CN109359389A - City electric car charging decision method based on typical load dynamic game - Google Patents

Abstract

The present invention provides a kind of city electric car charging decision method based on typical load dynamic game, is related to electricity and fills automobile charging technique field.This method clusters to obtain somewhere typical user load type according to power grid short term curve, and electric car ideal charging load curve is calculated, it is constrained with charge requirement, charging time constraint, electric car charging quantity limitation, grid power load curve, power distribution network capacity is as boundary condition, with load curve and the minimum constraint condition of ideal charging load curve error of charging, with the minimum objective function of user power utilization cost, establish optimization method, and it solves, obtain charging strategy, judge whether to charge to electric car by the result of decision of dynamic game.Method provided by the invention, can improve network load characteristic makes power grid fluctuate minimum after electric car access, improves power equipment rate of load condensate, the charging cost on this basis spending automobile user is minimum.

Description

City electric car charging decision method based on typical load dynamic game

Technical field

The present invention relates to electricity to fill automobile charging technique field more particularly to a kind of city based on typical load dynamic game Electric car charging decision method.

Background technique

Global energy crisis and climate change drives electric car rapid development worldwide and answers With.As electric car is popularized in following, it will there is large-scale electric car to be connected to the grid and carries out charge and discharge, however with Increasing substantially for electric car quantity, for electric car because of the randomness of its own charging behavior, extensive access can be to electricity The operation and control of Force system bring significant harm.How power grid receives large-scale electric car grid-connected and is no more than electricity The significant problem that the capacity of net has become the following Development of Electric Vehicles and must solve during popularizing.In order to solve this It is a little difficult, it is necessary to which that the charge and discharge behavior for regulating and controlling electric car using reasonable means needs more accurate electric car Load modeling method and more reasonable scheme guide the orderly charge and discharge of electric car, the as far as possible big rule of raising power grid consumption The grid-connected ability of mould electric car.And avoiding apex filling vale can guide electricity as a kind of control measures flexible to a certain extent The charging modes of electrical automobile and the selection of time, can orderly charge and discharge, can to avoid electric car addition to power grid Caused by endanger, while the Experience Degree of charge user can be increased, user is allowed to enjoy lower charging price.

Summary of the invention

It is a kind of dynamic based on typical load the technical problem to be solved by the present invention is in view of the above shortcomings of the prior art, provide The city electric car charging decision method of state game, can improve network load characteristic fluctuates power grid after electric car access Minimum improves power equipment rate of load condensate, and the charging cost on this basis spending automobile user is minimum.

In order to solve the above technical problems, the technical solution used in the present invention is:

A kind of city electric car charging decision method based on typical load dynamic game, this method is by a transformer Branch is installed a total intelligent scheduling device and is realized, n charging pile of the intelligent scheduling device control transformer branch, when When having electric car access, charging pile uploads electric car information and is sent to intelligent scheduling device, and intelligent scheduling device is through excessive Charging strategy is sent after analysis to charge to charging pile, is communicated between intelligent scheduling device and charging pile by TCP；It is described Intelligent scheduling device obtain the specific method of charging strategy by analysis the following steps are included:

Step 1: short-term electric load being clustered by improved AHPK-means algorithm, predicts certain regional power grid Typical load curve, the typical load curve of power grid are the load curve removed except electric car charging load, including as follows Step:

Step 1.1: the load data structure of terminal user is chosen from the database of Utilities Electric Co.'s SCADA system in certain city At matrix of loadings A；The load data include the user type of a branch of some transformer of city, active power, The measured value of reactive power and voltage and current, every 15min sampling is primary, samples 96 data points altogether daily；

Step 1.2: data prediction, identification and amendment abnormal data being carried out to matrix of loadings A, and place is normalized Reason；It further includes steps of

Step 1.2.1: the abnormal data in matrix of loadings A is identified and is corrected；Actual load data can be due to certain The prolonged failure of a measuring unit or failure of other related elements lacks load data in data acquisition, for 96 data points, the record that the active record data of record, load of missing wherein 30 and above data point are negative, are directly kicked It removes；

Step 1.2.2: for rising sharply in load curve, rapid drawdown data, being judged by calculated load change rate, is born Lotus change rate calculation formula are as follows:

