CN109359389B - Urban electric vehicle charging decision method based on typical load dynamic game - Google Patents

Abstract

The invention provides a typical load dynamic game-based urban electric vehicle charging decision method, and relates to the technical field of electric vehicle charging. The method comprises the steps of obtaining a typical user load type in a certain area according to power grid short-term load curve clustering, obtaining an ideal charging load curve of the electric automobile through calculation, establishing an optimization equation by taking charging demand constraint, charging time constraint, electric automobile charging quantity limitation, a power grid power load curve and power distribution network capacity as boundary conditions, taking the minimum error between the charging load curve and the ideal charging load curve as constraint conditions, taking the minimum user power consumption cost as a target function, solving to obtain a charging strategy, and judging whether to charge the electric automobile or not through a decision result of dynamic game. The method provided by the invention can improve the load characteristics of the power grid, so that the fluctuation of the power grid after the electric automobile is connected is minimum, the load rate of the power equipment is improved, and the charging cost of electric automobile users is minimum on the basis.

Description

Urban electric vehicle charging decision-making method based on typical load dynamic game

Technical Field

The invention relates to the technical field of charging of electric charging automobiles, in particular to a city electric automobile charging decision method based on a typical load dynamic game.

Background

The global energy crisis and climate change drive the rapid development and application of electric vehicles worldwide. With the popularization of electric vehicles in the future, large-scale electric vehicles are merged into a power grid for charging and discharging, however, with the great increase of the number of electric vehicles, the large-scale access of the electric vehicles brings obvious harm to the operation and control of a power system due to the randomness of the charging behaviors of the electric vehicles. How the power grid accommodates large-scale electric automobile grid connection and the capacity of the power grid is not exceeded has become a great problem which must be solved in the future electric automobile development and popularization process. In order to solve the difficulties, reasonable means are needed to regulate and control the charging and discharging behaviors of the electric automobile, a more accurate electric automobile load modeling method and a more reasonable scheme are needed to guide the ordered charging and discharging of the electric automobile, and the grid-connected capacity of the electric grid for consuming large-scale electric automobiles is improved as much as possible. The peak avoiding and valley filling are used as a flexible regulating and controlling means, the charging mode and time selection of the electric automobile can be guided to a certain degree, the electric automobile can be charged and discharged in order, harm to a power grid caused by the addition of the electric automobile can be avoided, meanwhile, the experience degree of a charging user can be increased, and the user can enjoy lower charging price.

Disclosure of Invention

The technical problem to be solved by the invention is to provide a city electric vehicle charging decision method based on a typical load dynamic game aiming at the defects of the prior art, which can improve the load characteristic of a power grid, minimize the fluctuation of the power grid after the electric vehicle is connected, improve the load rate of power equipment, and minimize the charging cost spent by electric vehicle users on the basis.

In order to solve the technical problems, the technical scheme adopted by the invention is as follows:

a city electric vehicle charging decision-making method based on typical load dynamic game is realized by installing a general intelligent scheduling device on a transformer branch, wherein the intelligent scheduling device controls n charging piles of the transformer branch, when an electric vehicle is accessed, the charging piles upload electric vehicle information and send the information to the intelligent scheduling device, the intelligent scheduling device sends a charging strategy to the charging piles for charging after analysis, and the intelligent scheduling device is communicated with the charging piles through TCP; the specific method for obtaining the charging strategy through analysis by the intelligent scheduling device comprises the following steps:

step 1: clustering short-term power loads through an improved AHPK-means algorithm, predicting a typical load curve of a regional power grid, wherein the typical load curve of the power grid is a load curve except for a charging load of an electric automobile, and the method comprises the following steps:

step 1.1: selecting load data of terminal users from a database of an SCADA system of a power company in a certain city to form a load matrix A; the load data comprises the user type, the active power, the reactive power and the measured value of the voltage and the current of one branch of one transformer of the city, the sampling is carried out once every 15min, and 96 data points are sampled every day;

step 1.2: carrying out data preprocessing on the load matrix A, identifying and correcting abnormal data, and carrying out normalization processing; further comprising the steps of:

step 1.2.1: identifying and correcting abnormal data in the load matrix A; actual load data can cause the load data to be lost due to long-time fault of a certain measuring unit or fault of other related elements in the data acquisition process, for 96 data points, the records of 30 or more data points are lost, the load active recorded data is negative record, and the load active recorded data is directly kicked away;

step 1.2.2: for the sudden rising and falling data in the load curve, judging by calculating the load change rate, wherein the load change rate calculation formula is as follows:

ρ＝|(p _d -p _d-1 )/p _d |；

where ρ represents the load change rate of the user load sequence P at point d, and P = { P = { P = _d ，d＝1，…，96}，p _d Active power at point d; when the load change rate rho exceeds a preset threshold value epsilon, determining the load change rate rho as abnormal data;

and filling the abnormal data and the load data containing the missing data by adopting a smooth window type according to the following formula:

wherein p is _d ' to fill the active power at point d, a, b denote forward and backward points, respectively, with a taken at maximum ₁ 、b ₁ ；

Step 1.2.3: carrying out normalization processing on the identified and corrected load matrix A by adopting a maximum normalization method, removing a base load part of a load curve, and emphasizing the similarity of load trends, wherein a normalization expression is as follows:

x _d ＝p _d ′/max(P)；

wherein x is _d In the normalized active power of the point d, max (P) is the maximum power in the point 96;

step 1.3: performing clustering analysis on the load matrix A processed in the step 1.2 by adopting an AHPK-means algorithm, and clustering by taking a double-layer weighted Euclidean distance as a similarity criterion; further comprising the steps of:

step 1.3.1: obtaining the user category and the number of the branch, including the number h of industrial users ₁ And the number h of the resident living users ₂ The number h of school users ₃ Number of business users h ₄ Public transformer user number h ₅ (ii) a The optimized cluster number k is calculated by:

wherein f (-) is a normal function, c _i Is a constant number of times, and is,

is a transposition of a triangular matrix, h _i For the number of ith user categories, i =1,2, \8230, n, n =5, epsilon is the measurement error;

step 1.3.2: the K-means algorithm based on the double-layer weighted Euclidean distance comprises the following specific steps:

step 1.3.2.1: taking the optimized clustering number k obtained in the step 1.3.1 as an initial clustering center;

step 1.3.2.2: classifying samples; all sample centers of class k are divided into class centers with the closest weighted Euclidean distance, sample A _k To the jth cluster center m _j ＝{m _j，1 ，m _j，2 ，…，m _j，q The weighted distance of is calculated by:

step 1.3.2.3: updating a clustering center; according to the result of the step 1.3.2.2, calculating the average value of each class as a new clustering center of each class;

step 1.3.2.4: performing iterative computation; judging whether the clustering center is converged, if not, returning to the step 1.3.2.2, otherwise, executing the next step;

step 1.3.2.5: combining all the clustering centers into a new data set, regarding each clustering center as a class, and calculating the distance between all the classes, namely the similarity between samples;

step 1.3.2.6: selecting categories of which the inter-category distance meets the requirements according to a set rule to perform inter-category merging operation;

step 1.3.2.7: for class C _i And C _j Selecting the average distance between the centers of the two classes as the distance between the two classes, and then selecting the class with the minimum class distance for combination:

wherein D is _ij Is C _i And C _j The combined result, n _i 、n _j Are respectively of class C _i And C _j X, x 'are class centers, d (x, x') is the average distance between the two class centers;

step 1.3.2.8: calculating the similarity between the new class and the previous class generated in the last step;

step 1.3.2.9: repeating the steps 1.3.2.6-1.3.2.8 until all the load samples are classified into one type, and ending the algorithm;

step 1.4: testing the clustering effectiveness; evaluating a clustering result by establishing an effectiveness index, and determining an optimal clustering number to obtain a final short-term load curve;

determining the optimal clustering number and evaluating the clustering quality by adopting the Davishenburg index (DBI index I) _DBI The smaller the clustering effect, the better the clustering effect, and the calculation formula is:

wherein K is a number of clusters, R _k Is the sum of the average distances within any two classes divided by the distance between the centers of the two classes, d (X) _k ) And d (X) _j ) Is the average distance of the distances in k and j classes, X _k 、X _j Respectively representing class k and j data, d (c) _k ，c _j ) Is the distance between the centers of two clusters, c _k 、c _j Representing the centers of classes k and j, respectively;

performing cluster analysis by using AHPK-means, repeating for 20 times, and selecting I _DBI The corresponding solution at the minimum is the best clustering result;

step 2: calculating an ideal charging load curve of the electric automobile through the obtained short-term load curve, and comprising the following steps of:

step 2.1: processing the power load data matrix A through step 1.4 to obtain a short-term load curve with obvious difference in N-type centers;

step 2.2: obtaining behaviors and constitutions of the users through the obtained N types of short-term load curves, and analyzing the behaviors and constitutions of the users by combining clustering results to obtain N types of user categories;

step 2.3: respectively calculating the maximum power value P of the N user classes _max Minimum power value P _min Time period t of peak occurrence _h Time period t of valley appearance ₁ And a peak-to-valley difference Δ P, wherein the maximum power value P _max Minimum power value P _min Time period t of peak occurrence _h Time period t of valley appearance ₁ Directly from the obtained short-term load curve, the peak-to-valley difference Δ p is calculated from the following formula: Δ P = P _max -P _min ；

