Day-ahead strategy optimization method and system for charging operators based on future expected cost

Abstract

The application provides a charging operator day-ahead strategy optimization method and system based on future expected cost, comprising the following steps: generating a charging operator day-ahead decision-making scene based on market information rolling prediction; establishing a target function and constraint condition of a day-ahead decision-making model considering future expected cost based on the day-ahead decision-making scene; adopting a future expected cost function reverse construction method to segment linearize a daily decision-making optimization problem of the day-ahead decision-making model; and solving the linearized optimization problem to obtain a charging operator day-ahead strategy. The application realizes a charging operator day-ahead strategy optimization technology based on future expected cost through market information rolling prediction, maximally utilizes constantly updated market information, and reasonably predicts a future electricity price scene; and the influence of maximum demand on the total cost of the charging operator in terms of demand electricity charge and subsequent optimization boundary is considered, and the influence of the maximum demand on the whole cycle cost is accurately modeled.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of charging operator day-ahead strategy optimization, in particular to a charging operator day-ahead strategy optimization method and system based on future expected cost. BACKGROUND

[0002] The day-ahead decision of the charging operator includes the decision results of the two aspects of the declared power of the day-ahead market and the daily maximum load limit. According to the decision results, the declared power and demand pre-control of the charging operator in the day-ahead operation stage can be supported, which has important significance in balancing the load, stabilizing the voltage frequency, improving the equipment utilization rate, and reducing the operation risk of the distribution network.

[0003] The existing day-ahead decision technology of the charging operator usually adopts the following method: considering the relationship between the energy cost and the charging income and the declared power, establishing a profit model of the charging operator, and constructing an optimization model to obtain the declared power in the day-ahead power market. This method fails to accurately consider the demand charge in the power cost, and has the following technical problems: it cannot guide the charging operator to make a reasonable demand control plan from the perspective of day-ahead decision, which may cause short-time power impact in the process of power consumption of the charging operator, thereby causing negative effects on balancing the load, stabilizing the voltage frequency, improving the equipment utilization rate, and reducing the operation risk of the distribution network. SUMMARY

[0004] The present application provides a charging operator day-ahead strategy optimization method and system based on future expected cost to solve the above problems in the prior art.

[0005] According to one aspect of the present application, a charging operator day-ahead strategy optimization method based on future expected cost is provided, comprising:

[0006] Based on the market information rolling prediction, a charging operator day-ahead decision scenario is generated;

[0007] Based on the day-ahead decision scenario, a day-ahead decision objective function and constraint condition considering future expected cost are established, a future electricity cost expected function containing demand charge is constructed, and a total energy cost containing deviation charge is obtained;

[0008] linearize the future electricity cost expectation function and the deviation electricity fee to obtain a linearized optimization problem; wherein, for the deviation electricity fee, two groups of continuous variables are introduced to respectively represent the part of the real-time power exceeding the set multiple of the day-ahead declared power and the part of the real-time power lower than the set multiple of the day-ahead declared power at the t period of the d day under the scenario s, and linearization conversion is performed; for the future electricity cost expectation function, the value range of the expectation function independent variable is divided into segments by segment points, and continuous variables and binary variables are introduced to convert the maximum demand into a linear combination of each segment point;

[0009] the linearized optimization problem is solved to obtain the charging operator day-ahead strategy.

[0010] According to another aspect of the present application, a charging operator day-ahead strategy optimization system based on future expected cost is provided, comprising:

[0011] a decision scenario generation module which generates a charging operator day-ahead decision scenario based on market information rolling prediction;

[0012] a day-ahead decision model construction module which establishes a day-ahead decision objective function and constraint condition considering future expected cost based on the day-ahead decision scenario, constructs a future electricity cost expectation function containing demand electricity fee, and obtains total electricity energy cost containing deviation electricity fee;

[0013] an optimization problem linearization module which is used to linearize the future electricity cost expectation function and the deviation electricity fee to obtain a linearized optimization problem; wherein, for the deviation electricity fee, two groups of continuous variables are introduced to respectively represent the part of the real-time power exceeding the set multiple of the day-ahead declared power and the part of the real-time power lower than the set multiple of the day-ahead declared power at the t period of the d day under the scenario s, and linearization conversion is performed; for the future electricity cost expectation function, the value range of the expectation function independent variable is divided into segments by segment points, and continuous variables and binary variables are introduced to convert the maximum demand into a linear combination of each segment point;

[0014] a day-ahead strategy optimization module which is used to solve the linearized optimization problem to obtain the charging operator day-ahead strategy.

[0015] Thanks to the above technical solutions, the present application has at least one of the following beneficial effects compared with the prior art:

[0016] The application provides a charging operator day-ahead strategy optimization method and system based on future expected cost.

[0017] The charging operator day-ahead strategy optimization method and system based on future expected cost provided by the application embeds the demand charge settled monthly into the day-ahead decision model, guides the charging operator to make a reasonable demand control plan from the perspective of day-ahead decision, prevents the charging operator from having a short-time power impact during power consumption, and ensures the safe operation of the power distribution network.

[0018] The charging operator day-ahead strategy optimization method and system based on future expected cost provided by the application constructs a future power consumption cost expectation function containing the demand charge, ensures the rationality of the day-ahead declared power through the demand charge, and provides a user side power consumption management basis for the charging operator, thereby reducing the maximum demand of the charging operator.

[0019] The charging operator day-ahead strategy optimization method and system based on future expected cost provided by the application guides the charging operator to reasonably plan the use of charging service capacity, and improves the use efficiency of charging service resources. BRIEF DESCRIPTION OF DRAWINGS

[0020] Other features, objects and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments with reference to the attached drawings:

[0021] Figure 1 The figure is a working schematic diagram of the charging operator day-ahead strategy optimization method based on future expected cost in a preferred embodiment of the application.

[0022] Figure 2 The figure is a component module schematic diagram of the charging operator day-ahead strategy optimization system based on future expected cost in a preferred embodiment of the application.

[0023] Figure 3 The figure is a working flowchart of the charging operator day-ahead strategy optimization method based on future expected cost in a specific application example of the application.

[0024] Figure 4 The figure is a regional charging station distribution in a specific application example of the application.

[0025] Figure 5 The figure is the full-month real load and the full-month predicted load in the day-ahead decision stage of the first day in a specific application example of the application.

[0026] Figure 6 The figure is the full-month real day-ahead price and the predicted day-ahead price in a specific application example of the application.

[0027] Figure 7 The figure shows the distribution of the percentage of the deviation between the real-time electricity price and the day-ahead electricity price in the historical data in a specific application example of the present application.

[0028] Figure 8 The figure shows the charging load curve of the charging load scenario in a specific application example of the present application.

[0029] Figure 9 The figure shows the future expected cost function of some dates obtained in the day-ahead decision stage on the first day in a specific application example of the present application; wherein (a)~(d) respectively correspond to the future expected cost function values on the fourth day, the eleventh day, the twentieth day and the twenty-sixth day.

[0030] Figure 10 The figure shows the full-month day-ahead declared power optimization result (taking the previous three days as an example) in a specific application example of the present application. DETAILED DESCRIPTION

[0031] The embodiments of the present application are described in detail as follows: The embodiments are implemented on the premise of the technical solutions of the present application, and detailed implementation manners and specific operation processes are given. It should be noted that, for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the protection scope of the present application.

[0032] In view of the problems in the prior art, an embodiment of the present application provides a charging operator day-ahead strategy optimization method based on future expected cost. The method considers the joint optimization of day-ahead electricity quantity declaration and maximum demand control target of demand charge. First, a day-ahead decision scenario generation framework based on market information rolling prediction is developed as the basis for day-ahead decision. Second, a day-ahead decision model considering future expected cost is established according to the Bellman optimality principle. Then, the future expected cost function reverse construction method is used to linearize the future cost including demand charge in the daily decision optimization problem, so as to convert the day-ahead decision problem into a MILP problem that can be solved efficiently.

