahead plans, ignoring real-time changes within the day. This results in insufficient flexibility in actual market operations, making it difficult to achieve optimal economic benefits. Furthermore, day-ahead plans fail to account for the impact of uncertainty, often hindering power plants from maximizing their profits. Summary of the Invention

[0003] The present invention addresses the problems existing in the prior art and provides a method for processing output and cost data of a virtual power plant taking into account air conditioning load. To achieve the above-mentioned purpose, the present invention adopts the following technical solution: A method for processing output and cost data of a virtual power plant taking into account air conditioning load, comprising:

[0004] In the day-ahead phase, the output power of each unit in the virtual power plant for the next day, as well as the predicted clearing electricity price and cost, are determined. The units in the virtual power plant include: a micro gas turbine unit, an energy storage system unit, an electric vehicle unit, an air conditioning unit, and a power grid unit.

[0005] The output power of each unit, as well as the clearing electricity price and cost, are input into a pre-set day-ahead transaction prediction model established based on information gap decision theory; the day-ahead transaction prediction model outputs a winning bid result and corresponding revenue for the virtual power plant; the winning bid result includes: an output plan that the generator sets of each unit need to execute in real time;

[0006] During the real-time scheduling phase of the next day, the real-time output deviation value of each unit is determined based on the winning bid result and the actual output of each unit;

[0007] Inputting the output real-time deviation value into a real-time deviation balance model; the real-time deviation balance model outputs a minimized deviation adjustment total cost;

[0008] The adjustment plan and the minimum adjustment cost of each unit are determined according to the minimized deviation adjustment total cost.

[0009] In some embodiments, the day-ahead transaction prediction model is provided with an objective function, constraints, and a clearing electricity price uncertainty set;

[0010] The objective function is the revenue obtained by the virtual power plant aggregator in the electric energy market;

[0011] The clearing electricity price uncertainty set includes a set of actual values of the market clearing electricity price.

[0012] In some embodiments, the constraints include reporting quantity constraints, micro gas turbine unit operation constraints, wind and solar output constraints, energy storage system charging and discharging power constraints, state of charge constraints, output capacity constraints, electric vehicle cluster and air conditioning output capacity constraints, and response time constraints;

[0013] In some embodiments, the day-ahead transaction prediction model includes: a robust sub-model and an opportunistic sub-model;

[0014] The output power of each unit, as well as the clearing electricity price and cost, are input into a pre-set day-ahead transaction prediction model established based on information gap decision theory; the day-ahead transaction prediction model outputs the winning bid result and corresponding revenue of the virtual power plant, specifically including:

[0015] Inputting the output power of each unit, as well as the clearing electricity price and cost into a pre-set day-ahead transaction prediction model established based on information gap decision theory to obtain a benchmark value of revenue;

[0016] Determining the robustness deviation factor according to the objective function and the risk aversion parameter of the robust sub-model;

[0017] Determining a first profit threshold according to the robustness deviation factor and the profit benchmark value;

[0018] determining a maximum robustness according to the first benefit threshold;

[0019] Determining the opportunity deviation factor according to the objective function and risk aversion parameter of the opportunity sub-model;

[0020] Determining a second profit threshold value according to the opportunistic deviation factor and the profit benchmark value;

[0021] determining a minimum chance degree according to the second profit threshold;

[0022] The winning bid result and corresponding income of the virtual power plant are determined according to the maximum robustness and the minimum chance.

[0023] In some embodiments, the objective function of the real-time deviation balance model is the total deviation adjustment cost of the virtual power plant at any time;

[0024] The total deviation adjustment cost of the virtual power plant at any time includes: micro gas turbine adjustment cost, energy storage system adjustment cost, electric vehicle adjustment cost, air conditioning adjustment cost and grid deviation penalty fee.

[0025] In some embodiments, the adjustment cost of the micro gas turbine is determined based on the unit adjustment cost and the adjustment output of the micro gas turbine;

[0026] determining the energy storage system adjustment cost according to the unit power adjustment cost and the adjustment power of the energy storage system;

[0027] determining the electric vehicle adjustment cost according to the unit power adjustment cost and the adjustment power of the electric vehicle;

[0028] determining the air conditioner adjustment cost according to the unit power adjustment cost and the adjustment power of the air conditioner;

[0029] The grid deviation penalty fee is determined according to the unit power adjustment cost and the adjustment power of the micro gas turbine.

[0030] In some embodiments, the constraints of the real-time deviation balance model include: supply and demand balance constraints, controllable unit operation constraints, energy storage system scheduling constraints, and flexible load constraints.

[0031] In some embodiments, predicting the clearing electricity price includes: using an ARIMA model to predict the clearing electricity price and cost.

[0032] In some embodiments, determining the adjustment plan and the lowest adjustment cost of each unit according to minimizing the total deviation adjustment cost includes:

[0033] A pre-trained machine learning model is used to determine the adjustment plan and minimum adjustment cost of each unit based on the total cost of minimizing the deviation adjustment.