Step 2.4: obtaining an ideal charging load curve of the electric automobile, namely a time period t occurring in a valley through the information ₁ Charging and at a maximum power value P _max Minimum power value P _min As a limiting condition;

and 3, step 3: the method comprises the following steps of establishing an optimization equation and solving to obtain a charging strategy by taking charging demand constraint, charging time constraint, electric vehicle charging quantity limitation, a power grid power load curve and power distribution network capacity as boundary conditions, taking the minimum error of the charging load curve and an ideal charging load curve as constraint conditions and taking the minimum power consumption cost of a user as an objective function; the method specifically comprises the following steps:

step 3.1: initializing the load information of the power distribution network on the same day;

step 3.2: the charging station system judges whether a new electric vehicle drives into the charging station or not; if yes, reading all useful data information of the newly accessed electric automobile, including: acquiring the battery capacity of the electric vehicle, the current charge state of the electric vehicle battery, the expected retention time of the electric vehicle, a charging power curve of the electric vehicle battery and the expected charge state level of the electric vehicle when the electric vehicle leaves; if not, the charging mode of the last time period is continued;

step 3.3: obtaining the maximum value t of the residence time of all vehicles according to the expected residence time of the electric vehicle at the charging station in the time period _M And calculates the optimization time length T at this time,

| x | is the smallest integer less than x;

step 3.4: aiming at the optimization of the electric vehicle charging, a new charging control strategy is formulated according to the system information; establishing an optimization equation by taking charging demand constraint, charging time constraint, electric vehicle charging quantity limitation, a power grid power load curve and power distribution network capacity as boundary conditions, taking the minimum error of a charging load curve and an ideal charging load curve as constraint conditions and taking the minimum electricity consumption cost of a user as a target function;

the charging demand is constrained to

Wherein x is _n，t When the nth car is accessedActive power at a point of time t, S _n，S To an initial state of charge, b _n For charging the battery SOC value which can be increased in a period of time, the battery SOC of the charged electric automobile at least reaches the final SOC S required at the beginning of charging in T time periods _n，E While charging should be stopped in the case of full charge;

the charging time constraint is t _n，E ≤T _n，E The charged electric automobile needs to be charged within the expected retention time set by the user; wherein t is _n，E Charging end time for the nth electric vehicle; t is a unit of _n，E An expected end-of-charge time set for the electric vehicle user;

the charging quantity of the electric automobile is limited to

Wherein n is _t For the active power of the nth electric vehicle at the moment t, the number of charging piles in the charging station is limited, and the number of the charging vehicles in each time period is limited by the number of the charging piles; x is the number of the charging piles in the charging station;

the peak-to-valley difference constraint is | P _max1 -P _min1 < Δ P; wherein, P _max1 And P _min1 Respectively starting from the morning of the current day to the end of the current optimization time period, wherein the system load is the maximum value and the minimum value; combining the peak-valley difference value of the time interval in the last 7 days, determining the initial value of the delta P as the minimum value of the peak-valley difference value of the time interval in the seven days, and if the value is small and the optimization target has no solution, adding 1% to the initial value of the delta P until the solution exists;

step 3.5: the selection of the charging start time adopts a nonlinear optimization method, a nonlinear equation is solved through MATLAB, and the appropriate charging start time is searched to minimize the fluctuation of the total load curve as an optimized objective function, wherein the specific method comprises the following steps:

step 3.5.1: the load variance is minimum in T time; after a charging load is connected, if charging is started at a certain moment, so that the variance of a total load curve added with the load in a T time period is minimum, the moment is the selected charging starting time; for the continuous function, the 2 nd order center distance representation variance is calculated

Wherein, P _var For the daily load variance, P, of the distribution network _av To adjust the daily average load, P _Et Is the total charging power, P, of the electric vehicle in the period of t _Lt The conventional load of the power distribution network in the time period t when the power distribution network does not contain the charging load of the electric automobile is obtained through load prediction;

step 3.5.2: combining the condition of time-of-use electricity price, and taking the lowest electricity cost of the user as an objective function, namely

Wherein S is _j Representing the electricity price of the power grid at the moment j, wherein the positive value represents the charging electricity price of the electric automobile, and the negative value represents the subsidy electricity price of the electric automobile user for feeding the power grid; p _ij The representation represents the required power of i automobiles in the j time period;

step 3.5.3: and converting the optimization problem of the 2 objective functions into single-objective optimization by adopting a linear weighted sum method, and simultaneously normalizing the functions respectively due to different dimensions of the two objective functions, wherein the converted objective functions are shown as the following formula:

min T ₁ ＝ω ₁ (T ₃ /T _3max )+ω ₂ (T ₄ /T _4max )；

in the formula, T _3max The mean square error of the original power grid load curve is obtained; t is _4max The daily charging cost of the vehicle owner is used for the traditional vehicle usage, namely the cost required by fully charging the battery from the lowest electric quantity in the stop time; t is ₃ 、T ₄ Representing two objective functions, ω, in step 3.3.1 and step 3.3.2, respectively ₁ 、ω ₂ Are respectively T ₃ 、T ₄ And satisfies ω ₁ +ω ₂ ＝1；

Step 3.6: respectively using the maximum power value P in the N user categories obtained in the step 2.3 _max Minimum power value P _min Time period t of peak occurrence _h Time period t of valley appearance _l And the peak-to-valley difference delta p is brought into the optimization equation established in the step 3.5, similarity judgment is respectively carried out on the real-time data of the power grid and the obtained data of the N typical load curves, and the ideal charging load curve is judged if an electric vehicle is accessed, so that the optimal charging strategy of each user category is obtained;

the similarity judgment formula is shown as follows:

wherein, K (x) ^new Is the real-time data of the current day,

q =1,2, \ 8230;, N;

and 4, step 4: judging whether the electric vehicle is charged or not according to a decision result of the dynamic game, namely judging whether the obtained optimal charging strategy has adjustability or not, wherein the specific method comprises the following steps of:

step 4.1: when a new electric vehicle is accessed into any one charging pile loaded in a certain power grid branch, the charging station operation system automatically updates system information to the next 15min control time point according to the operation state of the charging pile in the station, and firstly calculates the SOC state, the power required when the electric vehicle is fully charged and the number J of required charging time segments corresponding to the electric vehicle _i Number of parking time segments T _i And T of the system for leaving the vehicle from the current time period _i Charging load margin M in each time period _t ，M _t ＝A _t S _T λ, where T =1,2, \ 8230;, T _i ，T _i ＝96；S _T The rated capacity of the transformer is represented; a. The _t The proportion of power which represents that a charging station is allowed to charge the electric vehicle in the t-th time period in the day accounts for the capacity of the transformer; λ is the power factor of the charging load;

and 4.2: before a preferential time-of-use electricity price period facing the user is set, the system judges whether the charging requirement of the electric automobile can be met or not in advance, and the T of leaving the automobile from the current period is calculated by the computing system _i Charging load margin M in each time period _t The realization is as follows: h _i ＝|A _i |，A _i ＝{t|M _t ≥P _i ，t＝1，2，...，T _i }; wherein, | A _i Is set A _i The number of the elements in the solution;

step 4.3: if it corresponds to M of the time interval _t ≥P _i +P ₀ If not, adopting an optimization algorithm to reasonably arrange whether each automobile is charged; p _i Sum of the active power required for each charging pile when there is an electric vehicle access, P ₀ For the current real-time active power, M _t A larger value indicates a larger charging load margin for the corresponding period;

step 4.4: if H is _i ＜J _i ，J _i The number of the time segments indicating the full charge indicates that the system cannot meet the charging requirements input by all users in the parking time of the electric automobile, so that the electric automobile is charged by adopting an optimization algorithm to arrange the charging load margin M _t As a constraint condition, if the charging power required by a certain electric vehicle exceeds 80% of the SOC, the charging power is distributed according to a chi priority parameter, namely P _i ＝χP _i Then at max M _t ＝αP ₁ +βP ₂ +...+φP _n Is solved as an objective equation, where P _i The configured charging power of the electric automobile is alpha, beta, and phi are specific gravity coefficients, the number of the electric automobiles meeting the optimal charging requirement is obtained, and the charging piles can be known to charge the electric automobiles, and the power of the electric automobiles required by a user is displayed;

step 4.5: analyzing the result of the step 4.4, if the result can not be the electric automobileCharging, prompting the user to leave, if charging can be carried out on the electric automobile, prompting the user to charge the electric automobile, and calculating the state of charge of the electric automobile battery which is maximally met by the system when the user leaves:

wherein the content of the first and second substances,

representing the initial battery capacity, eta is the power parameter, Δ t is the length of time from the start of charging to the departure, B _i The capacity of the electric steam battery;

because the electric automobile needs to be charged all the time in the parking time, the charging station prompts the user that the state of charge of the battery which can be met by the system to the maximum extent