[0033] Specifically, as shown in the figure, the charging operator day-ahead strategy optimization method based on future expected cost provided by the embodiment can include: Figure 1

[0034] S1, generating a charging operator day-ahead decision scenario based on market information rolling prediction;

[0035] ​S2, based on the day-ahead decision scenario, a day-ahead decision objective function and constraint conditions considering future expected cost are established, a future electricity cost expectation function including demand charge is constructed, and total electricity energy cost cost including deviation charge is obtained; wherein the demand charge is used to ensure the rationality of the day-ahead declared power and to provide user-side electricity management basis for the charging operator, so as to reduce the maximum demand of the charging operator;

[0036] S3, the future electricity cost expectation function and the deviation charge are linearized to obtain a linearized optimization problem; wherein for the deviation charge, two groups of continuous variables are introduced, which respectively represent the part of the real-time power exceeding the set multiple of the day-ahead declared power and the part of the real-time power lower than the set multiple of the day-ahead declared power in the t period of the d day under the scenario s, and linearization conversion is performed; for the future electricity cost expectation function, the value range of the maximum demand generated in the current month by the end of the current day is divided into segments by segmentation points, and continuous variables and binary variables are introduced to convert the maximum demand into a linear combination of each segmentation point;

[0037] S4, the linearized optimization problem is solved to obtain the day-ahead strategy of the charging operator.

[0038] In some preferred embodiments, the above S1, based on market information rolling prediction, generates the day-ahead decision scenario of the charging operator, and can further include:

[0039] S11, real-time market information is obtained, including: historical day-ahead / real-time electricity price, market total load, weather factors (including temperature, light intensity, etc.);

[0040] S12, based on the market information, the hourly load from the d day to the D day is predicted; wherein the D day is the last day of the current month;

[0041] S13, based on the hourly load, the day-ahead electricity price prediction mean and variance of each hour from the d day to the D day are calculated;

[0042] S14, based on the day-ahead electricity price prediction mean and variance, day-ahead electricity price scenarios are generated by sampling, wherein each day-ahead electricity price scenario includes the day-ahead electricity price of each hour from the d day to the D day; for each day-ahead electricity price scenario, real-time electricity prices are randomly sampled to form day-ahead-real-time electricity price scenarios;

[0043] S15, according to the charging demand, vehicle charging behavior simulation is performed to calculate the hourly charging load of each station under the charging operator, and a charging load scenario is generated;

[0044] S16, clustering the above obtained scenarios to obtain a plurality of day-ahead decision scenarios, wherein each day-ahead decision scenario comprises day-ahead electricity price, real-time electricity price and charging load of each hour from the dth day to the Dth day.

[0045] In some preferred embodiments, S12 can further comprise:

[0046] S121, initialization , wherein, is the prediction time step of the rolling prediction, is the first time step in the rolling prediction interval, i.e., the interval from the dth day to the Dth day.

[0047] S122, obtaining historical load and predicted load through a market information release channel;

[0048] S123, splicing the historical load and the predicted load to obtain load;

[0049] S124, obtaining temperature and illumination information through weather forecast or historical weather database;

[0050] S125, obtaining predicted load by using a load prediction model; wherein the load prediction model can adopt a pre-trained LSTM-based load prediction model; wherein the historical load of and the predicted load of , the temperature and illumination information of are taken as inputs of the model, and the required predicted load is outputted;

[0051] S126, judging ? (i.e., judging whether is equal to ); if yes, obtaining load, i.e., the hourly load from the dth day to the Dth day; if no, making , updating the load feature, and returning to S14 to start execution again until obtaining the hourly load from the dth day to the Dth day.

[0052] In some preferred embodiments, S13 can further comprise:

[0053] S131, initialization , wherein, is the prediction time step of the rolling prediction, is the first time step in the rolling prediction interval, i.e., the interval from the dth day to the Dth day.

[0054] S132, obtaining the historical day-ahead price through a market information release channel;

[0055] S133, obtaining the load of

[0056] S134, obtaining the temperature and illumination information of

[0057] S135, obtaining the day-ahead price mean and variance of by using a day-ahead price interval prediction model based on Gaussian process regression; wherein:

[0058] Gaussian process regression regards the function to be modeled as a Gaussian process, and performs Bayesian inference in the function space. Assuming that the target function to be modeled is , the function is said to obey Gaussian process. It can be seen from the principle of Gaussian process regression that it can not only give the mean value of the prediction, but also describe the uncertainty of each prediction point through the variance. This makes Gaussian process regression particularly suitable for tasks that need to describe uncertainty, such as the day-ahead optimization scenario generation required by the present application. The historical day-ahead price of , the load of , the temperature and illumination information of are used as input data of the model, and based on the distribution of the future time point day-ahead price obtained by Gaussian process regression, the required scenario can be conveniently generated by Monte Carlo sampling. Therefore, Gaussian process regression model is selected for day-ahead price prediction in this step, which serves as the basis for scenario generation.

[0059] In Gaussian process regression, the kernel function (or covariance function) is used to calculate the correlation between any two points of the input. Different kernel functions may affect the ability to describe the function relationship, so in the training of the Gaussian process regression model, a suitable kernel function needs to be selected, which can include: square exponential kernel, which has the highest smoothness, and the corresponding function is infinitely differentiable, suitable for modeling extremely smooth target functions. Matern 3 / 2 kernel, which generates a function that is once differentiable, is rougher than the square exponential kernel, and is more suitable for modeling functions with certain discontinuities or inflection points. Matern 5 / 2 kernel, which generates a function that is once differentiable and twice differentiable, has a smoothness between Matern 3 / 2 kernel and square exponential kernel, providing a good trade-off between smoothness and flexibility. Rational quadratic kernel, which can be regarded as a weighted average of a group of square exponential kernels, is suitable for modeling multi-scale phenomena. Exponential kernel, which is rougher than Matern 3 / 2 kernel, corresponds to a function that is only continuous but not differentiable.

[0060] ​​​It is generally believed that the day-ahead price is greatly influenced by the market supply-demand ratio, and the change trend of the historical day-ahead price can reflect the possible change trend of the day-ahead price to some extent. Therefore, two groups of data are selected as input features, which are: historical data, including the day-ahead price, regional load, temperature and light intensity in the past 168 hours; market and environmental factors: regional load forecast value, temperature and light intensity in the future 24 hours. Among them, the future regional load forecast value is given by market forecast value or market load prediction method. The output feature is the day-ahead price (including mean and variance) of the future 1 hour. Unlike the load prediction method, the prediction of the day-ahead price is hour by hour, because the input features available for the day-ahead price can be updated hour by hour, while in the input features of the load prediction, the market forecast load is updated every 24 hours.

[0061] S136, judging whether is equal to , if yes, the day-ahead price mean and variance of , i.e. the day-ahead price prediction mean and variance of each hour from the dth day to the Dth day are obtained; if not, make , update the day-ahead price feature by mean, return to S133 to start execution again, until , the day-ahead price prediction mean and variance of each hour from the dth day to the Dth day are obtained.

[0062] In some preferred embodiments, the above S14 can further include:

[0063] S141, initialization , wherein is the day-ahead price scenario number;

[0064] S142, initialization , wherein is the prediction time step of the rolling prediction, is the first time step in the rolling prediction interval, i.e. the interval from the dth day to the Dth day;

[0065] S143, randomly generating the day-ahead price , i.e. the day-ahead price of each hour from the dth day to the Dth day, according to the day-ahead price mean and variance at

[0066] S144, initialization , wherein is the real-time price scenario number corresponding to each day-ahead price scenario;

[0067] S145, generating the deviation and the real-time price ​wherein, is the hourly day-ahead price from day d to day D; wherein:

[0068] Unlike the day-ahead price which is cleared by market bidding, the real-time price is calculated based on the ultra-short term load forecast. Although the real-time price is anchored by the day-ahead price, it is still affected by many random factors, thus forming a certain random deviation with the day-ahead price. In the generation of decision scenarios, the generation of real-time price needs to consider market dynamics and statistical reliability, so this step adopts the method of constructing real-time price scenarios based on historical deviation distribution. The core of this method is to capture the conditional dependence between the day-ahead price and the real-time price through data-driven probabilistic modeling.