[0034] In some embodiments, the machine learning model includes: a linear regression model, a decision tree model, a random forest model, a neural network model, or a support vector machine model.

[0035] Compared with the existing technology, the present invention has the following beneficial effects: the present application inputs the output power of each unit, as well as the clearing electricity price and cost into a pre-set day-ahead transaction prediction model based on information gap decision theory; the day-ahead transaction prediction model outputs the winning bid result and corresponding revenue of the virtual power plant; the winning bid result includes: the output plan that the generator set of each unit needs to execute in the real-time stage. This is conducive to obtaining a more scientific and reasonable winning bid result. In the real-time scheduling stage of the next day, the real-time deviation model is used to output the total cost of minimizing the deviation adjustment; the adjustment plan and the minimum adjustment cost of each unit are determined based on the total cost of minimizing the deviation adjustment. This is conducive to real-time adjustment of the power plant, thereby reducing costs and increasing revenue. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following detailed description, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings:

[0037] Figure 1 This is a flow chart of a method for processing output and cost data of a virtual power plant taking into account air conditioning load in one embodiment of the present application;

[0038] Figure 2 This is a structural diagram of a virtual power plant in one embodiment of the present application;

[0039] Figure 3 This is a structural diagram of another virtual power plant in an embodiment of the present application. DETAILED DESCRIPTION

[0040] The present invention will be further described below with reference to the accompanying drawings. The following examples are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application.

[0041] A method for processing output and cost data of a virtual power plant taking into account air conditioning load is shown in the appendix. Figure 1 ,include:

[0042] In step S102, in the day-ahead stage, the output power of each unit in the virtual power plant for the next day, as well as the predicted clearing electricity price and cost, are determined.

[0043] Among them, the units in the virtual power plant include: micro gas turbine unit, energy storage system unit, electric vehicle unit, air conditioning unit and power grid unit.

[0044] In this embodiment, see the attached Figure 2-3To address the technical, efficiency, and management challenges associated with the large-scale integration of distributed renewable energy, virtual power plants (VPPs) leverage the Internet of Things (IoT), advanced information and communication technologies, and software systems to aggregate distributed energy resources, including distributed power sources (DGs), electric vehicles, energy storage systems, and flexible loads. This creates a non-physical entity, transcending spatial limitations and allowing it to participate in electricity market transactions as a specialized power plant, thereby improving power system stability. Within the context of power market reform, VPPs participate in electricity market transactions as an aggregated entity while internally coordinating and controlling the output and load plans of their constituent units. This effectively addresses the widespread distribution, large number, small capacity, and difficult management of DGs, enabling flexible resource integration and allocation. VPP operations can be broadly divided into the underlying layers of power dispatching, demanders and responders, the intermediate VPP system operation platform, and the upper layer of the power trading center. VPP operators connect to the power trading center and its aggregated power resources, organizing resource participants to participate in various power market transactions, completing relevant settlements and profit distribution, and settling response compensation with the trading center.

[0045] In step S104, the output power of each unit, as well as the clearing electricity price and cost, are input into a pre-set day-ahead transaction forecasting model based on information gap decision theory. The day-ahead transaction forecasting model outputs the winning bid and corresponding revenue for the virtual power plant. The winning bid includes the output plan that each unit's generator set needs to execute in real time.

[0046] In this embodiment, a mixed integer linear programming (MILP) solver is used to solve the above model to obtain an optimal benefit value, which serves as a benchmark value for subsequent decisions.

[0047] In this embodiment, a day-ahead transaction prediction model is established based on the information gap decision theory. The day-ahead transaction prediction model adopts the information gap decision theory (IGDT) to tolerate the uncertainty of the data to improve the market performance and operational efficiency of the virtual power plant. When processing uncertain information, the IGDT method can tolerate a certain degree of data noise and imprecision, and only needs to know the possible range or limit of the uncertain parameters. Even when data is scarce or the probability distribution of uncertain parameters is difficult to estimate, this method can still provide useful decision support, does not rely on detailed information of the data, and reduces the risk of model overfitting. Establishing a day-ahead transaction model based on the information gap decision theory is conducive to helping market participants better cope with the risks brought about by electricity price fluctuations.

[0048] Power generators must develop a bidding strategy based on their own generation capacity and market-defined bidding rules, aiming to maximize profits. This strategy must be submitted to the market operator in the form of a bid curve. This application will allow virtual power plants to participate in market bidding as price takers. Given their relatively small size and inability to influence market clearing prices, winning bids will be settled at the declared price. A day-ahead trading model will be established to maximize profits for virtual power plant aggregators.

[0049] In the day-ahead stage, each trading entity determines the startup combination plan and power generation capacity of the units for the next day through centralized bidding, basically determines the equipment status of the units, forms the day-ahead time-of-use electricity price and node electricity price, and at the same time adjusts the power generation and consumption curve and medium- and long-term contract deviations to ensure the balance of electricity and power for the next day.