And prompts the user to press the peak price p _h Charging, wherein a user autonomously selects to receive charging service or give up charging service; if the user receives the charging service, arranging the electric automobile to be in the charging load margin M _t Greater than P _i +P ₀ Is charged and the charging load margin A is set in the corresponding time interval _t S _T λ is updated to A _t S _T λ-P _i ；

Step 4.6: if H _i ≥J _i If so, the system can meet the charging requirements of all the electric automobiles within the parking time of the electric automobiles, and peak clipping and valley filling are realized to the maximum extent on the basis of meeting the charging requirements of users; thus, the charging station ordered charge coordination system initially selects consecutive J _i The minimum value of the starting time interval with the maximum sum of the load margins is the starting time interval of the low-valley electricity price; specifically, the start period of the valley electricity price is preliminarily calculated according to the following expression:

consider the following

Beginning of the time period until the electric vehicle leaves, and there is J _i Charging load margin M corresponding to each time period _t ≥P _i +P ₀ In the case of (1), the off-peak electricity price initial period is adjusted to be J _i Charging load margin M corresponding to time period _t ≥P _i +P ₀ To ensure that the system can charge the SOC of the electric vehicle to a desired level after the user responds to the off-peak electricity prices; specifically, the start period of the valley electricity prices is adjusted according to the following expression:

wherein the content of the first and second substances,

n is an integer set;

step 4.7: at the initial time of the adjusted valley price obtained by calculation

Then, determining the time period

The charging price of the inner user is the valley electricity price p _l The charging electricity price in other time periods is the peak electricity price p _h The charging is started immediately or delayed to the valley power price by the autonomous selection of the user; wherein D is a time parameter and is 0-1;

if the user selects to start charging immediately, the electric vehicle is arranged to start charging in the next time period and the corresponding charging load margin M is arranged one by one _t ≥P _i +P ₀ Until schedule J _i Until a time period; simultaneously charging load margin A corresponding to charging period _t S _T λ is updated to A _t S _T λ-P _i ；

If the user selects to delay the charging to the valley power price, the electric automobile is arranged to start charging from

Initially, one by one at the corresponding charging load margin M _t ≥P _i +P ₀ Until schedule J _i Until a time period; simultaneously charging load margin A corresponding to charging period _t S _T λ is updated to A _t S _T λ-P _i 。

Further, the number N =4,4 user categories obtained in step 2.2 are:

(1) Late peak user group: the load change trend of a user accords with the form of a common daily load curve, namely, the load continuously rises from 5 am, the load is kept at a higher level to 9 am, the power consumption is reduced a little at noon 12 noon break, the power consumption continuously rises from 3 pm to 8 pm and reaches the highest peak of the power consumption all day long, and the power consumption slowly falls;

(2) Unimodal user group: the load change trend of the user has an electricity consumption peak, and in the overall trend, the amplitude difference of the electricity consumption power corresponding to the peak and the valley is not large, and the stable electricity consumption is continuously carried out from 5 am to 20 pm;

(3) Smooth user group: the load change trend of users is relatively smooth, the load change in one day is not large, but the active power level of the electric load of a smooth user group is very high, and the load is always at a high level;

(4) Avoiding peak type user groups: the load change trend of the user is very obvious from that of other users, the electricity utilization characteristics of the user are obviously different from those of other four types, the user presents a peak avoiding type characteristic that the two ends are high and the middle is low, electricity utilization is stopped at about 6 points in the morning, the electricity utilization is quickly reduced and kept at a lower level until 6 points in the evening, the rest time is in the electricity utilization peak period, and the peak value and the peak valley difference are also the highest in all types.

Adopt the produced beneficial effect of above-mentioned technical scheme to lie in: the typical load urban large-scale electric vehicle charging method based on the dynamic game can effectively reduce the peak-valley difference of the load of the power grid, improve the load characteristic of the power grid and improve the load rate of power equipment. Compared with the prior art, the method has the following advantages:

(1) The power load data used by the method is all load data of a certain area, and the charging optimization of the electric automobile in the area is more comprehensive, the precision is higher and the effect is better;

(2) The AHPK-means clustering algorithm has higher scalability and high efficiency, and the algorithm is more accurate and stable;

(3) In the process of solving the optimization equation, more limiting conditions are adopted, so that the obtained optimization strategy is more accurate;

(4) By solving the nonlinear equation, the total load curve fluctuation is minimized, and three optimization objective functions are adopted, so that the optimization is more comprehensive.

Drawings

Fig. 1 is a general flowchart of a charging method for a large-scale electric vehicle in a typical load city based on dynamic gaming according to an embodiment of the present invention;

FIG. 2 is a flowchart of clustering performed by the AHPK-means clustering algorithm provided in the embodiment of the present invention;

FIG. 3 is a diagram of the clustering effect of the AHPK-means clustering algorithm provided by the embodiment of the present invention;

fig. 4 is a graph of active power variation of electricity consumption generated by a late peak user group during a day according to an embodiment of the present invention;

fig. 5 is a graph of active power variation of power consumption generated by a single peak user group in a day according to an embodiment of the present invention;

fig. 6 is a graph of active power variation of electricity consumption generated by a stable user group in a day according to an embodiment of the present invention;

fig. 7 is a diagram of active power variation of power consumption generated by a peak avoidance type user group in a day according to an embodiment of the present invention;

FIG. 8 is a block diagram of an electric vehicle charging system according to an embodiment of the present invention;

fig. 9 is a schematic diagram of a monte carlo simulation method according to an embodiment of the present invention.

Detailed Description

The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.

To study the load composition of an area, it is necessary to know the real-time load data of all users, and the topological correlation information of each load line or private or public change in the area, and to specify to which class the various encoded officer attributes belong. The information such as power grid topology, user load, load property, power grid structure and the like can be acquired respectively through the power grid SCADA system and the information acquisition system, so that the proportion of loads such as industry, agriculture, commerce, residents and the like can be counted, and structural decomposition and structural analysis of the load of the whole system are realized. And the nodes with the same attribute are merged by utilizing a partition strategy, so that the analysis of the load structure of the whole area system is formed, the quantitative analysis of the load composition is realized, and the characteristics of the area load are mastered.

The existing load component composition refers to the proportion of loads with different properties in the total load, and the total classification of the loads is distinguished according to industrial, agricultural, commercial, residential and other loads, and specifically comprises the following steps:

(1) Industrial loads, mainly production, processing, manufacturing enterprises, such as mining, food processing, tobacco, textile, wood, such as industrial, furniture, paper, printing, chemical, petrochemical, and the like;

(2) Agricultural load mainly refers to farmland irrigation and the like;

(3) The commercial load mainly refers to an organization used for business, such as a commercial service organization, some enterprises and public institutions, schools, hospitals, governments and the like;

(4) The resident load means the electricity utilization in the aspects of living and rest, such as an apartment and the like;

(5) Other loads include power for street or highway lighting, power for electrified railways or subways, and power for auxiliary equipment in power plants.

As shown in fig. 1, the embodiment provides a city electric vehicle charging decision method based on a typical load dynamic game, which is implemented by installing a general intelligent scheduling device on a transformer branch, wherein the intelligent scheduling device controls n charging piles of the transformer branch, when an electric vehicle is accessed, the charging piles upload electric vehicle information and send the information to the intelligent scheduling device, the intelligent scheduling device sends a charging strategy to the charging piles for charging after analysis, and the intelligent scheduling device communicates with the charging piles through a TCP; the specific method for obtaining the charging strategy through analysis by the intelligent scheduling device is as follows.