[0069] The deviation between the real-time price and the day-ahead price is driven by the uncertainty in the market clearing process, including renewable energy output fluctuation, load forecast deviation, network congestion leading to node price difference, etc. The real-time price can be expressed as:

[0070]

[0071] wherein, is used to represent the deviation between the real-time price and the day-ahead price, is the historical deviation distribution, which can be regarded as the joint effect of these uncertain factors. Based on the functional central limit theorem, when the generating factors of the day-ahead price deviation satisfy the conditions of weak correlation and stationarity, the standardized process of historical deviation converges to a diffusion process. Therefore, by estimating the steady-state distribution of historical deviation using non-parametric or parametric methods, the generated real-time price scenarios based on have statistical consistency.

[0072] In order to preserve the original characteristics of as much as possible, a non-parametric method is used for empirical distribution sampling. The empirical distribution function is defined as:

[0073]

[0074] wherein, is the indicator function. This function is a right-continuous, monotonically increasing step function that jumps at each observation point . For each day-ahead price sample , a deviation is uniformly sampled according to to generate a real-time price sample:

[0075]

[0076] The advantages of generating real-time electricity prices according to the above method are that the deviation completely depends on data, tail risk underestimation caused by false assumption of parameter distribution (such as normality) is avoided, and the phenomena of spikes, thick tails and multimodality in historical data can be automatically captured.

[0077] S146, judging ? (i.e., judging whether it is equal to ), if yes, proceed to the next step; if no, make , return to S145 to start execution again;

[0078] S147, judging ? (i.e., judging whether it is equal to ), if yes, proceed to the next step; if no, make , return to S143 to start execution again;

[0079] S148, judging ? (i.e., judging whether it is equal to ), if yes, get hourly day-ahead electricity prices and real-time electricity prices; if no, make , return to S142 to start execution again.

[0080] In some preferred embodiments, the above S15 can further include:

[0081] S151, initializing , wherein is the number of charging load scenarios;

[0082] S152, simulating vehicle charging behavior according to a charging demand spatiotemporal distribution prediction method to generate hourly charging load scenarios of each station under the charging operator ; the charging demand spatiotemporal prediction method used can be any existing method, and each charging load scenario contains hourly charging load from the dth day to the Dth day;

[0083] S153, judging ? (i.e., judging whether it is equal to ), if yes, get hourly charging load scenarios; if no, make , return to the previous step to start execution again.

[0084] In some preferred embodiments, the above 16 can further include: ​​

[0085] S161, combine the obtained day-ahead electricity price scenario, day-ahead-real-time electricity price scenario, and charging load scenario to obtain One scenario;

[0086] S162, for K-means clustering is performed on each scenario;

[0087] S163, using clustering to... The scenario was reduced to Each scenario yields a day-ahead decision scenario, which is used to construct a stochastic optimization problem for day-ahead decisions.

[0088] In some preferred embodiments, S2 above establishes the objective function of the day-ahead decision-making model that takes into account future expected costs, including:

[0089] S21, Based on the day-ahead decision-making scenario, establish a day-ahead decision-making objective function that takes into account future expected costs. The objective function aims to minimize the expected future electricity costs when making decisions on the day-ahead declared power and the maximum daily demand, expressed as:

[0090]

[0091] In the formula, This represents the mathematical expectation of the total electricity cost for a charging operator from day d to day D, the last day of the month, from the perspective of the day-ahead phase of day d. This represents the expected electricity cost for charging operators from day d to the last day of the month, day D, which is the expected electricity cost function for day d. Indicates as of the date Maximum demand generated by the end of the day ,for The independent variable.

[0092] In some preferred embodiments, the above-described S2 constructs up to the [number]th [unit / section]. The objective function of the expected future electricity cost function at the end of the day includes:

[0093] S221, According to Bellman's optimality theorem, the objective function of the expected future electricity cost on day d is expressed as:

[0094]

[0095] In the formula, It represents mathematical expectation and is used to handle uncertainties arising from multiple scenarios; This represents the total cost of electricity generated by all charging stations under the jurisdiction of the charging operator on day d. represents the battery wear cost caused by the charging and discharging of the energy storage system of each station under the jurisdiction of the charging operator within the dth day; represents the mathematical expectation of the electricity cost of the charging operator within the time period from the d+1th day to the last day D of this month, that is, the future electricity cost expectation function of the maximum demand generated by the end of the dth day; represents the mathematical expectation of the electricity cost of the charging operator within the time period from the d+1th day to the last day D of this month, that is, the future electricity cost expectation function of the maximum demand generated by the end of the dth day; represents a set of each time period within the dth day; represents a set of all charging stations under the jurisdiction of the charging operator k; 、 is a decision variable in the day-ahead decision-making of the dth day, representing the day-ahead declared power of the charging operator in each time period within the dth day and the energy storage charging and discharging power of the charging station j in each time period within the dth day, respectively;

[0096] S222, and satisfies the state transition equation:

[0097]

[0098] In the formula, represents the total actual power of the charging operator in the tth time period of the dth day under the scenario s from the day-ahead decision-making perspective; represents a set of all scenarios.

[0099] In some preferred embodiments, the above S222 can further include:

[0100] At the last day D of this month, is the maximum demand of the whole month, and the recursive part in the future electricity cost expectation function of the Dth day becomes the actual monthly demand electricity fee of the user, that is:

[0101] min

[0102] In the formula, represents the future electricity cost expectation function of the Dth day; represents the total energy cost of the charging operator generated within the Dth day under the jurisdiction of each station; represents the battery wear cost caused by the charging and discharging of the energy storage system of each station under the jurisdiction of the charging operator within the Dth day; represents the calculation formula of the monthly demand electricity fee.

[0103] Further, The general calculation formula of is as follows:

[0104]

[0105] In the formula, is the calculation function of the demand charge, is the maximum demand generated from the beginning of the month to the end of the Dth day of the month, is the contract demand charge, is the demand charge corresponding to each step, is the contract maximum demand, is the demand threshold value of each step. If the actual maximum demand is charged, then .

[0106] In some preferred embodiments, the above S221 can further include:

[0107] S2211, the total electric energy cost of the charging operator in the dth day is equal to the sum of the product of the electric energy cost under each scenario and the probability of the occurrence of the corresponding scenario; wherein the electric energy cost under each scenario includes day-ahead electricity charge, real-time electricity charge, deviation electricity charge and charging income; the deviation electricity charge is obtained by balancing the day-ahead electricity charge and the real-time electricity charge, which is used to ensure the rationality of the day-ahead declared power, and additionally increase the unit penalty cost times of the day-ahead declared power or times of the day-ahead declared power; ;

[0108] S2212, the total actual power of the charging operator in the t period of the dth day under the scenario s is calculated according to the following formula:

[0109]

[0110] In the formula, represents the charging power of the charging station j in the t period of the dth day under the scenario s, represents the net charging power of the energy storage system of the charging station j in the t period of the dth day under the scenario s, wherein the positive value represents charging and the negative value represents discharging;

[0111]

[0112] In the formula, , respectively represent the net charging power and discharging power of the energy storage system of the charging station j in the t period of the dth day under the scenario s;

[0113] S2113, the battery loss cost is proportional to the cumulative charging and discharging amount, which is represented as:

[0114]

[0115] In the formula,​​ represents the probability of the occurrence of the scenario s; represents the battery life loss cost caused by the unit charge and discharge capacity of the energy storage system.