[0050] In step S106, during the real-time scheduling phase of the next day, the real-time deviation value of the output of each unit is determined according to the winning bid result and the actual output of each unit.

[0051] In this embodiment, in real-time, the day-ahead output plan may deviate due to factors such as uncertainty in wind and photovoltaic output, load forecast deviations, and temporary unit failures. This requires adjusting the generation and consumption curve and day-ahead forecast deviations based on grid operating conditions and ultra-short-term load forecasts to achieve real-time power balance. The virtual power plant operator's bidding mechanism for different resource combinations is key to encouraging various demand-side resources to actively participate in virtual power plant transactions and improving the overall efficiency of the virtual power plant.

[0052] In step S108, the output real-time deviation value is input into a real-time deviation balance model, and the real-time deviation balance model outputs a minimum deviation adjustment total cost.

[0053] In this embodiment, a real-time deviation balance model is established in advance. Due to the random fluctuations in wind speed and solar irradiance, the actual output of wind and solar power generation may deviate significantly from the predicted value, which directly affects the accuracy of the declaration in the electricity market and the supply and demand balance of the power system. In order to effectively deal with this uncertainty and ensure the stable operation and optimal scheduling of the power system, the power generation plan can be dynamically adjusted according to the latest wind and solar output forecast and actual operation data in each scheduling cycle, thereby minimizing the impact of the uncertainty of wind and solar output and ensuring the reliable power supply and economy of the power system. Considering the uncertainty of wind and solar output, it is necessary to temporarily call on controllable units such as micro gas turbines, energy storage systems, electric vehicles, and air conditioners to maximize the realization of the day-ahead declaration plan by changing their operating status or electricity consumption behavior. When different controllable units change the day-ahead declaration plan, an adjustment cost will be incurred. In order to achieve the greatest possible consistency between real-time execution and the day-ahead plan and avoid high deviation penalty costs, this application chooses to minimize the system deviation adjustment cost as the objective function.

[0054] In step S110 , the adjustment plan and the minimum adjustment cost of each unit are determined according to the minimized deviation adjustment total cost.

[0055] This application proposes an improved two-stage trading optimization model that uses information gap decision theory (IGDT) to tolerate data uncertainty and improve the market performance and operational efficiency of virtual power plants. During the next day's real-time scheduling phase, a real-time deviation model is used to output the minimum deviation adjustment total cost. Based on this minimum deviation adjustment total cost, the adjustment plan and minimum adjustment cost for each unit are determined. This helps power plants reduce costs and increase profits.

[0056] In some embodiments, the day-ahead transaction prediction model is provided with an objective function, constraints, and a clearing electricity price uncertainty set.

[0057] The objective function is the revenue obtained by the virtual power plant aggregator in the electric energy market.

[0058] In this embodiment, the objective function is:

[0059]

[0060] Where G is the revenue obtained by the virtual power plant aggregator in the electricity market, Q t is the total electricity declared at time t, P t is the quotation at time t.

[0061] The clearing electricity price uncertainty set includes a set of actual values of the market clearing electricity price.

[0062] The actual value of the market-clearing electricity price is determined according to the actual value and fluctuation range of the market-clearing electricity price.

[0063] In this embodiment, the clearing electricity price uncertainty set is:

[0064]

[0065] Where, is the predicted value of the market clearing price, λ t is the actual value of the market clearing price, α P is the fluctuation range of the clearing price. First, ignoring the uncertainty of the clearing price, the predicted clearing price is substituted into the model solution. The result is used as the fixed value of the objective function, which is the reference value for expected profit in subsequent decision-making. The actual clearing price is uncertain and deviates from the predicted value. Robust optimization models and opportunistic optimization models are constructed to address the impact of clearing price fluctuations.

[0066] In some embodiments, the constraints include reporting quantity constraints, micro gas turbine unit operation constraints, wind and solar output constraints, energy storage system charging and discharging power constraints, state of charge constraints, output capacity constraints, electric vehicle cluster and air conditioning output capacity constraints and response number constraints.

[0067] In this embodiment, the reporting quantity constraints are:

[0068]

[0069] Where λ t is the market-clearing price at time t. If the bid price exceeds the clearing price, the winning bid is zero. If the bid price's lower limit is higher than the clearing price, it fails to meet costs, and the aggregator will not bid. Only when the bid price meets costs and is lower than the clearing price will the aggregator bid and gain market share and trading opportunities.

[0070] 2) The operating constraints of the micro gas turbine unit mainly include start-stop constraints, output constraints, and ramp constraints. Therefore, the operating constraints of the micro gas turbine are as follows:

[0071]

[0072] Where μ MT,t It is the operating status of the unit at time t, 0 represents shutdown and 1 represents startup. on,t-1 is the running time of the unit up to time t, T minon is the shortest start-up time of the unit, T off,t-1 is the downtime of the unit up to time t, T minofm is the shortest downtime of the unit. g MT,t 、g MT,min 、g MT,max They are the unit output at time t and the upper and lower limits of the unit output respectively. They are the upper and lower limits of the unit's climbing capacity.