ρ=| (p_d-p_d-1)/p_d|；

Wherein, ρ indicates load changing rate of the customer charge sequence P in d point, P={ p_d, d=1 ..., 96 }, p_dFor d point Active power；When load changing rate ρ is more than preset threshold ε, then it is considered as abnormal data；

Load data for abnormal data and containing missing is filled up as the following formula using smooth window:

Wherein, p_d' it is the active power for filling up rear d point, a, b, which are respectively indicated, forwardly and rearwardly to be taken a little, and maximum is got respectively a₁、b₁；

Step 1.2.3: place is normalized with modified matrix of loadings A to by identification using maximum value method for normalizing Reason, the base lotus part of elimination capacity curve emphasize that the similitude of load tendency, normalization expression formula are as follows:

x_d=p_d′/max(P)；

Wherein, x_dFor the active power of d point after normalization, max (P) is the maximum power in 96 points；

Step 1.3: clustering being carried out to step 1.2 treated matrix of loadings A using AHPK-means algorithm, with double Layer weighted euclidean distance is that similarity criteria is clustered；It further includes steps of

Step 1.3.1: the class of subscriber and quantity of the branch, including industrial user's number h are obtained₁, resident living user Number h₂, school user number h₃, commercial user's number h₄, public become user's number h₅；It is calculate by the following formula optimization cluster numbers k:

Wherein, f () is normal function, c_iFor constant,For the transposition of triangular matrix, h_iFor i-th kind of class of subscriber number Amount, i=1,2 ..., n, n=5, ε are measurement error；

Step 1.3.2: the K-means algorithm based on double-weighing Euclidean distance, method particularly includes:

Step 1.3.2.1: the optimization cluster numbers k that step 1.3.1 is obtained is as initial cluster center；

Step 1.3.2.2: sample is sorted out；All center of a sample of k class are divided into the nearest class of its weighted euclidean distance The heart, sample A_kTo j-th of cluster centre m_j={ m_{J, 1}, m_{J, 2}..., m_{J, q}Weighted distance calculated by following formula:

Step 1.3.2.3: cluster centre updates；According to step 1.3.2.2's as a result, calculating the average value of every class as each The new cluster centre of class；

Step 1.3.2.4: iterative calculation；Judge whether cluster centre restrains, if not converged, return step 1.3.2.2, Otherwise it performs the next step；

Step 1.3.2.5: all cluster centres are combined into a new data set, each cluster centre is considered as itself One kind calculates all between class distances, i.e. similarity between sample；

Step 1.3.2.6: by established rule choose wherein between class distance reach requirement classification carry out class between union operation；

Step 1.3.2.7: for class C_iAnd C_j, distance of the average distance at two class centers as the two classes is selected, then The smallest class of class spacing is chosen to merge:Wherein, D_ijFor C_iAnd C_jIt is after merging as a result, n_i、n_jRespectively class C_iAnd C_jNumber of samples, x, x ' be class center, d (x, x ') be two class centers average distance；

Step 1.3.2.8: the new class of calculating previous step generation and the before similarity between class；

Step 1.3.2.9: repeating step 1.3.2.6-1.3.2.8, until all load samples are divided into one kind, algorithm knot Beam；

Step 1.4: Cluster Validity is examined；By establishing Validity Index, cluster result is evaluated, and determine best cluster Number, obtains final short term curve；

Using Dai Weisenbaoding index, that is, DBI index, for determining preferable clustering number and evaluating clustering result quality, DBI index I_DBISmaller expression Clustering Effect is better, calculation formula are as follows:

Wherein, K is cluster numbers, R_kThe sum of inter- object distance average distance for any two classification divided by two cluster centres away from From d (X_k) and d (X_j) be k and j inter- object distance average distance, X_k、X_jRespectively indicate k and j class data, d (c_k, c_j) it is two poly- Class centre distance, c_k、c_jRespectively indicate the center of k Yu j class；

Clustering is carried out using AHPK-means, is repeated 20 times, wherein I is chosen_DBICorresponding solution is most when minimum Good cluster result；

Step 2: electric car ideal charging load curve, including following step are calculated by obtained short term curve It is rapid:

Step 2.1: it is short at what there were significant differences that the class class center N being obtained by step 1.4 processing Power system load data matrix A Phase load curve；