Step 1: clustering short-term power loads through an improved AHPK-means algorithm, predicting a typical load curve of a regional power grid, wherein the typical load curve of the power grid is a load curve except for a charging load of an electric automobile, and the method comprises the following steps:

step 1.1: selecting load data of terminal users from a database of an SCADA system of a power company in a certain city to form a load matrix A; the load data comprises the user type, the active power, the reactive power and the measured value of the voltage and the current of one branch of one transformer of the city, the sampling is carried out once every 15min, and 96 data points are sampled every day;

in the embodiment, certain working day data of 7 months in 2018 of a certain city, including measured values of active power, reactive power, voltage and current and the like of users in the direct genus and the subordinate 7 counties and districts of a certain power company, the daily load curves of 2567 users are actually measured as research objects, the daily load sampling interval is 15min, 96 sampling points are counted daily, 8563 effective daily load curves are contained, and a 8563 × 96-order matrix A is formed;

step 1.2: carrying out data preprocessing on the load matrix A, identifying and correcting abnormal data, and carrying out normalization processing; further comprising the steps of:

step 1.2.1: identifying and correcting abnormal data in the load matrix A; the actual load data can cause the load data to be lost due to long-time fault of a certain measuring unit or fault of other related elements in the data acquisition process, and for 96 data points, records of 30 or more data points and load active record data are lost and are negative records, and the data are directly kicked out;

step 1.2.2: for the sudden rising and falling data in the load curve, judging by calculating the load change rate, wherein the load change rate calculation formula is as follows:

ρ＝|(p _d -p _d-1 )/p _d |；

wherein, rho represents the load change rate of the user load sequence P at the point d, P = { P _d ，d＝1，…，96}，p _d Active power at point d; when the load change rate rho exceeds a preset threshold value epsilon, the data are regarded as abnormal data;

and filling the abnormal data and the load data containing the missing data by adopting a smooth window type according to the following formula:

wherein p is _d ' is the active power of d points after filling, a, b respectively represent forward and backward points, and a is taken to the maximum ₁ 、b ₁ (ii) a In this example, a ₁ 、b ₁ May be 3, 4 or 5, respectively;

step 1.2.3: normalizing the identified and corrected load matrix A, removing the base load part of the load curve, emphasizing the similarity of the load trend, and specifically adopting a maximum value normalization method, wherein the expression is as follows:

x _d ＝p _d ′/max(P)；

wherein x is _d In the normalized active power of the point d, max (P) is the maximum power in the point 96;

step 1.3: performing clustering analysis on the load matrix A processed in the step 1.2 by adopting an AHPK-means algorithm, and clustering by taking a double-layer weighted Euclidean distance as a similarity criterion; as shown in fig. 2, the clustering process further includes the following steps:

step 1.3.1: obtaining the user category and the number of the branch, including the number h of industrial users ₁ The number h of resident life users ₂ The number h of school users ₃ Business, business and deliveryNumber of business users h ₄ Public transformer user number h ₅ (ii) a The optimized cluster number k is calculated by:

wherein f (-) is a normal function, c _i Is a constant number of times, and is,

is a transposition of a triangular matrix, h _i For the number of ith user categories, i =1,2, \8230, n, n =5, epsilon is the measurement error;

step 1.3.2: the K-means algorithm based on the double-layer weighted Euclidean distance comprises the following specific steps:

step 1.3.2.1: taking the optimized clustering number k obtained in the step 1.3.1 as an initial clustering center;

step 1.3.2.2: classifying samples; all sample centers of class k are divided into class centers with the closest weighted Euclidean distance, sample A _k To the jth cluster center m _j ＝{m _j，1 ，m _j，2 ，…，m _j，q The weighted distance of is calculated by:

step 1.3.2.3: updating a clustering center; according to the result of the step 1.3.2.2, calculating the average value of each class as a new clustering center of each class;

step 1.3.2.4: performing iterative computation; judging whether the clustering center is converged, if not, returning to the step 1.3.2.2, otherwise, executing the next step;

step 1.3.2.5: combining all the clustering centers into a new data set, regarding each clustering center as a class, and calculating the distance between all the classes, namely the similarity between samples;

step 1.3.2.6: selecting categories of which the inter-category distance meets the requirements according to a set rule to carry out inter-category merging operation;

step 1.3.2.7: for class C _i And C _j Selecting the average distance between the centers of the two classes as the distance between the two classes, and then selecting the class with the minimum class distance for combination:

wherein D is _ij Is C _i And C _j The combined result, n _i 、n _j Are respectively of class C _i And C _j X, x 'are class centers, d (x, x') is the average distance between the two class centers;

step 1.3.2.8: calculating the similarity between the new class and the previous class generated in the last step;

step 1.3.2.9: repeating the steps 1.3.2.6-1.3.2.8 until all the load samples are classified into one type, and ending the algorithm;

step 1.4: checking the clustering validity; evaluating a clustering result by establishing an effectiveness index, and determining the optimal clustering number to obtain a final short-term load curve;

and determining the optimal clustering number and evaluating the clustering quality by using the Davison baudin index, namely the DBI index. The DBI index is one of the commonly used effectiveness indexes in the existing clustering method, and the purpose of load curve clustering is to obtain different typical load curves, so that each type of curve has a similar mode and reflects the same type of user electricity utilization characteristics.

DBI index comprehensively considers the dispersion between classes and the compactness in the class, and the DBI index I _DBI The smaller the cluster effect, the better the calculation formula:

wherein K is a cluster number, R _k Is the sum of the mean distances between any two classes of intra-class distances divided by the distance between the centers of the two clusters, d (X) _k ) And d (X) _j ) Is the average distance of the distances within the k and j classes, X _k 、X _j Respectively representing class k and j data, d (c) _k ，c _j ) Is the distance between the centers of two clusters, c _k 、c _j Representing the centers of classes k and j, respectively;

performing cluster analysis by using AHPK-means, repeating for 20 times, and selecting I _DBI The corresponding solution at the minimum is the best clustering result;

the resulting clustering effect is shown in fig. 3.

Step 2: calculating an ideal charging load curve of the electric automobile through the obtained short-term load curve, and comprising the following steps of:

step 2.1: processing the power load data matrix A through step 1.4 to obtain a short-term load curve with obvious difference in N-type centers;

step 2.2: obtaining the behaviors and the constitutions of the users through the obtained N types of short-term load curves, and analyzing the behaviors and the constitutions of the users by combining clustering results to obtain N types of users;

in this embodiment, 4 user categories are obtained, which are:

(1) Late peak user group: the load change trend of a user accords with the form of a common daily load curve, namely, the load continuously rises from 5 am, the load is kept at a higher level from 9 am, the power consumption is reduced a little at noon 12 noon break, the power consumption continuously rises from 3 pm to 8 pm, the highest peak of the power consumption in the whole day is reached, and the power consumption slowly falls; the active power change of electricity generated by this type of user in a day is shown in fig. 4;

(2) Unimodal user group: the load change trend of the user has an electricity consumption peak, and in the overall trend, the amplitude difference of the electricity consumption power corresponding to the peak and the valley is not large, and the stable electricity consumption is continuously carried out from 5 am to 20 pm; the active power change of electricity generated by this type of user in a day is shown in fig. 5;

(3) Smooth user group: the load change trend of the user is relatively smooth, the load change in one day is not large, but the active power level of the electric load of the smooth user group is very high, and the load is always at a high level; the active power change of electricity generated by this type of user in a day is shown in fig. 6;

(4) Avoiding peak user groups: the load change trend of the user is very obvious from other users, the electricity utilization characteristics of the user are obviously different from other four types, the user presents a peak avoiding type characteristic that two ends are high and the middle is low, electricity utilization is stopped at about 6 points in the morning, the electricity utilization is rapidly reduced and kept at a lower level until 6 points in the evening, the rest time is in the electricity utilization peak period, and the peak value and peak valley difference are also the highest in all the types; the active power change of electricity generated by this type of user in a day is shown in fig. 7;

step 2.3: respectively calculating the maximum power value P of the N user classes _max Minimum power value P _min Time period t of peak occurrence _h Time period t of valley appearance _l And a peak-to-valley difference Δ P, wherein the maximum power value P _max Minimum power value P _min Time period t of peak occurrence _h Time period t of valley appearance _l Directly from the obtained short-term load curve, the peak-to-valley difference Δ p is calculated from the following formula:

Δp＝P _max -P _min ；

step 2.4: the ideal charging load curve of the electric automobile is obtained through the information, namely the electric automobile is charged in the time period t1 when the valley occurs and is charged at the maximum power value P _max Minimum power value P _min As a limiting condition.