[0116] In some preferred embodiments, the constraint condition of S2 for establishing the day-ahead decision model considering the future expected cost can further include:

[0117] S23, the constraint condition of the day-ahead decision model includes: energy storage system operation constraints, electric vehicle charging constraints, day-ahead declared power constraints, and real-time power constraints; wherein:

[0118] S231, the energy storage system operation constraints include: charge and discharge power limits and power limits; set represents the set from the dth day to the Dth day, then for satisfies:

[0119]

[0120]

[0121]

[0122]

[0123]

[0124]

[0125] In the formula, is a 01 variable used to represent the charge and discharge state of the energy storage configured by the charging station j on the dth day t period under the scenario s, and takes 1 to represent charging and takes 0 to represent discharging; , respectively represent the net charging power and discharging power of the energy storage system of the charging station j on the dth day t period under the scenario s; , respectively represent the maximum charging power and maximum discharging power of the energy storage configured by the charging station j; represents the storage power of the energy storage configured by the charging station j on the dth day t period; , respectively represent the charging efficiency and discharging efficiency of the energy storage configured by the charging station j; is the unit of the optimized time interval; , respectively represent the minimum storage power and maximum storage power of the energy storage configured by the charging station j;

[0126] S232, the electric vehicle charging constraint is represented as:

[0127]

[0128] wherein, represents the charging power of charging station j at time period t on day d under scenario s; represents the charging demand of charging station j at time period t on day d under scenario s in the day-ahead decision-making scenario of the charging operator;

[0129] S233, day-ahead declared power constraints, including:

[0130] The deviation between real-time power and day-ahead declared power should not exceed a times the day-ahead declared power, then for , the following constraint condition is established:

[0131]

[0132] wherein, represents the total actual power of the charging operator under scenario s at time period t on day d from the day-ahead decision-making perspective; represents the day-ahead declared power of the charging operator at each time period on day d;

[0133] When the deviation between real-time power and day-ahead declared power exceeds a times the day-ahead declared power, then for , the following constraint condition is established:

[0134]

[0135] wherein, represents the maximum charging power of charging station j, which is limited by the installed capacity of charging piles in the charging station;

[0136] S234, real-time power constraints, including:

[0137] The real-time power at each time period on day d should not exceed the maximum demand on day d determined in the day-ahead decision-making stage , then for , the following constraint condition is established:

[0138]

[0139] In some preferred embodiments, the above S3, which personalizes the deviation electricity fee, can further include:

[0140] S311, for the deviation electricity fee , two sets of continuous variables , are introduced, which respectively represent times the day-ahead declared power and The part of the multiple is doubled, while constraints are introduced:

[0141]

[0142]

[0143]

[0144]

[0145] where, is the real-time power of each time period on day d; denotes the day-ahead declared power of each time period on day d by the charging operator;

[0146] S312, the deviation electricity fee is denoted as:

[0147]

[0148] In some preferred embodiments, the above S3, linearizing the future electricity cost expectation function, can further include:

[0149] S321, for the future electricity cost expectation function, the value range of the function independent variable, the maximum demand already generated in the current month by the end of day d is divided into segment points by segments, and continuous variables , binary variables and the following constraints are introduced, to transform the maximum demand already generated in the current month into a linear combination of each segment point:

[0150]

[0151]

[0152]

[0153]

[0154]

[0155] where, the binary variable is used to indicate which segment the independent variable falls into, denotes falling into the lth segment ; by introducing the binary variable The last two constraints mentioned above are used to ensure that only a maximum of two adjacent elements exist. Non-zero values ​​ensure strict piecewise linearization.

[0156] The expected future cost function up to day d can be expressed as a linear combination of the function values ​​at each piecewise point:

[0157]

[0158] S322, therefore, the solution of the expected future cost function is transformed into a MILP problem. Based on this, a reverse construction method of the expected future cost function is adopted. By recursively working backward from day D to day d, the transformed MILP problem is solved to obtain the values ​​of the expected future cost function at each segment point. ; where, the expected future cost function on day D Values ​​at each segmentation point By using the values ​​of each segment point Substitution The solution yields the expected future cost function from day D-1 to day d. Values ​​at each segmentation point By using the values ​​of each segment point Substitution The solution yields the day-ahead declared power and the maximum demand boundary for day d, which can be obtained by recursively applying this sequence to day d.

[0159] Based on the same inventive concept, one embodiment of the present invention also provides a charging operator day-ahead strategy optimization system based on future expected costs.

[0160] Specifically, such as Figure 2 As shown, the day-ahead strategy optimization system for charging operators based on future expected costs provided in this embodiment may include:

[0161] The decision-making scenario generation module generates the day-ahead decision-making scenarios for charging operators based on rolling market information forecasts.

[0162] The day-ahead decision model construction module establishes a day-ahead decision objective function and constraints that take into account future expected costs based on the day-ahead decision scenario, constructs a future electricity cost expectation function that includes demand electricity charges, and obtains the total electricity cost including deviation electricity charges.

[0163] An optimization problem linearization module is configured to linearize the future electricity cost expectation function and the deviation electricity fee to obtain a linearized optimization problem; for the deviation electricity fee, two sets of continuous variables are introduced to represent the parts of the real-time power exceeding the set multiple of the day-ahead declared power and the parts of the real-time power lower than the set multiple of the day-ahead declared power at the t time period of the d day under the scenario s, and linearization conversion is performed; for the future electricity cost expectation function, the value range of the maximum demand generated in the current month by the time the current day ends is divided into segments by segmentation points, and continuous variables and binary variables are introduced to convert the maximum demand into a linear combination of the segmentation points;

[0164] A day-ahead strategy optimization module is configured to solve the linearized optimization problem to obtain the charging operator day-ahead strategy.

[0165] It should be noted that the steps in the method provided by the present application can be implemented by using corresponding components in the system, and those skilled in the art can refer to the technical solution of the system to implement the step flow of the method, or refer to the technical solution of the method to implement the composition of the system, that is, the embodiments in the system and the embodiments in the method can be understood as preferred examples, which will not be described here.

[0166] The charging operator day-ahead strategy optimization method and system based on future expected cost provided by the above embodiments of the present application realize a charging operator day-ahead strategy optimization technology based on future expected cost through market information rolling prediction, which maximally utilizes the continuously updated market information to reasonably predict the future electricity price scenario; the influence of the maximum demand on the total cost of the charging operator in the demand electricity fee and the subsequent optimization boundary is considered, and the maximum demand in the whole cycle cost influence is accurately modeled. Among them, the electricity fee cost as a guide signal can guide the charging operator to actively reduce the peak load in the day-ahead decision, thereby reducing the operation pressure of the power grid load peak period; and can be used as the basis for deviation electricity settlement in the day-ahead decision of the charging operator, to ensure the rationality of the operator day-ahead declared power.

[0167] The technical solutions and technical principles of the above embodiments of the present application will be further described in detail below.

[0168] In the operation of the charging operator, the day-ahead decision of each day is only one stage of the whole month, therefore, the day-ahead decision of the charging operator is a typical multi-stage decision process. The core idea of Bellman optimality principle is that the optimal strategy has "self-consistency" or "optimal substructure", that is, the solution of any sub-problem is also optimal. In mathematical language, if an optimization problem contains B stages of sequential decisions , and the strategy is the optimal strategy of this multi-stage optimization problem, then for any integer , if the decisions up to the b-th decision (called stage b) are , then in the system state at this time, the subsequent optimal strategy must be Based on the Bellman optimality principle, the multi-stage optimization problem can be solved.

[0169] For the day-ahead decision problem, the uncertainty of the subsequent state mainly lies in the uncertainty of the future market information and charging load relative to the decision time. Therefore, based on the Bellman optimality principle, the uncertainty of the future value function can be included in the decision objective function by the following formula, and the monthly demand charge is included in the daily day-ahead decision, thereby avoiding the inaccurate modeling of the demand charge in the traditional model:

[0170]

[0171] In the formula, is the value function, indicating the system state corresponding to the system value, is the feasible strategy set in stage b, is the system state after the strategy is adopted in stage b, which is jointly determined by the system state before stage b decision and the strategy adopted in stage b, and the process of the system state change can be described by the transition state equation . is the system value of adopting strategy in stage b, indicates the mathematical expectation of the next stage system value.