Step 2.2: the behavior and composition of user being obtained by obtained N class short term curve, and combines cluster result The behavior and composition of user are analyzed, N kind class of subscriber is obtained；

Step 2.3: calculating separately the maximum power value P of this N kind class of subscriber_max, minimal power values P_min, peak occur Time period t_h, low ebb occur time period t₁And peak-valley difference Δ p, wherein maximum power value P_max, minimal power values P_min, peak The time period t of appearance_h, low ebb occur time period t₁It is directly obtained by obtained short term curve, peak-valley difference Δ p is by following formula It is calculated: Δ p=P_max-P_min；

Step 2.4: electric car ideal charging load curve, i.e., the period occurred in low ebb are obtained by information above t₁It charges and with maximum power value P_max, minimal power values P_minAs restrictive condition；

Step 3: bent with charge requirement constraint, charging time constraint, the limitation of electric car charging quantity, grid power load Line, power distribution network capacity are as boundary condition, with load curve and the minimum constraint condition of ideal charging load curve error of charging, With the minimum objective function of user power utilization cost, optimization method is established, and solve, obtain charging strategy；Specifically include following step It is rapid:

Step 3.1: initialization same day power distribution network load information；

Step 3.2: charging station system judges whether there is new electric car and drives into charging station；If so, reading all new Access electric car useful data information, comprising: obtain batteries of electric automobile capacity, the current state-of-charge of battery of electric vehicle, The charge power curve and electric vehicle of residence time, batteries of electric automobile expected from electric car desired electric car when leaving State-of-charge is horizontal；If not provided, continuing to use the charge mode of a period；

Step 3.3: according to the expected downtime of charging station electric car in the period, when obtaining all stoppage of vehicle Between maximum value t_M, and this suboptimization time span T is calculated,| x | for the smallest positive integral less than x；

Step 3.4: turning to target so that electric car charging is optimal, according to system information, formulate new charge control strategy； With charge requirement constraint, charging time constraint, electric car charging quantity limitation, grid power load curve, power distribution network capacity As boundary condition, with load curve and the minimum constraint condition of ideal charging load curve error of charging, with user power utilization at This minimum objective function, establishes optimization method；

The charge requirement is constrained toWherein, x_{N, t}When t when being accessed for n-th automobile Carve certain point active power, S_{N, S}For initial state-of-charge, b_nFor charge a period increasable battery SOC numerical value, at T Between in section, the battery charge state of the electric car being electrically charged should required final charged shape when at least up to charging starts State S_{N, E}, while should stop charging in the case where fully charged；

The charging time is constrained to t_{N, E}≤T_{N, E}, the electric car needs being electrically charged are in expected stop set by user Interior charging complete；Wherein t_{N, E}It charges the end time for n-th electric car；T_{N, E}For the expection of automobile user setting It charges the end time；

The electric car charging quantity is limited toWherein, n_tFor the active of n-th electric car t moment Power, charging pile limited amount in charging station, Rechargeable vehicle quantity is limited by charging pile quantity in each period；X is to fill Charging pile quantity in power station；

The peak- valley load constraint is | P_max1-P_min1| < Δ P；Wherein, P_max1And P_min1Respectively since morning on the same day to working as The preceding optimization period terminates system loading maximum value and minimum value in this period；In conjunction with the peak-valley difference of the period in past 7 days Value, Δ P initial value is set to the period peak-valley difference minimum value in this seven days, if the value causes optimization aim without solution due to less than normal, Δ P initial value progressively increases 1%, until there is solution；

Step 3.5: the selection for time started of charging uses nonlinear optimization method, solves non-linear side by MATLAB Journey finds the suitable charging time started, fluctuates total load curve minimum, in this, as the objective function of optimization, specifically Method are as follows:

Step 3.5.1:T time internal loading variance is minimum；After a charging load accesses, if at a time starting to fill Electricity, the total load curve after making to be added the load is minimum in T period internal variance, then the moment be the charging time started selected； Its 2 rank center is calculated away from variance is indicated for continuous function, can be obtained

Wherein, P_varFor power distribution network daily load variance, P_avFor the per day load before adjustment, P_EtFor t period electric car Total charge power, P_LtThe conventional load of t period when for power distribution network without electric car charging load, that is, it is pre- by load It measures；