And 3, step 3: the method comprises the steps of establishing an optimization equation and solving to obtain a charging strategy by taking charging demand constraint, charging time constraint, electric vehicle charging quantity limitation, a power grid power load curve and power distribution network capacity as boundary conditions, taking the minimum error between a charging load curve and an ideal charging load curve as constraint conditions and taking the lowest electricity consumption cost of a user as an objective function; the method specifically comprises the following steps:

step 3.1: initializing the load information of the power distribution network on the same day;

step 3.2: the charging station system judges whether a new electric vehicle drives into the charging station or not; if yes, reading all useful data information of the newly accessed electric automobile, including: acquiring the capacity of an electric vehicle battery, the current charge state of the electric vehicle battery, the expected residence time of the electric vehicle, the charging power curve of the electric vehicle battery and the expected charge state level of the electric vehicle when the electric vehicle leaves; if not, the charging mode of the last time period is continued;

step 3.3: obtaining the maximum value t of the residence time of all vehicles according to the expected residence time of the electric vehicle at the charging station in the time period _M And calculates the optimization time length T of this time,

| x | is the smallest integer less than x;

step 3.4: aiming at the optimization of the electric vehicle charging, a new charging control strategy is formulated according to the system information; establishing an optimization equation by taking a charging demand constraint, a charging time constraint, an electric vehicle charging quantity limitation, a power grid power load curve and a power distribution network capacity as boundary conditions, taking a charging load curve and an ideal charging load curve with the minimum error as constraint conditions and taking the lowest power consumption cost of a user as a target function;

the charging demand is constrained to

Wherein x is _n，t Active power at a point of time t when the nth car is accessed, S _n，S To an initial state of charge, b _n For charging the battery SOC value which can be increased in a period of time, the battery SOC of the charged electric automobile at least reaches the final SOC S required at the beginning of charging in T time periods _n，E While charging should be stopped in the case of full charge;

the charging time constraint is

t _n，E ≤T _n，E ；

The charged electric automobile needs to be charged within the expected residence time set by a user; wherein, t _n，E Charging end time for the nth electric vehicle; t is a unit of _n，E An expected end-of-charge time set for the electric vehicle user;

the charging quantity of the electric automobile is limited to

Wherein n is _t For the active power of the nth electric vehicle at the moment t, the number of charging piles in the charging station is limited, and the number of the charging vehicles in each time period is limited by the number of the charging piles; x is the number of charging piles in the charging station;

the peak-to-valley difference constraint is

|P _max1 -P _min1 |＜ΔP；

Wherein, P _max1 And P _min1 Respectively starting from the morning of the current day to the end of the current optimization time period, and then respectively determining the maximum value and the minimum value of the system load in the time period; combining the peak-valley difference value of the time interval in the last 7 days, and determining the initial value of the delta P as the minimum value of the peak-valley difference value of the time interval in the seven days, wherein the value may cause no solution of the optimization strategy due to small deviation, and if no solution exists, the initial value of the delta P is increased by 1% until a solution exists;

step 3.5: the selection of the charging start time adopts a nonlinear optimization method, a nonlinear equation is solved through MATLAB, and the appropriate charging start time is searched to minimize the fluctuation of the total load curve as an optimized objective function, wherein the specific method comprises the following steps:

step 3.5.1: the load variance is minimum in T time; after a charging load is connected, if charging is started at a certain moment, so that the variance of a total load curve added with the load in a T time period is minimum, the moment is the selected charging starting time; for the continuous function, the 2 nd order center distance representation variance is calculated

Wherein, P _var For the daily load variance, P, of the distribution network _av To adjust the daily average load, P _Et Is the total charging power, P, of the electric vehicle in the period of t _model The typical load obtained in the step 1 in the period t does not contain the charging load of the electric automobile;

step 3.5.2: combining the time-of-use electricity price condition, and taking the lowest electricity cost of the user as an objective function, namely

Wherein x is _n，t Active power at a certain point of time t when the nth automobile is accessed, S _n，s To an initial state of charge, b _n Increasing SOC value of a battery for a period of time for charging, S _j Representing the electricity price of the power grid at the moment j, wherein the positive value represents the charging electricity price of the electric automobile, and the negative value represents the subsidy electricity price of the electric automobile user for feeding electricity to the power grid;

step 3.5.3: the 2 objective functions are mutually influenced, and in order to realize the comprehensive optimization of the two objective functions, a linear weighted sum method is adopted for processing, namely, a multi-objective optimization problem is converted into single-objective optimization; meanwhile, because the dimensions of the two objective functions are different, the two objective functions are normalized respectively, and the converted objective functions are shown as follows:

min T ₁ ＝ω ₁ (T ₃ /T _3max )+ω ₂ (T ₄ /T _4max )；

in the formula, T _3max The mean square error of the original power grid load curve is obtained; t is _4max The daily charging cost of the vehicle owner is used for the traditional vehicle usage, namely the cost required by fully charging the battery from the lowest electric quantity in the stop time; t is ₃ 、T ₄ Each representsTwo objective functions, ω, in step 3.3.1 and step 3.3.2 ₁ 、ω ₂ Are respectively T ₃ 、T ₄ And satisfies the weight coefficient of ω ₁ +ω ₂ ＝1；

Step 3.6: respectively using the maximum power value P in the N user categories obtained in the step 2.3 _max Minimum power value P _min Time period t of peak occurrence _h Time period t of valley appearance _l And the peak-to-valley difference delta p is brought into the optimization equation established in the step 3.5, similarity judgment is respectively carried out on the real-time data of the power grid and the obtained data of the N typical load curves, and the ideal charging load curve is judged if an electric vehicle is accessed, so that the optimal charging strategy of each user category is obtained;

the similarity judgment formula is shown as follows:

wherein, K (x) ^new Is the real-time data of the current day,

for the qth exemplary payload data, q =1,2, \8230;, N.

And 4, step 4: judging whether the electric vehicle is charged or not according to a decision result of the dynamic game, namely judging whether the obtained optimal charging strategy has adjustability or not, wherein the specific method comprises the following steps of:

step 4.1: when a new electric vehicle is accessed into any one charging pile loaded in a certain power grid branch, the charging station operation system automatically updates the system information to the next 15min control time point according to the running state of the charging pile in the station, and firstly calculates the SOC state, the power required when the charging is fully performed and the number J of charging time segments required by the electric vehicle _i Number of parking time segments T _i And T of the system at the departure of the vehicle from the current time period _i Charging load margin M in each time period _t ，M _t ＝A _t S _T λ, wherein t =1,2,…，T _i ，T _i ＝96；S _T representing the rated capacity of the transformer; a. The _t The proportion of power which allows the charging station to charge the electric vehicle in the t-th time period in the day to the capacity of the transformer is represented; λ is the power factor of the charging load;

step 4.2: before the preferential time-of-use electricity price time period for the user is formulated, the system judges whether the charging requirement of the electric automobile can be met in advance, and the T of leaving the automobile is calculated from the current time period by the calculation system _i Charging load margin M in each time period _t The realization is as follows:

H _i ＝|A _i |，A _i ＝{t|M _t ≥P _i ，t＝1，2，...，T _i }；

wherein, | A _i Is set A _i The number of the elements in the solution;

step 4.3: if it corresponds to M of the time interval _t ≥P _i +P ₀ If not, adopting an optimization algorithm to reasonably arrange whether each automobile is charged; p _i Sum of the active power required for each charging pile when there is an electric vehicle access, P ₀ For the current real-time active power, M _t A larger value indicates a larger charging load margin for the corresponding period;

step 4.4: if H is _i ＜J _i ，J _i The number of the time segments indicating the full charge indicates that the system cannot meet the charging requirements input by all users in the parking time of the electric automobile, so that the electric automobile is charged by adopting an optimization algorithm to arrange the charging load margin M _t As a constraint condition, if the charging power required by a certain electric vehicle exceeds 80% of the SOC, the charging power is distributed according to a chi priority parameter, namely P _i ＝χP _i Then at max M _t ＝αP ₁ +βP ₂ +...+φP _n Is solved as an objective equation, where P _i The configured charging power of the electric automobile, alpha, beta, eta and phi are specific gravity coefficients, the number of the electric automobiles meeting the optimal charging requirement is obtained, and the charging piles can be used for charging the electric automobiles, and the same time, the charging piles can be used for charging the electric automobilesThe time display shows how much power the user should charge;

step 4.5: and 4.4, analyzing the result of the step 4.4, prompting the user to leave if the electric automobile cannot be charged, prompting the user to charge the electric automobile if the electric automobile can be charged, and calculating the state of charge of the battery of the electric automobile, which is maximally met by the system when the user leaves:

wherein the content of the first and second substances,

representing the initial battery capacity, η is the power parameter, Δ t is the length of time from start of charge to departure, B _i Is the capacity of the electric steam battery;

because the electric automobile needs to be charged all the time in the parking time, the charging station prompts the user that the state of charge of the battery which can be met by the system to the maximum extent

And prompts the user to press the peak price p _h Charging, wherein a user autonomously selects to receive charging service or give up charging service; if the user receives the charging service, arranging the electric automobile to be in the charging load margin M _t Greater than P _i +P ₀ Is charged and the charging load margin A is set in the corresponding time interval _t S _T λ is updated to A _t S _T λ-P _i ；