[0073] 3) Wind and solar power output constraints:

[0074]

[0075] Where, They represent the maximum power generation capacity of photovoltaic and wind power at time t respectively.

[0076] 4) The main constraints considered for energy storage systems participating in virtual power plant operations include charge and discharge power constraints and state of charge constraints. Therefore, the energy storage system operation constraint model is as follows:

[0077]

[0078] Where, are the maximum charging and discharging power of the energy storage system at time t.

[0079] They are respectively the lower and upper limits of the state of charge required by the energy storage system.

[0080] 5) Electric vehicle clusters and air conditioners respond to virtual power plant dispatch by adjusting their operating modes to obtain economic benefits, such as incentives or preferential electricity prices. The main constraints to consider are output capacity constraints and response time constraints. Therefore, the operating constraints are modeled as follows:

[0081]

[0082] Where g EVC,t 、g AC,t are the equivalent outputs that the electric vehicle group and air conditioner can provide to the virtual power plant at time t, are the maximum equivalent output that the electric vehicle group and air conditioner can provide to the virtual power plant at time t, η max 、μ max are the maximum response times of the electric vehicle group and air conditioner in a complete scheduling cycle, respectively.

[0083] In some embodiments, in step S104, the day-ahead transaction prediction model includes: a robust sub-model and an opportunity sub-model.

[0084] The output power of each unit, as well as the clearing electricity price and cost, are input into a pre-set day-ahead transaction prediction model established based on information gap decision theory; the day-ahead transaction prediction model outputs the winning bid result and corresponding revenue of the virtual power plant, specifically including:

[0085] The output power of each unit, the clearing electricity price and the cost are input into a pre-set day-ahead transaction prediction model based on information gap decision theory to obtain the profit benchmark value G b .

[0086] The robustness deviation factor β is determined according to the objective function and the risk aversion parameter of the robust sub-model.

[0087] According to the robustness deviation factor β and the return benchmark value G b Determine the first profit threshold G exp ;

[0088] According to the first income threshold G exp Determine the maximum robustness.

[0089] In this embodiment, the robust model seeks to find a decision solution that satisfies the decision maker's minimum expectations even in the worst-case scenario. This emphasizes robustness under the worst-case scenario and is suitable for decision makers with low risk tolerance who strongly desire to avoid the worst-case outcome. The opportunity model considers the decision solution that maximizes the objective function value under the best-case scenario. This emphasizes optimization under the best-case scenario and is suitable for decision makers with high risk tolerance who seek the maximum possible benefit.

[0090] Set the robust deviation factor of the return in the robust optimization model to β, and the model solves the return not less than (1-β)G b The robust optimization model includes the above constraints and the following:

[0091]

[0092] Where a is the robustness factor. A larger value indicates a more robust decision solution to uncertain fluctuations. β is the robustness deviation factor, which measures the deviation between the decision maker's expected target value and the target's determined value. This robust model maximizes the robustness of the decision solution, i.e., maximizes robustness, while ensuring that the acceptable critical target is not exceeded. Decision variables are also obtained, ensuring that even if the uncertain parameters fluctuate to extremes, the results are acceptable.

[0093] The opportunity deviation factor δ is determined according to the objective function and the risk aversion parameter of the opportunity sub-model.

[0094] According to the opportunistic deviation factor δ and the return benchmark value G b Determine the second profit threshold G op .

[0095] According to the second profit threshold G op Determine the minimum chance.

[0096] In this embodiment, the opportunity deviation factor of the benefit in the opportunity sub-model is set to δ, and the benefit obtained by the opportunity sub-model is not less than (1+δ)G b The opportunity sub-model includes the following in addition to the above constraints:

[0097]

[0098] Where b is the degree of chance; the smaller its value, the greater the risk of obtaining a better outcome. δ is the chance bias factor. This chance model limits the fluctuation of uncertain information. The decision variables derived from this model are likely to achieve the decision maker's desired goal when the uncertain parameters fluctuate within a given range. When the fluctuation range of the uncertain parameters is too small, the risk is greater. Therefore, this model is suitable for risk-takers who seek to achieve higher returns through speculation.

[0099] The winning bid result and corresponding income of the virtual power plant are determined according to the maximum robustness and the minimum chance.

[0100] In some embodiments, the objective function of the real-time deviation balance model is the total deviation adjustment cost of the virtual power plant at any time.

[0101] The total deviation adjustment cost of the virtual power plant at any time includes: micro gas turbine adjustment cost, energy storage system adjustment cost, electric vehicle adjustment cost, air conditioning adjustment cost and grid deviation penalty fee.