[0172] The daily maximum load of the charging operator in the whole month is directly related to the decisions of future stages, so the Bellman optimality principle can be used to determine this objective.

[0173] The day-ahead decision stage of the charging operator in each day needs to determine the day-ahead declared power of each period of the day and the maximum demand to be maintained in the day, and gives the maximum power constraint that cannot be broken in the real-time operation stage of the day.

[0174] Based on the above objective, a day-ahead decision model considering the expected future cost is established. Specifically:

[0175] The objective function is established:

[0176] Suppose that in the day-ahead decision stage of the charging operator on the d-th day, the maximum demand generated by the charging operator k in the month is . Updates can be obtained based on actual electricity consumption after the charging operator's daily operations conclude (for...). The situation makes Therefore, for the day-ahead decision on day d, It is a definite constant.

[0177] When charging operator k makes its day-ahead decision on day d, it needs to consider the expected costs over the period from day d until the last day of the month (day D), referred to as the expected future electricity costs. During the daily day-ahead decision-making phase, the objective function is to minimize the expected future electricity costs when deciding on the day-ahead declared power and the maximum daily demand. This objective function can be expressed as:

[0178] (1)

[0179] in, This represents the mathematical expectation of the total electricity cost for a charging operator from day d to the end of the month, from the perspective of the day-ahead phase. This represents the expected electricity cost function for day d. As the expected function of future electricity costs The independent variable represents the value up to the [number]. The maximum demand generated at the end of the day. According to Bellman's optimality theorem, this expectation function is a recursive function, that is:

[0180] (2)

[0181] in, It refers to mathematical expectation, used to handle uncertainties arising from various scenarios. It is the total electricity charge generated by the charging operator's various stations within day d. It is the cost of battery loss caused by the charging and discharging of the energy storage systems of the charging operators' stations within day d. It is the set of all time periods within day d. It is the collection of all charging stations under the jurisdiction of charging operator k. , ( , ) are the decision variables in the day-ahead decision-making process for day d, representing the day-ahead declared power of the charging operator and the energy storage charging / discharging power of charging station j in each time period of day d, respectively. It should be noted that although a charging / discharging power plan for the energy storage system on day d is formulated by solving an optimization problem during the day-ahead decision-making phase, this plan is not used as a boundary condition in the real-time phase. In contrast, the day-ahead declared power will be actually reported to ISO after the strategy solution is completed. is the maximum demand generated in this month up to the end of day d, which is the sum of the maximum demand generated in this month up to the end of day d-1 and the demand generated in day d, i.e., satisfies the state transition equation:

[0182]

[0183] In particular, at the end of the month on the last day (D), is the maximum demand in the whole month, so the recursive part in the future electricity cost expectation function of day D becomes the actual monthly demand electricity bill of the user, i.e.,

[0184] min (3)

[0185] wherein, is the calculation formula of the monthly demand electricity bill, which is expressed as:

[0186]

[0187] wherein, is the maximum demand generated in the whole month to the end of day D at the end of the month, is the contract demand electricity price, ( ) is the demand electricity price corresponding to each step, is the contract maximum demand, ( ) is the demand threshold value of each step. If the actual maximum demand is charged, then .

[0188] According to the nature of the objective function, it can be found that only for the future cost expectation function of day D, this optimization problem is a mixed integer linear programming (MILP) problem.

[0189] According to the idea of stochastic programming, the total electricity energy cost of the charging operator in day d is equal to the sum of the product of the electricity energy cost in each scenario and the probability of the occurrence of the scenario. The electricity energy cost in each scenario includes day-ahead electricity bill , real-time electricity bill , deviation electricity bill and charging income . The deviation electricity bill is obtained by deviation settlement of the day-ahead electricity bill and the real-time electricity bill, which is used to ensure the rationality of the day-ahead declaration. For the part of the real-time electricity consumption that is more than times of the day-ahead declared power or less than times of the day-ahead declared power, an additional unit penalty cost of is added.

[0190] ​

[0191]

[0192]

[0193]

[0194]

[0195] where, is the set of all scenarios, represents the probability of scenario s occurring, and the sum of all scenario probabilities . and represent the day-ahead market price and intra-day market price at time period t on day d under scenario s, respectively, represents the charging price of charging station j at time period t on day d under scenario s. Since there are a large number of loads participating in bidding in the market, a single charging operator can be approximated as a price taker, whose bid in the market does not affect the market clearing price. is the total actual power of the charging operator at time period t on day d under scenario s from the day-ahead decision perspective, which can be calculated according to the following formula:

[0196]

[0197] where, is the charging power of charging station j at time period t on day d under scenario s, is the net charging power of the energy storage system of charging station j at time period t on day d under scenario s (a positive value represents charging and a negative value represents discharging), which can be further decomposed into charging power and discharging power:

[0198]

[0199] To simplify the model, the battery wear cost is simplified to be proportional to the cumulative charging and discharging amount, i.e.,

[0200]

[0201] where, is the battery life loss cost per unit of charging / discharging of the energy storage system. Generally speaking, each charging and discharging cycle will cause the life of the energy storage system to decay, and this decay is usually considered to be related to the charging and discharging amount and the depth of discharge.

[0202] Therefore, the objective function of the day-ahead decision model is a decision problem in stages of days, with the maximum demand of each day coupling between stages.

[0203] Establish the constraint conditions:

[0204] The main constraints of the decision model include energy storage system operation constraints, electric vehicle charging constraints, day-ahead declared power constraints, and real-time power constraints.

[0205] 1. Energy storage system operation constraints

[0206] The operation of the energy storage system needs to consider the limits of charging / discharging power and electric quantity. Let represent the set from day d to day D, then for , it needs to satisfy:

[0207] (4)

[0208] (5)

[0209] (6)

[0210] (7)

[0211] (8)

[0212] (9)

[0213] wherein is a 01 variable, used to represent the charging / discharging state of the energy storage configured by charging station j on day d at time period t under scenario s, taking 1 when charging and 0 when discharging; , and represent the maximum charging power and the maximum discharging power of the energy storage configured by charging station j, respectively; , represent the charging efficiency and the discharging efficiency of the energy storage configured by charging station j, respectively; is the optimized time interval unit; , represent the minimum and maximum storage electric quantity of the energy storage configured by charging station j, respectively.

[0214] Constraints (4)-(5) describe the charging / discharging power limits of the station-configured energy storage, constraint (6) describes the relationship between the storage electric quantity and the charging / discharging power of the station-configured energy storage, and constraints (7)-(9) represent the storage electric quantity limits.

[0215] 2. Electric vehicle charging constraints

[0216] The charging power provided by the charging station for the electric vehicle should not exceed the charging demand in this scenario, which also means that in certain cases, the real-time charging power may be lower than the charging demand in the corresponding scenario, that is, the charging operator may refuse part of the charging demand in the real-time operation phase according to the maximum demand reduction or other operation needs:

[0217]

[0218] 3. Day-ahead declared power constraint

[0219] According to existing research, ISO has a deviation assessment limit between the day-ahead declared power and the real-time power of each load in the power market. The deviation between the real-time power and the day-ahead declared power should not exceed α times the day-ahead declared power. Therefore, for :

[0220]

[0221] In addition, if the day-ahead declared power is too large and exceeds α times the sum of the maximum power consumption of all charging stations under the jurisdiction of the charging operator, the real-time power cannot be kept within the deviation assessment interval through power scheduling in the real-time phase. Therefore, for :

[0222]

[0223] wherein is the maximum charging power of charging station j, which is limited by the installed capacity of the charging pile in the charging station.