Step 4.6: if H _i ≥J _i If so, the system can meet the charging requirements of all the electric automobiles within the parking time of the electric automobiles, and peak clipping and valley filling are realized to the maximum extent on the basis of meeting the charging requirements of users; thus, the charging station ordered charging coordination system initially selects consecutive J _i The minimum value of the initial time interval with the maximum sum of the load margins is the initial time interval of the low-valley electricity price; specifically, the start time of the valley electricity rate is preliminarily calculated according to the following expressionSection (2):

consider from

Beginning of the time period until the electric vehicle leaves, and there is J _i Charging load margin M corresponding to each time period _t ≥P _i +P ₀ In the case of (1), the off-peak electricity price initial period is adjusted to be J _i Charging load margin M corresponding to time period _t ≥P _i +P ₀ To ensure that the system can charge the SOC of the electric vehicle to a desired level after the user responds to the valley price; specifically, the start period of the valley electricity prices is adjusted according to the following expression:

wherein the content of the first and second substances,

n is an integer set;

step 4.7: at the initial time of the calculated adjusted off-peak electricity price

Then, determining the time period

The charging price of the inner user is the valley electricity price p _l The charging electricity price in other time periods is the peak electricity price p _h The user is prompted to select to start charging immediately or delay to the valley power price for starting charging; wherein D is a time parameter and is 0-1;

if the user selects to start charging immediately, the electric automobile is arranged to start charging in the corresponding charging load margin M one by one from the next time period _t ≥P _i Until schedule J _i Until a time period; simultaneously charging load margin A corresponding to charging period _t S _T λ is updated to A _t S _T λ-P _i ；

If the user selects to delay the charging to the valley price, the electric vehicle is arranged to start charging

At the beginning, one by one at the corresponding charging load margin M _t ≥P _i +P ₀ Until schedule J _i Until a time period; simultaneously charging load margin A corresponding to charging period _t S _T λ is updated to A _t S _T λ-P _i 。

After the charging strategy is obtained by the method of this embodiment, the charging strategy is completed by the electric vehicle charging system shown in fig. 8, where the system includes a transformer (a current transformer and a voltage transformer), an intelligent electric meter, an intelligent charging control device, a charging pile, and a distribution transformer monitoring terminal. The transformers (current transformers and voltage transformers) refer to current transformers and voltage transformers which are arranged on the low-voltage side of a distribution transformer of a residential area, and the output ends of the current transformers and the voltage transformers are connected with the input end of a monitoring terminal of the distribution transformer; the intelligent charging control device is communicated with the distribution transformer monitoring terminal in an RS485 or RS232 mode, an optimized charging plan of each charging pile is formulated and issued to each charging pile, the charging state of each charging pile is monitored, the charging load of the charging pile is stored at regular intervals, and the obtained charging load of the charging pile is transmitted to the distribution transformer monitoring terminal; the charging pile and the intelligent charging control device of the electric automobile are communicated in a power line carrier mode; the distribution transformer monitoring terminal acquires the voltage, the current and the like of the low-voltage side of the distribution transformer in a certain area, calculates all the power loads and load curves except for the electric automobile, and sends the power loads and the load curves to the intelligent charging control device.

In the implementation of the method of the invention, the following reasonable simplifications and assumptions are made:

(1) Assuming that all users of the electric vehicle in the area choose to stay the electric vehicle in a charging station for charging, and the parking and charging places of the electric vehicle are fixed;

(2) A dispatching system is arranged in the charging station, and when a new electric vehicle drives into the charging station and is connected into the charging pile, the system can automatically read the battery capacity b of the electric vehicle _n And initial state of charge S of electric vehicle _n，S And the like, and meanwhile, in order to make a corresponding optimization strategy, the user of the electric automobile needs to input the expected stay time T of the electric automobile into the system _n，E And the electric vehicle charge state S expected to be reached by the user when the electric vehicle leaves the charging station _n，E ；

(3) A simulation analysis method based on Monte Carlo simulation; in fact, many useful information such as the time when the electric vehicle enters the charging station, the initial state of charge, the expected residence time, etc. cannot be predicted, and for the convenience of calculation, the present embodiment randomly generates more required electric vehicle charging data by using the monte carlo method, which is shown in fig. 9.

The monte carlo method is a method for solving physical and mathematical problems using repeated statistical experiments. When the Monte Carlo method is used for processing the problem, the solution is often constructed into the mathematical expectation of a certain random variable, the arithmetic mean value of the specific values of the random variable is obtained by a hypothetical test of a certain number on a computer, and the arithmetic mean value is used as the approximate solution of the problem.

1. Determining the random number, generating the random number by adopting a mixed congruence method, and carrying out randomness test to obtain a relatively real random number, wherein the recursion formula of the mixed congruence method is x _i ＝mod(Ax _i-1 + C, M); in the formula, mod is a remainder function, A is a multiplier, M is a modulus, C is an increment, all of the three are positive integers, and the initial value is x ₀ Then x can be recurred ₁ ，x ₂ ，…，x _n . Dividing this sequence of values by M, we obtain [0,1 ]]Random number sequence r with uniformly distributed intervals _i . For the randomness test of random number sequence, the N random numbers which appear in sequence are divided into two types or k types according to the sizes, and whether the occurrence of each type of numbers is coherent or not is checked, thereby determining the real random number sequenceA random number.

2. The method for calculating the charging load of the electric automobile comprises the steps of firstly obtaining the charging load of various electric automobiles, and sequentially accumulating to obtain a total charging load curve. The unit of charging load calculation is day, and the day is divided into 1440 minutes, and when i minutes is reached, the total charging load is all vehicles, and then the charging load can be expressed as follows:

in the formula, L _i I =1,2, \ 8230;, 1440; n is the total number of the electric automobiles; p is _n，i Charging power for the nth vehicle in the ith minute.

3. The difficulty of calculating the charging load of the electric vehicle is the randomness of the initial charging time and the initial SOC. Assuming that the power grid cannot play a role in determining the charging behavior of the electric automobile, the charging is started after the power grid is connected, and the initial charging time and the initial SOC of each automobile are extracted. The initial charging time is different according to different vehicle types, and the initial SOC accords with normal distribution. After charging load models and parameters of different vehicle types are determined, rapid charging and conventional charging are required to be modeled respectively, and system input information comprises the following contents, such as the total amount of electric vehicles, the occurrence rate of different charging behaviors, possible charging time periods, charging duration constraints and the like, and initial SOC probability distribution corresponding to different types of charging behaviors is included.

Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit of the corresponding technical solutions and scope of the present invention as defined in the appended claims.

Claims (6)

1. A city electric vehicle charging decision method based on a typical load dynamic game is characterized by comprising the following steps: the method is realized by installing a general intelligent scheduling device on a transformer branch, wherein the intelligent scheduling device controls n charging piles of the transformer branch, when an electric automobile is connected, the charging piles upload electric automobile information and send the electric automobile information to the intelligent scheduling device, the intelligent scheduling device sends a charging strategy to the charging piles for charging after analysis, and the intelligent scheduling device is communicated with the charging piles through TCP; the specific method for obtaining the charging strategy through analysis by the intelligent scheduling device comprises the following steps:

step 1: clustering short-term power loads through an improved AHPK-means algorithm, predicting a typical load curve of a regional power grid, wherein the typical load curve of the power grid is a load curve except for a charging load of an electric automobile, and the method comprises the following steps:

step 1.1: selecting load data of terminal users from a database of an SCADA system of a power company in a certain city to form a load matrix A; the load data comprises the user type, the active power, the reactive power and the measured value of the voltage and the current of one branch of one transformer of the city, the sampling is carried out once every 15min, and 96 data points are sampled every day;

step 1.2: carrying out data preprocessing on the load matrix A, identifying and correcting abnormal data, and carrying out normalization processing;

step 1.3: performing clustering analysis on the load matrix A processed in the step 1.2 by adopting an AHPK-means algorithm, and clustering by taking a double-layer weighted Euclidean distance as a similarity criterion;

step 1.4: checking the clustering validity; evaluating a clustering result by establishing an effectiveness index, and determining an optimal clustering number to obtain a final short-term load curve;

and 2, step: calculating an ideal charging load curve of the electric automobile through the obtained short-term load curve, and comprising the following steps of:

step 2.1: processing the power load data matrix A through the step 1.4 to obtain a short-term load curve with obvious difference in N types of centers;

step 2.2: obtaining the behaviors and the constitutions of the users through the obtained N types of short-term load curves, and analyzing the behaviors and the constitutions of the users by combining clustering results to obtain N types of users;

step 2.3: respectively calculating the maximum power value P of the N user classes _max Minimum power value P _min Time period t of peak occurrence _h Time period t of valley appearance ₁ And a peak-to-valley difference Δ P, wherein the maximum power value P _max Minimum power value P _min Time period t of peak occurrence _h Time period t of valley appearance ₁ Directly from the obtained short-term load curve, the peak-to-valley difference Δ p is calculated from the following formula: Δ P = P _max -P _min ；