[0102] In some embodiments, the adjustment cost of the micro gas turbine is determined based on the unit adjustment cost and the adjustment output of the micro gas turbine.

[0103] The energy storage system adjustment cost is determined according to the unit power adjustment cost and the adjustment power of the energy storage system.

[0104] The electric vehicle adjustment cost is determined according to the unit power adjustment cost and the adjustment power of the electric vehicle.

[0105] The air conditioner adjustment cost is determined according to the unit power adjustment cost and the adjustment power of the air conditioner.

[0106] The grid deviation penalty fee is determined according to the unit power adjustment cost and the adjustment power of the micro gas turbine.

[0107] In this embodiment, considering the uncertainty of wind and solar power output, it is necessary to temporarily call on controllable units such as micro gas turbines, energy storage systems, electric vehicles, and air conditioners to maximize the implementation of the day-ahead plan by changing their operating status or electricity consumption behavior. When different controllable units change the day-ahead plan, adjustment costs will be incurred. In order to ensure that real-time execution is as consistent as possible with the day-ahead plan and avoid high deviation penalty costs, this application chooses to minimize the system deviation adjustment cost as the objective function:

[0108] Objective function:

[0109]

[0110]

[0111] Where C t is the total cost of deviation adjustment of the virtual power plant at time t;

[0112] ΔC MT,t , ΔC ESS,t , ΔC EVC,t , ΔC AC,t , ΔC grid,tThey are the adjustment costs of micro gas turbines, energy storage systems, electric vehicles, air conditioners, and grid deviation penalty fees;

[0113] are the unit adjustment costs of the above units respectively;

[0114] Δg MT,t 、 Δg EVC,t , Δg AC,t , Δg grid,t The output of each of the above units is adjusted respectively.

[0115] In some embodiments, the constraints of the real-time deviation balance model include: supply and demand balance constraints, controllable unit operation constraints, energy storage system scheduling constraints, and flexible load constraints.

[0116] Supply and demand balance constraints:

[0117] Δg pv,t +Δg wpp,t =Δg MT,t +Δg ESS,t +Δg EVC,t +Δg AC,t +Δg grid,t ;

[0118] Where Δg pv,t , Δg wpp,t The deviation of the declared output of PV and WPP, Δg MT,t , Δg ESS,t , Δg EVC,t , Δg AC,t , Δg grid,t They represent the adjusted output and reported shortage output of each unit at time t respectively.

[0119] Controllable unit operation constraints: In the real-time stage, the micro-turbine output adjustment is limited by the upper and lower limits of the unit output and the upper and lower limits of the ramp power at time t. The specific constraints are as follows:

[0120]

[0121] Where g MT,t Planned output for the unit at time t.

[0122] Energy storage system scheduling constraints: During the real-time phase, the energy storage system is still subject to state of charge and power constraints. Furthermore, the charge and discharge adjustments of the energy storage system are limited to the maximum adjustable value during the period, and the battery's operating status must be consistent with the day-ahead plan. Specific constraints are as follows:

[0123]

[0124] Where, is the maximum floating value of the energy storage system, It is a real-time charging and discharging 0-1 state variable.

[0125] Flexible load constraints: Electric vehicles and air conditioners are flexible loads that can be adjusted. Their real-time adjustment capabilities are limited and the adjustment range cannot be too large. The specific constraints are as follows:

[0126]

[0127] Where η EVC,t 、μ AC,t It is a 0-1 state variable, 0 means that the output of electric vehicles and air conditioners cannot be adjusted in real time, and 1 means that the output of electric vehicles and air conditioners can be adjusted in real time. max,t 、g min,t Adjust the upper and lower limits for output.

[0128] In some embodiments, in the spot market bidding, the bidding strategy of the virtual power plant is crucial to its revenue and market competitiveness. The bidding range analysis is the basis for the virtual power plant to formulate a reasonable bidding strategy, which directly affects its revenue and risk in the market. In view of the fact that the existing research on virtual power plants focuses on scheduling optimization and handling of uncertainty in the output of renewable energy such as wind power, and lacks in-depth analysis and discussion of the bidding range, this application will analyze the bidding range of virtual power plants participating in the spot market bidding, and construct a model for calculating the upper and lower limits of the bidding range electricity price, where the upper limit of the interval is the clearing electricity price and the lower limit is the total cost calculation model.

[0129] In some embodiments, predicting the clearing electricity price includes: using an AutoRegressive Integrated Moving Average (ARIMA) model to predict the clearing electricity price.

[0130] In this embodiment, the upper limit of the electricity price within the quotation range is calculated using a model. my country's spot market clearing rules need to be improved, and there is insufficient disclosed transaction data. The ARIMA model does not require extensive historical data and can be effective even with relatively small amounts of data. Therefore, the present invention uses the ARIMA model for predictive modeling.