[0224] 4. Real-time power constraint

[0225] In the day-ahead decision-making model, the real-time power of each period on the dth day should not exceed the maximum demand of the dth day determined in the day-ahead decision-making phase, that is, for :

[0226]

[0227] Linearization of the decision-making model:

[0228] In the above optimization problem, the deviation cost in the objective function and the future cost expectation function are nonlinear functions, which makes it difficult to solve the optimization problem by using existing solvers. Therefore, the piecewise linearization method is used to convert these two functions into linear functions.

[0229] For , two sets of continuous variables , , respectively, represent the part of the day-ahead declared power that is exceeded and the part of the day-ahead declared power that is undershot, respectively, by the real-time power at time t on day d under scenario s.

[0230] (10)

[0231] (11)

[0232] (12)

[0233] (13)

[0234]

[0235] (14)

[0236] Since the optimization objective is to minimize the objective function, in the process of the optimizer solving, will naturally take its minimum value that meets the constraints, so the constraints (10)-(13) and equation (14) can completely replace the original calculation of .

[0237] For , the value range of the function argument is divided into segments by segment points , and the continuous variable , the binary variable , and the following constraints are introduced, which can convert into a linear combination of each segment point.

[0238] (15)

[0239] (16)

[0240] (17)

[0241] (18)

[0242] (19)

[0243] where the binary variable is used to indicate which segment of the value range falls into, ​​​​​​It falls in the first paragraph By introducing binary variables The constraints (18)-(19) ensure that there are at most two adjacent elements. Non-zero values ​​ensure strict segmentation of the linearization.

[0244] Furthermore, the expected function of future costs This can be expressed as a linear combination of the function values ​​at each piecewise point:

[0245]

[0246] Thus, the optimization problem (2) is transformed into a MILP problem, which can be easily solved using a commercial solver. Based on this, the values ​​of the function at each segment point can be obtained by recursively solving the planning problem (2) from day D back to day d. Among them, the expected future cost function on day D. Values ​​at each segmentation point By using the values ​​of each segment point Substitution The solution yields the expected future cost function from day D-1 to day d. Values ​​at each segmentation point By using the values ​​of each segment point Substitution The solution is obtained. Following this sequence, we can recursively extend to day d. This process of solving the MILP problem can be called the reverse construction of the future expected cost function, represented by the following pseudocode.

[0247]

[0248] After completing the reverse construction of the future expected cost function, the actual Substituting these values ​​into the optimization problem, we can obtain the optimal day-ahead strategy for day d, which includes the application power for each time period in the day-ahead market on day d and the maximum demand that needs to be maintained during the real-time operation phase on day d.

[0249] Through the technical solutions and technical principles, the day-ahead decision of the charging operator is a rolling updating process. In the day-ahead decision stage of each decision day d, the charging operator needs to first predict the day-ahead electricity price and the real-time electricity price from the dth day to the end of the month (Dth day) according to the latest market information of the electricity market, and predict the charging load from the dth day to the end of the month according to the vehicle charging demand, to generate sufficient future market-load scenarios, and then select representative scenarios through a clustering algorithm. Based on these scenarios, the charging operator will first construct a piecewise linear future cost function through a future expected cost function algorithm, and then construct a piecewise linear future cost function from the Dth day to the dth day. Finally, the day-ahead declared power and the daily maximum demand boundary of the dth day are obtained through an optimization model based on the future expected cost function. For the day-ahead decision of each decision day, the complete day-ahead decision process is as shown in Figure 3 .

[0250] The application effect of the day-ahead decision technology provided by the above embodiments of the application is verified through a specific verification example.

[0251] A typical region is selected for example analysis in the specific verification example. There are 83 charging stations in the region, belonging to 3 charging operators, of which 28 belong to A operator, 46 belong to B operator, and 9 belong to State Grid. In the example of this chapter, A operator is taken as the research object to analyze the day-ahead optimization decision of A operator. The distribution of charging stations in the region is as shown in Figure 4 .

[0252] At present, the electricity market in which the typical region is located is still in the initial construction stage, and a long-term stable market operation and information disclosure mechanism has not yet been formed. Therefore, in order to simulate the day-ahead optimization decision in the market environment, the NYISO electricity market data in the United States is used for verification in the specific verification example. The market load prediction model and the day-ahead electricity price prediction model use the load and electricity price data of the NYC (New York City) region of NYISO from 00:00 on April 1, 2024 to 23:59 on August 31, 2024 for training. The simulation of day-ahead optimization decision in this example uses the data of this region from 00:00 on September 1, 2024 to 23:59 on September 30, 2024. In addition, in order to match the market information, the weather data of the same period from the WeatherAPI website is used. The price of demand charge is 38.4 yuan / kW, and the maximum demand that has occurred at the beginning of the month is set to 0.

[0253] Assume that operator A has 28 charging stations equipped with 8 sets of station-mounted energy storage systems, each with a capacity of 200kWh and a maximum charging / discharging power of 100kW. The charging / discharging efficiency of the energy storage and the charging efficiency of the charging piles are set to 0.95. The maximum power of each charging station's connection point to the grid is set to 2000kW.

[0254] In the rolling forecasting framework for current decision-making scenarios, the overall market load forecast is achieved through a two-layer LSTM. To verify the effectiveness of the proposed forecasting method, Figure 5 The hourly total load of the NYC region in September 2024 was compared with the hourly total load forecast for the entire month at the pre-decision stage on Day 1. The line graph above shows the Mean Absolute Percentage Error (MAPE) of the forecast results from Day 1 to the end of each day. The results show that the proposed method generally captures the cyclical characteristics of the load well. Forecast results closer to the forecast time point perform well, but the performance decreases somewhat for loads far from the forecast time point. This is because for periods far from the forecast time point, the input characteristics consist entirely of the forecast values ​​and do not reflect the actual supply and demand relationship and environmental conditions in the market; therefore, a larger deviation between the forecast values ​​and the actual values ​​is logical. However, overall, the MAPE of the forecast results does not exceed 8%, indicating that the forecast performance of the proposed method is good in absolute terms.

[0255] The training of current electricity price prediction models requires first comparing the performance of various kernel functions on the training set and selecting the best-performing kernel function. Table 1 compares the root mean square error (RMSE) and coefficient of determination of five kernel functions. (Coefficient of Determination). The root mean square error (RMSE) is the square root of the average of the squared prediction errors, used to measure the absolute deviation between the predicted and actual values ​​over time; the coefficient of determination reflects the model's explanatory power for fluctuations in the target variable, and its range is... The closer the value is to 1, the better the model's interpretability. Based on the comparison of these two indicators, it can be seen that among the five kernel functions tested, the exponential kernel performs best. Therefore, the day-ahead electricity price forecasting model uses the exponential kernel for training.

[0256] Table 1. Indicators of the 5 kernel functions

[0257]

[0258] Figure 6 The results of the day-ahead electricity price forecasting model are presented. Figure 6In the middle, the purple broken line represents the actual day-ahead price of NYISO in September 2024, and the light blue color band represents the hourly day-ahead price prediction results of the whole month obtained by the above rolling prediction framework in the first day-ahead decision-making stage. The upper and lower boundaries of the color band are ±2 standard deviations, i.e. 95% confidence interval. Similarly, the light brown color band represents the hourly day-ahead price predicted in the 16th day-ahead decision-making stage until the end of the month. From the results, it can be seen that, similar to the characteristics of the load prediction results, the day-ahead price rolling prediction framework proposed in the present application will produce larger deviations for the day-ahead prices far from the prediction time point, but the prediction effect for the near future day-ahead price at the prediction time point is better. Within 3 days after the prediction time point, the actual day-ahead price is basically within the prediction interval. Since the day-ahead decision-making method proposed in the present application gives the bidding strategy within the day in each daily day-ahead decision-making stage, and the long-term day-ahead price is only used as a reference to construct the future expected cost, the proposed rolling prediction framework will not have a great impact on the daily day-ahead decision-making.