Step 2.4: obtaining an ideal charging load curve of the electric automobile, namely a time period t occurring in a valley through the information ₁ Charging and at a maximum power value P _max Minimum power value P _min As a limiting condition;

and step 3: the method comprises the steps of establishing an optimization equation and solving to obtain a charging strategy by taking charging demand constraint, charging time constraint, electric vehicle charging quantity limitation, a power grid power load curve and power distribution network capacity as boundary conditions, taking the minimum error between a charging load curve and an ideal charging load curve as constraint conditions and taking the lowest electricity consumption cost of a user as an objective function;

and 4, step 4: whether the electric vehicle is charged or not is judged according to the decision result of the dynamic game, namely whether the obtained optimal charging strategy has adjustability or not is judged, and the specific method is as follows:

step 4.1: when a new electric vehicle is accessed into any one charging pile loaded in a certain power grid branch, the charging station operation system automatically updates the system information to the next 15min control time point according to the running state of the charging pile in the station, and firstly calculates the SOC state, the power required when the charging is fully performed and the number J of charging time segments required by the electric vehicle _i And the number of parking time segments T _i And T of the system for leaving the vehicle from the current time period _i Charging load margin M in each time period _t ，M _t ＝A _t S _T λ, where T =1,2, \ 8230;, T _i ，T _i ＝96；S _T Representing the rated capacity of the transformer; a. The _t The proportion of power which represents that a charging station is allowed to charge the electric vehicle in the t-th time period in the day accounts for the capacity of the transformer; λ is the power factor of the charging load;

and 4.2: before the preferential time-of-use electricity price time period for the user is formulated, the system judges whether the charging requirement of the electric automobile can be met in advance, and the T of leaving the automobile is calculated from the current time period by the calculation system _i Charging load margin M in each time period _t The realization method comprises the following steps: h _i ＝|A _i |，A _i ＝{t|M _t ≥P _i ，t＝1，2，...，T _i }; wherein, | A _i Is set A _i The number of the elements in the solution;

step 4.3: if it corresponds to M of the time interval _t ≥P _i +P ₀ That is, all the electric vehicles can be charged in the time period, otherwise, whether each vehicle is charged or not is reasonably arranged by adopting an optimization algorithm; p is _i Sum of the active power required for each charging pile when there is an electric vehicle access, P ₀ For the current real-time active power, M _t A larger value indicates a larger charge load margin for the corresponding period;

step 4.4: if H _i ＜J _i ，J _i The number of the time segments indicating the full charge indicates that the system can not meet the charging requirements input by all users in the parking time of the electric automobile, so that the electric automobile is charged by adopting an optimization algorithm, and the charging load margin M is obtained _t As a constraint condition, if the charging power required by a certain electric automobile exceeds 80% of the SOC, the charging power is distributed according to a chi priority parameter, namely P _i ＝χP _i Then with maxM _t ＝αP ₁ +βP ₂ +...+φP _n Is solved as an objective equation, where P _i The configured charging power of the electric automobiles, alpha, beta, eta and phi are specific gravity coefficients, the number of the electric automobiles meeting the optimal charging requirement is obtained, the charging piles can be known to charge the electric automobiles, and the number of the electric automobiles which are needed to be charged by a user is displayed at the same timeCharging with large power;

step 4.5: analyzing the result of the step 4.4, if the electric automobile cannot be charged, prompting the user to leave, if the electric automobile can be charged, prompting the user to charge the electric automobile, and calculating the state of charge of the battery of the electric automobile, which is maximally met by the system when the user leaves:

wherein the content of the first and second substances,

representing the initial battery capacity, eta is the power parameter, Δ t is the length of time from the start of charging to the departure, B _i Is the capacity of the electric steam battery;

because the electric automobile needs to be charged all the time during the parking time, the charging station prompts the user of the state of charge of the battery which can be met by the system to the maximum

And prompts the user to press the peak price p _h Charging, wherein a user autonomously selects to receive charging service or give up charging service; if the user receives the charging service, arranging the electric automobile to be in the charging load margin M _t Greater than P _i +P ₀ Is charged and the charging load margin A is set in the corresponding time interval _t S _T λ is updated to A _t S _T λ-P _i ；

Step 4.6: if H is _i ≥J _i If so, the system can meet the charging requirements of all the electric automobiles within the parking time of the electric automobiles, and peak clipping and valley filling are realized to the maximum extent on the basis of meeting the charging requirements of users; thus, the charging station ordered charging coordination system initially selects consecutive J _i The minimum value of the starting time interval with the maximum sum of the load margins is the starting time interval of the low-valley electricity price; specifically, the start period of the valley electricity price is preliminarily calculated according to the following expression:

consider from

Beginning of the time period until the electric vehicle leaves, and there is J _i Charging load margin M corresponding to each time period _t ≥P _i +P ₀ In the case of (1), the off-peak electricity price initial period is adjusted to be J _i Charging load margin M corresponding to time period _t ≥P _i +P ₀ To ensure that the system can charge the SOC of the electric vehicle to a desired level after the user responds to the valley price; specifically, the start period of the valley electricity prices is adjusted according to the following expression:

wherein, the first and the second end of the pipe are connected with each other,

n is an integer set;

step 4.7: at the initial time of the calculated adjusted off-peak electricity price

Thereafter, a period of time is determined

The charging price of the inner user is the valley electricity price p _l The charging price in other time periods is the peak price p _h The charging is started immediately or is delayed to the valley power price by the user's own choice; wherein D is a time parameter and is 0-1;

if the user selects to start charging immediately, the electric vehicle is arranged to start charging in the next time period and the corresponding charging load margin M is arranged one by one _t ≥P _i +P ₀ Until schedule J is charged _i Until a time period; simultaneously charging load margin A corresponding to charging period _t S _T λ is updated to A _t S _T λ-P _i ；

If the user selects to delay the charging to the valley price, the electric vehicle is arranged to start charging

Initially, one by one at the corresponding charging load margin M _t ≥P _i +P ₀ Until schedule J _i Until a time period; simultaneously charging load margin A corresponding to charging period _t S _T λ is updated to A _t S _T λ-P _i 。

2. The urban electric vehicle charging decision method based on the typical load dynamic game as claimed in claim 1, wherein: said step 1.2 further comprises the steps of:

step 1.2.1: identifying and correcting abnormal data in the load matrix A; the actual load data can cause the load data to be lost due to long-time fault of a certain measuring unit or fault of other related elements in the data acquisition process, and for 96 data points, records of 30 or more data points and load active record data are lost and are negative records, and the data are directly kicked out;

step 1.2.2: for the sudden rising and falling data in the load curve, judging by calculating the load change rate, wherein the load change rate calculation formula is as follows:

ρ＝|(p _d -p _d-1 )/p _d |；

where ρ represents the load change rate of the user load sequence P at point d, and P = { P = { P = _d ，d＝1，…，96}，p _d Active power at point d; when the load change rate rho exceeds a preset threshold value epsilon, determining the load change rate rho as abnormal data;

and filling abnormal data and load data containing the missing data by adopting a smooth window type according to the following formula:

wherein p is _d ' to fill the active power at point d, a, b denote forward and backward points, respectively, with a taken at maximum ₁ 、b ₁ ；

Step 1.2.3: carrying out normalization processing on the identified and corrected load matrix A by adopting a maximum normalization method, removing a base load part of a load curve, and emphasizing the similarity of load trends, wherein a normalization expression is as follows:

x _d ＝p _d ′/max(P)；

wherein x is _d Max (P) is the maximum power in 96 points for the normalized active power at d points.