[0131] The basic concept of the ARIMA model is to generalize the ARMA model to deal with non-stationary time series. It achieves stabilization by performing d-order differences. After the data is stable, the relationship between the current value of the predicted object and the past value and past prediction error is analyzed to establish a prediction model.

[0132] In some embodiments, establishing the ARIMA model comprises the following steps:

[0133] 1) Original time series trend judgment and stabilization processing.

[0134] Collect the disclosed electricity market clearing price data for the region to be forecasted, check data availability, address missing values and outliers, extract temporal features, and convert the data into a usable time series. Use a time series plot to determine whether the series is stationary. Given the cyclical and seasonal nature of electricity prices, stabilization is generally necessary. Use the differencing method to generate a time series plot. Once the differencing sequence is confirmed to be stationary, proceed to the next step.

[0135] 2) Model construction and preliminary determination of parameters.

[0136] The p-value and q-value in ARIMA(p,d,q) are determined by the ACF and PACF plots of the differenced series.

[0137] The ACF, or autocorrelation function, measures the correlation between the current value in a time series and the value lagged q time points. The result is typically presented as a correlation coefficient, which ranges between -1 and 1. Values close to 1 or -1 indicate a strong positive or negative correlation; values close to 0 indicate no significant correlation.

[0138] The PACF, or partial autocorrelation function, measures the correlation between the current value in a time series and the value lagged p time points after controlling for the effects of intermediate lags. It is typically obtained by solving an autoregressive model and can be expressed as a correlation coefficient, interpreted in the same way as the ACF.

[0139] The ACF and PACF plots can be used to identify the order p of the AR component and the order q of the MA component. If the ACF plot shows truncation and the PACF plot shows tailing, an MA model is obtained. If the ACF plot shows tailing and the PACF plot shows truncation, an AR model is obtained. By observing the order after which the ACF and PACF plots show tailing, combined with the d-order variance obtained in the previous step, all the parameters of the ARIMA (p, d, q) can be determined.

[0140] 3) Model parameter estimation and model fitting.

[0141] According to the possible values of all parameters obtained in the first two steps, repeated fitting is performed, the model prediction fitting graph is drawn, the prediction effects of different parameter combinations are compared, and the optimal model parameters are obtained.

[0142] 4) Model verification.

[0143] After establishing an ARIMA (p, d, q) model, you need to perform diagnostic tests to check whether the residuals are white noise, that is, whether the residuals are random. Generally speaking, the variances at different time points must be independent of each other and free of autocorrelation. Furthermore, you want the model residuals to conform to a normal distribution with a mean of 0 and a constant variance. If the residuals are not white noise, you need to readjust the model parameters and perform a model fit evaluation until the residuals are white noise.

[0144] Observing the ACF and PACF plots of the residuals can be used to test for autocorrelation in the residual series, that is, whether the correlation coefficient falls within the confidence interval. The QQ plot can be used to test whether the residuals conform to a normal distribution. The QQ plot, also known as the quantile-quantile plot, is a graphical method used to compare the quantiles of two distributions, providing a visual indication of whether the data conform to a particular distribution.

[0145] 5) Predictive applications.

[0146] After the model passes the white noise test, there is no need to continue modeling. The most recent historical observed electricity price data can be directly input to generate the predicted clearing electricity price.

[0147] A model for calculating the lower limit of the electricity price within the bidding range. Virtual power plants aggregate internal resources and act as independent market entities, participating in the electricity market. Aggregators need to understand the generation and utilization capabilities of their internal components and comprehensively consider their bidding strategies, prioritizing maximum revenue. Specifically, the bid price must meet the cost of purchasing electricity from internal resources.

[0148]

[0149] Where, P MIN,t is the lower limit of the unit quotation at time t, is the total cost of internal power purchases by the virtual power plant aggregator at time t, η is the expected profit margin, U = MT, PV, WPP, ESS, EVC, AC is the set of units aggregated by the virtual power plant, It is the total amount of electricity that the virtual power plant aggregator can claim at time t. is the output that the micro gas turbine in the VPP can provide at time t; is the output that the PV units in the VPP can provide at time t; is the output that the wind turbines in the VPP can provide at time t; is the maximum equivalent output that the electric vehicle cluster in the VPP can provide at time t; is the regulated output that the energy storage device in the VPP can provide at time t; is the maximum equivalent output that can be adjusted by the air conditioner in the VPP at time t. A 0-1 variable indicating whether the aggregation unit participates in the declaration, 0 means not participating in the declaration at time t, and 1 means participating in the declaration at time t. U The unit cost of purchasing electricity from each unit for the virtual power plant.

[0150] In some embodiments, determining the adjustment plan and the lowest adjustment cost of each unit according to minimizing the total deviation adjustment cost includes:

[0151] A pre-trained machine learning model is used to determine the adjustment plan and minimum adjustment cost of each unit based on the total cost of minimizing the deviation adjustment.