Step 3.5.2: in conjunction with the case where tou power price, with the minimum objective function of user power utilization cost, i.e.,Wherein, S_jPower grid is represented in the electricity price at j moment, indicates electric car charging electricity price for positive value, The subsidy electricity price that automobile user is fed to power grid is then represented for negative value；P_ijNeeded for indicating i platform automobile within the j period Power；

Step 3.5.3: linear weight sum method is used, it is excellent to convert single goal for the optimization problem of above-mentioned 2 objective functions Change, simultaneously as the dimension of two objective functions is different, each function is normalized respectively, the objective function after conversion is as follows Shown in formula:

min T₁=ω₁(T₃/T_3max)+ω₂(T₄/T_4max)；

In formula, T_3maxFor the mean square deviation of original grid load curve；T_4maxIt is accustomed to being charged to day for lower car owner with vehicle for tradition This, i.e., be full of required cost by minimum electricity for battery in down time；T₃、T₄Respectively represent step 3.3.1 and step 3.3.2 In two objective functions, ω₁、ω₂Respectively T₃、T₄Weight coefficient, and meet ω₁+ω₂=1；

Step 3.6: the maximum power value P in N kind class of subscriber for respectively obtaining step 2.3_max, minimal power values P_min, peak occur time period t_h, low ebb occur time period t_lAnd peak-valley difference Δ p is brought into the optimization of step 3.5 foundation In equation, similarity judgement is carried out with obtained N kind typical load curve data by real-time data of power grid respectively, if judgement When having electric car access, it should which kind of ideal charging load curve used, in the hope of the optimal charging plan of every kind of class of subscriber Slightly；

Similarity judgment formula is shown below:

Wherein, K (x)^newFor the real time data on the same day,For q kind typical load data, q=1,2 ..., N；

Step 4: judging whether to charge to electric car by the result of decision of dynamic game, i.e. judgement obtains most Whether excellent charging strategy has controllability, and the specific method is as follows:

Step 4.1: when in a certain grid branch with there is new electric car access in any one charging pile carried, filling System information is updated to next 15min automatically according to charging pile operating status in standing and controls time point by power station operation system, and Calculate first the corresponding SOC state of this electric car, it is fully charged when required power, required charging time number of segment J_i, parking Time hop counts T_iThe T left in the vehicle is counted from present period with system_iCharging load margin M in a period_t, M_t=A_tS_T λ, wherein t=1,2 ..., T_i, T_i=96；S_TIndication transformer rated capacity；A_tWith indicating in one day to permit in t-th of period Perhaps charging station accounts for the ratio of transformer capacity to the power that electric car charges；λ is the power factor of charging load；

Step 4.2: before formulating the preferential tou power price period towards the user, can system is prejudged meet this The charge requirement of electric car counts the T left in the vehicle from present period by computing system_iCharging in a period is negative Lotus nargin M_tIt realizes: H_i=| A_i|, A_i=t | M_t≥P_i, t=1,2 ..., T_i}；Wherein, | A_i| it is set A_iInterior element Number；

Step 4.3: if the M of corresponding period_t≥P_i+P₀, i.e., all electric cars can be arranged to charge in the period, otherwise adopted Reasonably arrange whether each automobile charges with optimizing algorithm,；P_iIt is required active when having an electric car access for each charging pile The summation of power, P₀For current real-time active power, M_tIt is bigger to be worth the bigger charging load margin for indicating the corresponding period；

Step 4.4: if H_i< J_i, J_iIndicate the time hop counts that electricity is full of, then explanation is in the electric car berthing time Interior, system is unable to satisfy the charge requirement of all user's inputs, thus arrange electric car to charge using optimizing algorithm, with Charge load margin M_tIt is excellent with χ if charge power needed for certain electric car is more than the 80% of SOC as constraint condition First parameter is allocated, i.e. P_i=χ P_i, later with max M_t=α P₁+βP₂+...+φP_nIt is solved as target equation, Middle P_iBy the charge power of electric car configured, α, β ..., φ be specific gravity factor, ask and be satisfied optimal charge demand Electric car quantity, and learn which charging pile can charge for electric car, while showing that user should be with great function Rate charges；