[0259] For the daily day-ahead decision-making process, on the basis of the day-ahead price prediction interval, 20 groups of day-ahead prices are randomly generated by the Monte Carlo simulation method. For each group of day-ahead prices, 5 groups of real-time prices are generated by randomly sampling the deviation between the real-time price and the day-ahead price in the historical data and adding it to the day-ahead price. According to the data from February 1, 2024 to August 31, 2024, the deviation between the real-time price and the day-ahead price of the NYC region of NYISO is as shown in Figure 7 . Except for the 5 samples with a deviation of more than 20% which are not shown in Figure 7 for drawing effect, it can be seen that the deviation of most samples is within ±5%.

[0260] Through the above process, for each decision-making day, a total of 20x5=100 future market price scenarios are obtained. Corresponding to the market price scenarios, based on the charging demand simulation method (charging demand space-time distribution prediction method), 100 charging load simulations of the whole Shanghai are performed, and the charging load of the public charging stations in the research target range is taken to obtain 100 hourly charging load scenarios. In order to establish an intuitive understanding of the charging scenarios, Figure 8 the first 7 days of the 100 charging load scenarios generated in the day-ahead decision-making stage on September 1, 2024 are shown. Among them, each scenario is represented by a curve with a transparency of 5%, so Figure 8 the color depth roughly reflects the probability density distribution of the charging load.

[0261] For the daily day-ahead decision-making phase, the 100 future market electricity price scenarios and 100 charging loads obtained above are combined one by one to form a total of 10,000 day-ahead decision-making scenarios. K-means clustering is then performed on these scenarios to obtain the 100 day-ahead decision-making scenarios and their probabilities of occurrence, which serve as boundary conditions for the optimization problem.

[0262] During the daily pre-decision phase, the expected cost function for the remaining days of the month needs to be calculated backwards based on future scenarios. This is specifically represented by the expected cost function values ​​at various segment points. In this example, eight segment points are set, evenly distributed between 0.75 and 1.2 times the maximum charging load. To visually represent the shape of the expected cost function, the expected cost function values ​​for the 4th, 11th, 20th, and 26th days of September 1, 2024, obtained during the pre-decision phase, are displayed respectively. Figure 9 In (a) to (d), each point marked with an asterisk is... The segmentation points and their corresponding Function value.

[0263] from Figure 9 As can be seen, the expected cost function initially remains constant and then monotonically increases. This characteristic mathematically aligns with general intuition: for the day-ahead decision-making stage on day d, if the maximum demand up to day d is low, the already generated maximum demand will not affect future electricity costs because subsequent maximum demands will exceed the already generated maximum demand. Conversely, if the already generated maximum demand is large, it will have an irreversible impact on the total monthly electricity cost. Although the already generated maximum demand relaxes the boundary conditions for subsequent optimization, this effect is insufficient to offset the impact of the already generated maximum demand on the total monthly electricity cost.

[0264] In addition, from Figure 9As can be seen, the inflection point where the slope of the future expected cost function curve changes from 0 to a positive real number shifts towards the negative x-axis as the date progresses. This is because the later the date, the greater the impact of the maximum demand already generated on subsequent future expected costs. In other words, the later the date, the more dominant demand-based electricity costs become in future expected costs. To illustrate this trend, a visual example is that in the day-ahead decision-making stage on day 1, the demand-based electricity costs for the entire month, along with the total electricity costs for the following D days, constitute future expected costs. However, in the day-ahead decision-making stage on day D, the demand-based electricity costs for the entire month, along with the electricity costs for the following day, constitute future expected costs, thus significantly increasing the proportion of demand-based electricity costs. By day 26, the future expected cost function is already above 0 in the latter part of the curve. This means that if charging operators fail to reasonably control maximum demand before day 26, the charging revenue for the following days will be insufficient to cover the electricity costs, including the total demand-based electricity costs for the entire month. This illustrates the importance of maximum demand control from another perspective.

[0265] The total monthly reporting power obtained by the reporting strategy proposed in this invention is as follows: Figure 10 As shown (images are for clear display, drawn from the previous 3 days). From Figure 10 As can be seen, the proposed day-ahead reporting strategy is largely consistent with the weighted average trend of charging load scenarios, and can basically reflect the actual demand for charging load. In addition, the proposed day-ahead reporting strategy limits the maximum daily demand to 7000kW, which reduces the peak load generated by charging operators by nearly half compared with the most extreme charging demand scenarios. This helps to alleviate the short-term power surge caused by electric vehicle charging, reduce the risk of grid accidents, and improve the overall safety and stability of the distribution network operation.

[0266] Any matters not covered in the above embodiments of the present invention are well-known in the art.

[0267] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various modifications or variations within the scope of the claims, which do not affect the essence of the present invention.

Claims

1. A method for optimizing the day-ahead strategy of charging operators based on expected future costs, characterized in that, include: Based on rolling market information forecasts, generate day-ahead decision-making scenarios for charging operators; Based on the aforementioned day-ahead decision-making scenario, a day-ahead decision-making objective function and constraints that take into account future expected costs are established, a future electricity cost expectation function that includes demand electricity charges is constructed, and the total electricity cost including deviation electricity charges is obtained. The expected future electricity cost function and the deviation electricity fee are linearized to obtain the linearized optimization problem. For the deviation electricity fee, two sets of continuous variables are introduced, representing the portion of real-time power exceeding a set multiple of the previously reported power and the portion falling below a set multiple of the previously reported power during time period t on day d, respectively, under scenario s. These are then linearized. For the expected future electricity cost function, the range of values ​​for the maximum demand generated this month up to the end of day d is determined by... Each segment point is divided into The maximum demand generated this month up to the end of day d is transformed into a linear combination of the segment points by introducing continuous and binary variables; the expected future electricity cost function up to the end of day d is expressed as a linear combination of the function values ​​of the segment points. Solving the linearized optimization problem yields the day-ahead strategy for charging operators; The method for generating day-ahead decision-making scenarios for charging operators based on rolling market information forecasts includes: Get real-time market information; Based on the market information, predict the hourly load from day d to day D; where day D is the last day of the month. Based on the hourly load, calculate the mean and variance of the day-ahead electricity price forecast for each hour from day d to day D; Based on the mean and variance of the day-ahead electricity price forecast, day-ahead electricity price scenarios are generated by sampling, wherein each day-ahead electricity price scenario includes the hourly day-ahead electricity price from day d to day D. For each day-ahead electricity price scenario, real-time electricity prices are generated by random sampling to form day-ahead-real-time electricity price scenarios; Based on charging demand, vehicle charging behavior is simulated, the hourly charging load of each station under the charging operator is calculated, and charging load scenarios are generated. Clustering the day-ahead electricity price scenario, day-ahead-real-time electricity price scenario, and charging load scenario obtained above, we get multiple day-ahead decision scenarios. Each day-ahead decision scenario includes the hourly day-ahead electricity price, real-time electricity price, and charging load from day d to day D. Establish a day-ahead decision objective function that takes into account expected future costs, including: Based on the aforementioned day-ahead decision-making scenario, a day-ahead decision-making objective function considering future expected costs is established. This objective function aims to minimize the expected future electricity costs when making decisions regarding day-ahead declared power and the maximum daily demand, and is expressed as follows: ； In the formula, This represents the mathematical expectation of the total electricity cost for a charging operator from day d to day D, the last day of the month, from the perspective of the decision made on day d. This represents the expected electricity cost for charging operators from day d to the last day of the month, day D, which is the expected electricity cost function for day d. Indicates as of the date The maximum demand generated this month by the end of the day was The independent variable; According to Bellman's optimality theorem, the objective function of the expected future electricity cost on day d is expressed as: ； In the formula, It represents mathematical expectation and is used to handle uncertainties arising from multiple scenarios; This represents the total cost of electricity generated by all charging stations under the jurisdiction of the charging operator on day d. This represents the battery loss cost caused by the charging and discharging of the energy storage systems at each station under the charge operator's jurisdiction during day d. This represents the mathematical expectation of electricity costs for charging operators from day d+1 to day D, the last day of the month, relative to the maximum demand generated by the end of day d. The function is the expected future electricity cost function up to the end of day d; This represents the set of time periods within day d. This represents the set of all charging stations under the jurisdiction of charging operator k; , Let be the decision variables in the decision-making process on day d, and let represent the day-ahead declared power of the charging operator and the energy storage charging and discharging power of charging station j in each time period on day d, respectively. and Satisfies the state transition equation: ； In the formula, This represents the total actual power output of charging operators during time period t on day d, under scenario s, from the perspective of recent decision-making. This represents the set of all scenarios, occurring on the last day of the month, day D. Given the maximum demand for the entire month, the recursive part of the expected electricity cost function for day D represents the user's actual monthly electricity demand cost, i.e.: min ; In the formula, This represents the expected future electricity cost function on day D; This represents the total cost of electricity generated by the charging operator's various stations within day D. This represents the battery loss cost caused by the charging and discharging of the energy storage systems at each station under the charging operator's jurisdiction during day D. The formula for calculating monthly electricity demand costs; The total cost of electricity generated by the charging operator's various stations within day d. It equals the sum of the products of the electricity cost in each scenario and the probability of that scenario occurring; where the electricity cost in each scenario includes day-ahead electricity cost, real-time electricity cost, deviation electricity cost, and charging revenue; the deviation electricity cost is calculated from the day-ahead electricity cost and the real-time electricity cost, and also includes the deviation electricity cost for real-time electricity consumption exceeding the day-ahead declared power. The power output is twice or less than the power output declared previously. The additional unit penalty cost is added to the portion that is doubled. Establish day-ahead decision constraints that take into account expected future costs, including: energy storage system operation constraints, electric vehicle charging constraints, day-ahead power reporting constraints, and real-time power constraints; among which: The electric vehicle charging constraint is expressed as: ； In the formula, This represents the charging power of charging station j during time period t on day d in scenario s; This indicates the charging demand at charging station j during time period t on day d in scenario s of the charging operator's recent decision-making scenario.