[0152] In this embodiment, the output plans of units such as MT, ESS, EVC, and AC can be adjusted based on the output capacity model of each flexibility unit, with the goal of minimizing the total cost of deviation adjustment. The deviation correction is achieved by comprehensively considering whether the actual output of each unit can meet the total power quantity bid on the previous day. After the flexibility unit has exerted its maximum adjustment capacity, if it still cannot meet the deviation adjustment needs, the virtual power plant needs to bear the deviation penalty fee, which is included in the deviation adjustment cost in the present invention. Use a MILP solver (such as Gurobi or CPLEX) to solve and obtain an adjustment plan that minimizes the total cost of deviation adjustment.

[0153] When using a pre-trained machine learning model to determine the adjustment plan and minimum adjustment cost of each unit based on the minimized deviation adjustment total cost, it is necessary to collect a large amount of historical data and learn the relationship between the adjustment cost of each unit and the total adjustment cost.

[0154] In some embodiments, the machine learning model includes: a linear regression model, a decision tree model, a random forest model, a neural network model, or a support vector machine model.

[0155] Among them, the linear regression model can establish a linear relationship between the total adjustment cost and each unit, learn the appropriate weights through training data, and distribute the total adjustment cost to each unit according to a certain proportion.

[0156] Decision tree model. It can allocate the total adjustment cost to different units based on different characteristic conditions, and determine the allocation of each unit through the decision rules of the tree structure.

[0157] Random forest model. This model consists of multiple decision trees and combines their results to adjust the total cost distribution. It generally has better generalization ability and stability.

[0158] Neural network models, such as the multilayer perceptron (MLP), can learn complex nonlinear relationships and are suitable for adjusting complex mapping relationships between total cost allocation and individual units.

[0159] Support vector machine model. It can find the optimal hyperplane in feature space to partition data. It can be used to reasonably allocate the total adjustment cost to different units. It performs particularly well when dealing with allocation problems with nonlinear boundaries.

[0160] In practical applications, it is necessary to select a suitable machine learning model based on the specific problem characteristics, data features, and assigned goals and requirements, and perform corresponding model training and optimization.

[0161] The following uses the linear regression model as an example to illustrate how to adjust the relationship between the total cost and each unit for allocation:

[0162] 1. Data Collection and Preparation: The virtual power plant collects historical operating data for the steam turbine unit, energy storage system unit, electric vehicle unit, air conditioning unit, and grid unit. This data includes each unit's power output, energy consumption, cost-related data (such as steam turbine fuel costs and energy storage charging and discharging costs), as well as the total cost of the entire virtual power plant. The data is cleaned, missing values and outliers are processed, and then divided into training and test sets.

[0163] 2. Linear Regression Model Training: A linear regression model is trained using the training data, using relevant data for each unit (such as power and energy consumption) as the independent variables and total cost as the dependent variable. The model learns the linear relationship between each unit and total cost, determining the coefficients (weights) for each independent variable. For example, the model might determine that for every increase in steam turbine unit power, total cost increases by a certain amount.

[0164] 3. Deviation Calculation and Analysis: Use the trained model to predict the test set data and obtain the predicted total cost. Compare the predicted total cost with the actual total cost and calculate the deviation. Analyze the causes of the deviation and determine which unit changes have the greatest impact on the deviation. For example, if the model predicts a larger deviation when EV units are charging in clusters, this indicates that the relationship between this unit and the total cost may need further refinement.

[0165] 4. Adjustment Plan Development: Based on the deviation analysis results and the operational constraints of each unit (such as turbine power limits and energy storage capacity range), an adjustment plan is developed. If the model indicates that increasing the discharge capacity of the energy storage system unit can reduce the forecast deviation and total cost, and is within the allowable operating range, then increasing the discharge capacity of the energy storage system can be considered.

[0166] 5. Cost Calculation and Optimization: Based on the established adjustment plan, calculate the total cost after adjustment. By repeatedly trying different adjustment schemes and using a linear regression model to predict the total cost, find the adjustment plan that minimizes the total cost, i.e., the one with the lowest adjustment cost. For example, try different air conditioning unit power adjustment ranges and combine the model's total cost prediction to find the adjustment value that meets comfort requirements while minimizing the total cost.

[0167] 6. Implementation and Feedback: Adjust each unit according to the determined adjustment plan and monitor the virtual power plant's operation and actual costs in real time. Feed this new data back to the linear regression model to update and optimize it, allowing for more accurate adjustment plans next time.

[0168] Through the above steps, the virtual power plant can use the linear regression model to adjust the total cost based on minimizing the deviation, optimize the adjustment plan of each unit and determine the lowest adjustment cost.

[0169] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, rather than to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions of the technical solution of the present invention by ordinary technicians in this field do not deviate from the essence and scope of the technical solution of the present invention.