3. The urban electric vehicle charging decision method based on the typical load dynamic game as claimed in claim 2, wherein: said step 1.3 further comprises the steps of:

step 1.3.1: obtaining the user category and the number of the branch, including the number h of industrial users ₁ The number h of resident life users ₂ The number h of school users ₃ Number of business users h ₄ Public transformer user number h ₅ (ii) a The optimized cluster number k is calculated by:

wherein f (-) is a normal function, c _i Is a constant number of times, and is,

is a transposition of a triangular matrix, h _i For the number of ith user categories, i =1,2, \8230, n, n =5, epsilon is the measurement error;

step 1.3.2: the K-means algorithm based on the double-layer weighted Euclidean distance comprises the following specific steps:

step 1.3.2.1: taking the optimized clustering number k obtained in the step 1.3.1 as an initial clustering center;

step 1.3.2.2: classifying samples; all sample centers of k classes are divided into class centers with the closest weighted Euclidean distance, sample A _k To the jth cluster center m _j ＝{m _j，1 ，m _j，2 ，…，m _j，q The weighted distance of is calculated by:

step 1.3.2.3: updating a clustering center; according to the result of the step 1.3.2.2, calculating the average value of each class as a new clustering center of each class;

step 1.3.2.4: performing iterative computation; judging whether the clustering center is converged, if not, returning to the step 1.3.2.2, otherwise, executing the next step;

step 1.3.2.5: combining all the clustering centers into a new data set, regarding each clustering center as a class, and calculating the distance between all the classes, namely the similarity between samples;

step 1.3.2.6: selecting categories of which the inter-category distance meets the requirements according to a set rule to carry out inter-category merging operation;

step 1.3.2.7: for class C _i And C _j Selecting the average distance between the centers of the two classes as the distance between the two classes, and then selecting the class with the minimum class distance for combination:

wherein D is _ij Is C _i And C _j The combined result, n _i 、n _j Are respectively class C _i And C _j X, x 'are class centers, d (x, x') is the average distance between the two class centers;

step 1.3.2.8: calculating the similarity between the new class and the previous class generated in the last step;

step 1.3.2.9: and (5) repeating the steps 1.3.2.6-1.3.2.8 until all the load samples are classified into one type, and ending the algorithm.

4. The urban electric vehicle charging decision-making method based on the typical load dynamic game as claimed in claim 3, wherein: step 1.4 adopts Thevenin bauxid index (DBI index) for determining the optimal clustering number and evaluating clustering quality, and the DBI index I _DBI The smaller the clustering effect, the better the clustering effect, and the calculation formula is:

wherein K is a cluster number, R _k Is the sum of the average distances within any two classes divided by the distance between the centers of the two classes, d (X) _k ) And d (X) _j ) Is the average distance of the distances in k and j classes, X _k 、X _j Respectively representing class k and j data, d (c) _k ，c _j ) Is the distance between the centers of two clusters, c _k 、c _j Representing the centers of classes k and j, respectively;

performing cluster analysis by using AHPK-means, repeating for 20 times, and selecting I _DBI And the corresponding solution is the best clustering result when the minimum value is reached.

5. The urban electric vehicle charging decision method based on the typical load dynamic game as claimed in claim 4, wherein: the number N =4 of user categories obtained in step 2.2, and the 4 user categories are:

(1) Late peak user group: the load change trend of a user accords with the form of a common daily load curve, namely, the load continuously rises from 5 am, the load is kept at a higher level to 9 am, the power consumption is reduced a little at noon 12 noon break, the power consumption continuously rises from 3 pm to 8 pm and reaches the highest peak of the power consumption all day long, and the power consumption slowly falls;

(2) Unimodal user group: the load change trend of the user has an electricity consumption peak, and in the overall trend, the amplitude difference of the electricity consumption power corresponding to the peak and the valley is not large, and the stable electricity consumption is continuously carried out from 5 am to 20 pm;

(3) Smooth user group: the load change trend of users is relatively smooth, the load change in one day is not large, but the active power level of the electric load of a smooth user group is very high, and the load is always at a high level;

(4) Avoiding peak user groups: the load change trend of the user is very obvious from other users, the electricity utilization characteristics of the user are obviously different from other four types, the user presents a peak avoiding type characteristic that two ends are high and the middle is low, electricity utilization is stopped at about 6 points in the morning, the electricity utilization is rapidly reduced and kept at a lower level until 6 points in the evening, the rest time is in the electricity utilization peak period, and the peak value and peak valley difference are also the highest in all the types.

6. The urban electric vehicle charging decision method based on the typical load dynamic game as claimed in claim 4, wherein: the step 3 further comprises the following steps:

step 3.1: initializing the load information of the power distribution network on the same day;

step 3.2: the charging station system judges whether a new electric vehicle drives into the charging station or not; if yes, reading all useful data information of the newly accessed electric automobile, including: acquiring the battery capacity of the electric vehicle, the current charge state of the electric vehicle battery, the expected retention time of the electric vehicle, a charging power curve of the electric vehicle battery and the expected charge state level of the electric vehicle when the electric vehicle leaves; if not, continuing to use the charging mode of the last time period;

step 3.3: obtaining the maximum value t of the residence time of all vehicles according to the expected residence time of the electric vehicle at the charging station in the time period _M And calculates the optimization time length T at this time,

| x | is the smallest integer less than x;

step 3.4: aiming at the optimization of the electric vehicle charging, a new charging control strategy is formulated according to the system information; establishing an optimization equation by taking charging demand constraint, charging time constraint, electric vehicle charging quantity limitation, a power grid power load curve and power distribution network capacity as boundary conditions, taking the minimum error of a charging load curve and an ideal charging load curve as constraint conditions and taking the minimum electricity consumption cost of a user as a target function;

the charging demand is constrained to

Wherein x is _n，t Active power at a point of time t when the nth car is accessed, S _n，S To an initial state of charge, b _n For charging the battery SOC value which can be increased in a period of time, the battery SOC of the charged electric automobile at least reaches the final SOC S required at the beginning of charging in T time periods _n，E While charging should be stopped in the case of full charge;

the charging time constraint is t _n，E ≤T _n，E The charged electric automobile needs to be charged within the expected retention time set by the user; wherein t is _n，E Charging end time for the nth electric vehicle; t is _n，E An expected end-of-charge time set for the electric vehicle user;

the charging quantity of the electric automobile is limited to

Wherein n is _t For the active power of the nth electric vehicle at the moment t, the number of the charging piles in the charging station is limited, and the number of the charging vehicles in each time period is limited by the number of the charging piles; x is the number of the charging piles in the charging station;

the peak-to-valley difference constraint is | P _max1 -P _min1 < Δ P; wherein, P _max1 And P _min1 Respectively in the morning of the dayThe maximum value and the minimum value of the system load in the period from the beginning to the end of the current optimization period; combining the peak-valley difference value of the time interval in the last 7 days, determining the initial value of the delta P as the minimum value of the peak-valley difference value of the time interval in the seven days, and if the value is small and the optimization target has no solution, adding 1% to the initial value of the delta P until the solution exists;

step 3.5: the selection of the charging start time adopts a nonlinear optimization method, a nonlinear equation is solved through MATLAB, and the appropriate charging start time is searched to minimize the fluctuation of the total load curve and take the fluctuation as an optimized objective function, and the specific method comprises the following steps:

step 3.5.1: the load variance is minimum in T time; after a charging load is connected, if charging is started at a certain moment, so that the variance of a total load curve added with the load in a T time period is minimum, the moment is the selected charging starting time; for the continuous function, the 2 nd order center distance representation variance is calculated

Wherein, P _var For the daily load variance, P, of the distribution network _av To adjust the daily average load, P _Et Is the total charging power, P, of the electric vehicle in the period of t _Lt The conventional load in the t time period when the power distribution network does not contain the charging load of the electric automobile is obtained through load prediction;

step 3.5.2: combining the condition of time-of-use electricity price, and taking the lowest electricity cost of the user as an objective function, namely

Wherein S is _j Representing the electricity price of the power grid at the moment j, wherein the positive value represents the charging electricity price of the electric automobile, and the negative value represents the subsidy electricity price of the electric automobile user for feeding electricity to the power grid; p _ij The representation represents the required power of i automobiles in the j time period;

step 3.5.3: the optimization problem of the 2 objective functions is converted into single-objective optimization by adopting a linear weighted sum method, meanwhile, due to the fact that the two objective functions have different dimensions, normalization is respectively carried out on the functions, and the converted objective functions are shown as the following formula:

minT ₁ ＝ω ₁ (T ₃ /T _3ma x)+ω ₂ (T ₄ /T _4max )；

in the formula, T _3max The mean square error of the original power grid load curve is obtained; t is _4max The daily charging cost of the vehicle owner is used for the traditional vehicle usage, namely the cost required by fully charging the battery from the lowest electric quantity in the stop time; t is ₃ 、T ₄ Representing two objective functions, ω, in step 3.3.1 and step 3.3.2, respectively ₁ 、ω ₂ Are respectively T ₃ 、T ₄ And satisfies ω ₁ +ω ₂ ＝1；

Step 3.6: respectively comparing the maximum power values P in the N user categories obtained in the step 2.3 _max Minimum power value P _min Time period t of peak occurrence _h Time period t of valley appearance _l And the peak-to-valley difference delta p is brought into the optimization equation established in the step 3.5, similarity judgment is respectively carried out on the real-time data of the power grid and the obtained N types of typical load curve data, and the ideal charging load curve is judged if an electric vehicle is accessed, so that the optimal charging strategy of each user category is obtained;

the similarity judgment formula is shown as follows:

wherein, K (x) ^new Is the real-time data of the current day,

for the qth exemplary payload data, q =1,2, \8230;, N.

Images