2. The charging operator day-ahead strategy optimization method based on future expected cost according to claim 1, characterized in that, In scenario s, the total actual power of the charging operator during time period t on day d. The following formula is used to calculate: ； In the formula, This represents the charging power of charging station j during time period t on day d in scenario s. This represents the net charging power of the energy storage system of charging station j during time period t on day d under scenario s, where positive values ​​represent charging and negative values ​​represent discharging. ； In the formula, , Let represent the net charging power and discharge power of the energy storage system of charging station j during time period t on day d under scenario s, respectively; Battery loss cost Proportional to the cumulative charge and discharge amount, expressed as: ； In the formula, Indicates the probability of scenario s occurring; This represents the cost of battery life loss per unit charge / discharge of an energy storage system.

3. The charging operator day-ahead strategy optimization method based on future expected cost according to claim 1, characterized in that, The operating constraints of the energy storage system include: charging and discharging power limits and energy limits; [The following is a list of constraints:] Let represent the set from day d to day D. Then for ,satisfy: ； ； ； ； ； ； In the formula, It is a 01 variable used to represent the charging and discharging status of the energy storage configured at charging station j in the t period of day d under scenario s. When it is 1, it means that it is charging, and when it is 0, it means that it is discharging. , Let represent the net charging power and discharge power of the energy storage system of charging station j during time period t on day d under scenario s, respectively; , These represent the maximum charging power and maximum discharging power of the energy storage configured at charging station j, respectively. This represents the stored energy capacity of the energy storage system configured at charging station j during time period t on day d in scenario s. , These represent the charging efficiency and discharging efficiency of the energy storage configured at charging station j, respectively. For optimized time interval units; , These represent the minimum and maximum energy storage capacities of the energy storage system configured for charging station j, respectively. The previously declared power constraints include: The deviation between real-time power consumption and the previously reported power consumption should not exceed α times the previously reported power consumption. Establish the following constraints: ； In the formula, This represents the total actual power output of charging operators during time period t on day d, under scenario s, from the perspective of recent decision-making. This indicates the day-ahead reported power output of the charging operator for each time period within day d; When the deviation between the real-time power consumption and the previously reported power consumption exceeds α times the previously reported power consumption, then for Establish the following constraints: ； In the formula, This indicates the maximum charging power of charging station j, which is limited by the installed capacity of the charging piles within the charging station. The real-time power constraint includes: In scenario s, the total actual power of the charging operator during time period t on day d. It should not exceed the maximum demand generated in the month as determined during the previous decision-making phase up to the end of day d. For Establish the following constraints: 。 4. The charging operator day-ahead strategy optimization method based on future expected cost according to claim 1, characterized in that, Linearizing the aforementioned deviation electricity charges specifically includes: For deviation electricity charges Introduce two sets of continuous variables , , respectively, represent the time period t on day d under scenario s, where the real-time power exceeds the power declared the day before. The power output is twice that of the power output reported earlier. For the multiple portion, constraints are introduced simultaneously: ； ； ； ； In the formula, This indicates the day-ahead reported power output of the charging operator for each time period within day d; Deviation in electricity charges Represented as: ； In the formula, Penalty cost per unit.

5. The method for optimizing the day-ahead strategy of charging operators based on future expected costs according to claim 1, characterized in that, Linearizing the expected function of future electricity costs specifically includes: For the expected electricity cost function, the independent variable of the expected function is the maximum demand generated this month up to the end of day d. The range of values ​​is determined by Segmentation points Divided into Segment, and introduce continuous variables binary variables And the following constraints, which will determine the maximum demand generated in the current month up to the end of day d. Transform it into a linear combination of the segment points: ； ； ； ； ； In the formula, the two variables Used to indicate the independent variable Which segment of the range of values ​​does it fall into? express It falls in the first paragraph By introducing binary variables The last two constraints mentioned above are used to ensure that only a maximum of two adjacent elements exist. Non-zero values ​​ensure strict piecewise linearization. The expected future electricity cost up to the end of day d can be expressed as a linear combination of the function values ​​at each piecewise point: ； Therefore, solving the expected function of future electricity costs is transformed into a MILP problem. Based on this, by recursively solving the transformed MILP problem from day D back to day d, the values ​​of the expected function of future electricity costs at each segment point are obtained. ; where, the expected function of future electricity costs on day D. Values ​​at each segmentation point By using the values ​​of each segment point Substitution The solution yields the expected function of future electricity costs from day D-1 to day d. Values ​​at each segmentation point By using the values ​​of each segment point Substitution The solution yields the day-ahead declared power and the maximum demand boundary for day d, which can be obtained by recursively applying this sequence to day d.

6. A charging operator day-ahead strategy optimization system based on future expected cost, used to implement the charging operator day-ahead strategy optimization method based on future expected cost according to any one of claims 1-5, characterized in that, include: The decision-making scenario generation module generates day-ahead decision-making scenarios for charging operators based on rolling market information forecasts. The day-ahead decision model construction module, based on the day-ahead decision scenario, establishes a day-ahead decision objective function and constraints that take into account future expected costs, constructs a future electricity cost expectation function that includes demand electricity charges, and obtains the total electricity cost including deviation electricity charges; The optimization problem linearization module is used to linearize the expected future electricity cost function and the deviation electricity fee, resulting in a linearized optimization problem. For the deviation electricity fee, two sets of continuous variables are introduced, representing the portion of real-time power exceeding a set multiple of the previously reported power and the portion falling below a set multiple of the previously reported power during time period t on day d, respectively, under scenario s. These are then linearized. For the expected future electricity cost function, the range of values ​​for the maximum demand generated this month up to the end of day d is determined by... Each segment point is divided into The maximum demand generated this month up to the end of day d is transformed into a linear combination of the segment points by introducing continuous and binary variables; the expected future electricity cost function up to the end of day d is expressed as a linear combination of the function values ​​of the segment points. The day-ahead strategy optimization module is used to solve the linearized optimization problem to obtain the day-ahead strategy of the charging operator.

Images