Claims

1. A method for processing output and cost data of a virtual power plant taking into account air conditioning load, characterized in that: include: In the day-ahead phase, the output power of each unit in the virtual power plant for the next day, as well as the predicted clearing electricity price and cost, are determined. The units in the virtual power plant include: a micro gas turbine unit, an energy storage system unit, an electric vehicle unit, an air conditioning unit, and a power grid unit. The output power of each unit, as well as the clearing electricity price and cost, are input into a pre-set day-ahead transaction prediction model established based on information gap decision theory; the day-ahead transaction prediction model outputs a winning bid result and corresponding revenue for the virtual power plant; the winning bid result includes: an output plan that the generator sets of each unit need to execute in real time; During the real-time scheduling phase of the next day, the real-time output deviation value of each unit is determined based on the winning bid result and the actual output of each unit; Inputting the output real-time deviation value into a real-time deviation balance model; the real-time deviation balance model outputs a minimized deviation adjustment total cost; The adjustment plan and the minimum adjustment cost of each unit are determined according to the minimized deviation adjustment total cost.

2. The method for processing output and cost data of a virtual power plant taking into account air conditioning load according to claim 1, characterized in that: In the day-ahead transaction prediction model, an objective function, constraints and a clearing electricity price uncertainty set are set; The objective function is the revenue obtained by the virtual power plant aggregator in the electric energy market; The clearing electricity price uncertainty set includes a set of actual values of the market clearing electricity price.

3. The method for processing output and cost data of a virtual power plant taking into account air conditioning load according to claim 2, characterized in that: The constraints include reporting quantity constraints, micro gas turbine unit operation constraints, wind and solar output constraints, energy storage system charging and discharging power constraints, charge state constraints, output capacity constraints, electric vehicle cluster and air conditioning output capacity constraints and response number constraints.

4. The method for processing output and cost data of a virtual power plant taking into account air conditioning load according to claim 1, characterized in that: The day-ahead transaction prediction model includes: a robust sub-model and an opportunity sub-model; The output power of each unit, as well as the clearing electricity price and cost, are input into a pre-set day-ahead transaction prediction model established based on information gap decision theory; the day-ahead transaction prediction model outputs the winning bid result and corresponding revenue of the virtual power plant, specifically including: Inputting the output power of each unit, as well as the clearing electricity price and cost into a pre-set day-ahead transaction prediction model established based on information gap decision theory to obtain a benchmark value of revenue; Determining the robustness deviation factor according to the objective function and the risk aversion parameter of the robust sub-model; Determining a first profit threshold according to the robustness deviation factor and the profit benchmark value; determining a maximum robustness according to the first benefit threshold; Determining the opportunity deviation factor according to the objective function and risk aversion parameter of the opportunity sub-model; Determining a second profit threshold value according to the opportunistic deviation factor and the profit benchmark value; determining a minimum chance degree according to the second profit threshold; The winning bid result and corresponding income of the virtual power plant are determined according to the maximum robustness and the minimum chance.

5. The method for processing output and cost data of a virtual power plant taking into account air conditioning load according to claim 1, characterized in that: The objective function of the real-time deviation balance model is the total cost of deviation adjustment of the virtual power plant at any time; The total deviation adjustment cost of the virtual power plant at any time includes: micro gas turbine adjustment cost, energy storage system adjustment cost, electric vehicle adjustment cost, air conditioning adjustment cost and grid deviation penalty fee.

6. The method for processing output and cost data of a virtual power plant taking into account air conditioning load according to claim 5, characterized in that: determining an adjustment cost of the micro gas turbine according to a unit adjustment cost and an adjustment output of the micro gas turbine; determining the energy storage system adjustment cost according to the unit power adjustment cost and the adjustment power of the energy storage system; determining the electric vehicle adjustment cost according to the unit power adjustment cost and the adjustment power of the electric vehicle; determining the air conditioner adjustment cost according to the unit power adjustment cost and the adjustment power of the air conditioner; The grid deviation penalty fee is determined according to the unit power adjustment cost and the adjustment power of the micro gas turbine.

7. The method for processing output and cost data of a virtual power plant taking into account air conditioning load according to claim 5, characterized in that: The constraints of the real-time deviation balance model include: supply and demand balance constraints, controllable unit operation constraints, energy storage system scheduling constraints and flexible load constraints.

8. The method for processing output and cost data of a virtual power plant taking into account air conditioning load according to claim 1, characterized in that: Predicting the clearing electricity price, including: using an autoregressive integrated moving average model to predict the clearing electricity price.

9. The method for processing output and cost data of a virtual power plant taking into account air conditioning load according to claim 1, characterized in that: Determining the adjustment plan and minimum adjustment cost of each unit according to the minimized deviation adjustment total cost includes: A pre-trained machine learning model is used to determine the adjustment plan and minimum adjustment cost of each unit based on the total cost of minimizing the deviation adjustment.

10. The method for processing output and cost data of a virtual power plant taking into account air conditioning load according to claim 1, characterized in that: Machine learning models include: Linear regression model, decision tree model, random forest model, neural network model, or support vector machine model.

Images