Step 4.5: the result of step 4.4 being analyzed, if cannot charge for electric car, prompts user It leaves, if can charge for electric car, reminds user that can charge for its electric car, and be calculated When user leaves, the batteries of electric automobile state-of-charge of system maximum satisfaction:Wherein, Indicate that initial battery capacity, η are power parameter, Δ t is the when length from starting to charge to leaving, B_iFor electronic vapour battery appearance Amount；

Since the electric car needs to charge always in berthing time, charging station prompts custom system maximum can meet Battery charge stateAnd prompt user by peak electricity tariff p_hIt charges, independently selected to receive charging service by user or puts Abandon charging service；If user receives charging service, arrange electric car in charging load margin M_tGreater than P_i+P₀Period into Row charging, and will charging load margin A in the corresponding period_tS_Tλ is updated to A_tS_Tλ-P_i；

Step 4.6: if H_i≥J_i, then explanation system in the electric car berthing time can meet all electric cars Charge requirement, and on the basis of meeting user's charge requirement, peak load shifting is realized to the maximum extent；Therefore, charging station has The sequence charging continuous J of coordination system initial option_iThe minimum value of the maximum start periods of the sum of a load margin is low ebb electricity price Start periods；Specifically, according to the start periods of following expression primary Calculation low ebb electricity price:

Consider fromPeriod starts when the electric car leaves, and there are J_iA period corresponding charging load Nargin M_t≥P_i+P₀The case where, then it needs to adjust low ebb electricity price start periods to there is J thereafter_iPeriod is abundant to inductive charging load Spend M_t≥P_i+P₀Position, required for ensuring that the SOC of electric car can be charged to by user's system after responding low ebb electricity price It is horizontal；Specifically, it is adjusted according to start periods of the following expression to low ebb electricity price:

Wherein,N is set of integers；

Step 4.7: the low ebb electricity price start periods adjusted are being calculatedAfterwards, the period is determined The charging electricity price of interior user is paddy electricity electricity price p_l, remaining period charging electricity price is peak electricity tariff p_h, to prompt user, certainly by user Main selection immediately begins to charge, or is delayed to paddy electricity electricity price and starts to charge；Wherein, D is time parameter, takes 0~1；

If user's selection immediately begins to charge, arrange the electric car since the next period one by one negative to inductive charging Lotus nargin M_t≥P_i+P₀Period charging, until arrange J_iUntil a period；Simultaneously by the charging load of corresponding charge period Nargin A_tS_Tλ is updated to A_tS_Tλ-P_i；

If user's selection is delayed to paddy electricity electricity price and starts to charge, arrange the electric car fromStart, is filled one by one in correspondence Electric load nargin M_t≥P_i+P₀Period charging, until arrange J_iUntil a period；Simultaneously by the charging of corresponding charge period Load margin A_tS_Tλ is updated to A_tS_Tλ-P_i。

Further, class of subscriber number N=4 obtained in step 2.2,4 kinds of class of subscribers are respectively as follows:

(1) evening peak type user group: the load variations trend of its user meets the form of common daily load curve, i.e., from When morning 5, load constantly rises, and starts to be maintained at a higher level when to morning 9, and lunch break when noon 12 causes Electric power is reduced on a small quantity, later from afternoon 3 when at night 8 when electricity consumption persistently rise, and reach the top of whole day electricity consumption, it Electricity consumption has beginning slowly decline afterwards；

(2) single peak type user group: there is a peaks of power consumption for the load variations trend of its user, and take advantage of a favourable situation integrally On, peak and the amplitude difference of electric power corresponding to low ebb are little, when morning 5, at night 20 points continue relatively to put down Steady electricity consumption；

(3) leveling style user group: the load variations trend of its user is more gentle, and one day load variations is little, but flat The power load active power level of slow class user group is very high, and load is constantly in higher level；

(4) peak type user group: the load variations trend of its user and the difference highly significant of other users is kept away, electricity consumption is special Property with other four classes there is significant difference, show two head heights, it is intermediate it is low keep away peak type feature, 6:00 AM or so stops Electricity consumption, electricity consumption quickly reduce and are maintained at a lower level, until 6 points of night, remaining time is then in peak times of power consumption, Peak value, peak-valley difference are also highest in all classes.

The beneficial effects of adopting the technical scheme are that the typical case provided by the invention based on dynamic game is negative The extensive electric car charging method in lotus city can effectively reduce network load peak-valley difference, improve network load characteristic, mention High power equipment rate of load condensate.There is following advantage compared with prior art:

(1) Power system load data that the present invention uses is the whole load data in a certain area, to the electronic of this region Automobile charging optimization is more fully, precision is higher and effect is more preferable；

(2) there is higher scalability and high efficiency using AHPK-means clustering algorithm, and makes algorithm more smart Really, more stable；

(3) during solving optimization equation, more restrictive conditions are used, so that obtained optimisation strategy is more quasi- Really；

(4) it by solution nonlinear equation, fluctuates total load curve minimum, uses three kinds of optimization object functions, So that optimization is more fully.

Detailed description of the invention

Fig. 1 is the typical load city extensive electric car charging side provided in an embodiment of the present invention based on dynamic game The general flow chart of method；

Fig. 2 is the flow chart that AHPK-means clustering algorithm provided in an embodiment of the present invention is clustered；

Fig. 3 is the Clustering Effect figure of AHPK-means clustering algorithm provided in an embodiment of the present invention；

Fig. 4 is the electricity consumption active power variation that evening peak type user group provided in an embodiment of the present invention generates in one day Figure；

Fig. 5 is the electricity consumption active power variation diagram that single peak type user group provided in an embodiment of the present invention generates in one day；

Fig. 6 is the electricity consumption active power variation diagram that leveling style user group provided in an embodiment of the present invention generates in one day；

Fig. 7 is the electricity consumption active power variation diagram provided in an embodiment of the present invention keeping away peak type user group and generating in one day；

Fig. 8 is charging system for electric automobile structural block diagram used in the embodiment of the present invention；

Fig. 9 is Monte Carlo Analogue Method schematic diagram provided in an embodiment of the present invention.

Specific embodiment

With reference to the accompanying drawings and examples, specific embodiments of the present invention will be described in further detail.Implement below Example is not intended to limit the scope of the invention for illustrating the present invention.

For the composition for studying certain region load, need to know the real-time load data and every load line of all users Or specially change or public affairs become topological correlation information in the zone, and to specify which point be various encoded negative government office in feudal China's properties belong to Class.By power grid SCADA system, teach news information system can obtain power network topology, customer charge amount, load character, power grid respectively The information such as structure realize the knot of total system load so as to count the proportion of the loads such as industry, agricultural, business, resident Structure decomposes and structural analysis.Processing is merged to same attribute node using partitioning strategies, and then is formd to whole region system The parsing of system load configuration, realizes the quantitative analysis to load structure, grasps region load character.

The composition of existing load ingredient refers to load of different nature specific gravity shared in total load, total classification of load It is to be distinguished according to industry, agricultural, business, resident and other loads, specific as follows:

(1) industrial load is primarily referred to as production, processing, manufacturing enterprise, such as mining, food processing, tobacco, weaving, timber Such as work, furniture, papermaking, printing, chemistry, petrochemical industry；

(2) agriculture load, is primarily referred to as field irrigation etc.；

(3) Commercial Load is primarily referred to as the mechanism for business, such as commerce services mechanism, some enterprises and institutions, School, hospital, government etc.；

(4) resident load, refer to occupying Cui, apartment etc., main includes the electricity consumption of life, rest aspect；

(5) other loads, including street or highway lighting electricity consumption, electric railway or subway electricity consumption, balance of power plant Electricity consumption.

As shown in Figure 1, the present embodiment provides a kind of city electric car charging decisions based on typical load dynamic game Method, this method are installed a total intelligent scheduling device by a transformer branch and are realized, the intelligent scheduling device control N charging pile of transformer branch, when there is electric car access, charging pile uploads electric car information and is sent to intelligent scheduling Device, intelligent scheduling device send charging strategy after analysis and charge to charging pile, intelligent scheduling device and charging pile Between communicated by TCP；The intelligent scheduling device obtains charging strategy by analysis the specific method is as follows described.

Step 1: short-term electric load being clustered by improved AHPK-means algorithm, predicts certain regional power grid Typical load curve, the typical load curve of power grid are the load curve removed except electric car charging load, including as follows Step:

Step 1.1: the load data structure of terminal user is chosen from the database of Utilities Electric Co.'s SCADA system in certain city At matrix of loadings A；The load data include the user type of a branch of some transformer of city, active power, The measured value of reactive power and voltage and current, every 15min sampling is primary, samples 96 data points altogether daily；

The present embodiment with certain city in July, 2018 working days evidence, including certain Utilities Electric Co. directly under and its subordinate other 7 The user of counties and districts is active and reactive and the measured values such as voltage and current, and the daily load curve for surveying 2567 users is research object, The daily load sampling interval is 15min, amounts to 96 sampled points daily, contains 8563 effective daily load curves altogether, constitutes 8563*96 Rank matrix A；

Step 1.2: data prediction, identification and amendment abnormal data being carried out to matrix of loadings A, and place is normalized Reason；It further includes steps of

Step 1.2.1: the abnormal data in matrix of loadings A is identified and is corrected；Actual load data can be due to certain The prolonged failure of a measuring unit or failure of other related elements lacks load data in data acquisition, for 96 data points, the record that the active record data of record, load of missing wherein 30 and above data point are negative, are directly kicked It removes；

Step 1.2.2: for rising sharply in load curve, rapid drawdown data, being judged by calculated load change rate, is born Lotus change rate calculation formula are as follows:

ρ=| (p_d-p_d-1)/p_d|；

Wherein, ρ indicates load changing rate of the customer charge sequence P in d point, P={ p_d, d=1 ..., 96 }, p_dFor d point Active power；When load changing rate ρ is more than preset threshold ε, then it is considered as abnormal data；

Load data for abnormal data and containing missing is filled up as the following formula using smooth window:

Wherein, p_d' it is the active power for filling up rear d point, a, b, which are respectively indicated, forwardly and rearwardly to be taken a little, and maximum is got respectively a₁、b₁；In the present embodiment, a₁、b₁It can be respectively 3,4 or 5；

Step 1.2.3: it is normalized to by identification with modified matrix of loadings A, the base of elimination capacity curve The similitude of load tendency is emphasized in lotus part, specifically uses maximum value method for normalizing, and expression formula is as follows:

x_d=p_d′/max(P)；

Wherein, x_dFor the active power of d point after normalization, max (P) is the maximum power in 96 points；

Step 1.3: clustering being carried out to step 1.2 treated matrix of loadings A using AHPK-means algorithm, with double Layer weighted euclidean distance is that similarity criteria is clustered；As shown in Fig. 2, cluster process further includes steps of

Step 1.3.1: the class of subscriber and quantity of the branch, including industrial user's number h are obtained₁, resident living user Number h₂, school user number h₃, commercial user's number h₄, public become user's number h₅；It is calculate by the following formula optimization cluster numbers k:

Wherein, f () is normal function, c_iFor constant,For the transposition of triangular matrix, h_iFor i-th kind of class of subscriber number Amount, i=1,2 ..., n, n=5, ε are measurement error；

Step 1.3.2: the K-means algorithm based on double-weighing Euclidean distance, method particularly includes:

Step 1.3.2.1: the optimization cluster numbers k that step 1.3.1 is obtained is as initial cluster center；

Step 1.3.2.2: sample is sorted out；All center of a sample of k class are divided into the nearest class of its weighted euclidean distance The heart, sample A_kTo j-th of cluster centre m_j={ m_{J, 1}, m_{J, 2}..., m_{J, q}Weighted distance calculated by following formula:

Step 1.3.2.3: cluster centre updates；According to step 1.3.2.2's as a result, calculating the average value of every class as each The new cluster centre of class；

Step 1.3.2.4: iterative calculation；Judge whether cluster centre restrains, if not converged, return step 1.3.2.2, Otherwise it performs the next step；

Step 1.3.2.5: all cluster centres are combined into a new data set, each cluster centre is considered as itself One kind calculates all between class distances, i.e. similarity between sample；

Step 1.3.2.6: by established rule choose wherein between class distance reach requirement classification carry out class between union operation；

Step 1.3.2.7: for class C_iAnd C_j, distance of the average distance at two class centers as the two classes is selected, then The smallest class of class spacing is chosen to merge:

Wherein, D_ijFor C_iAnd C_jIt is after merging as a result, n_i、n_jRespectively class C_iAnd C_jNumber of samples, x, x ' be class center, D (x, x ') is the average distance at two class centers；

Step 1.3.2.8: the new class of calculating previous step generation and the before similarity between class；

Step 1.3.2.9: repeating step 1.3.2.6-1.3.2.8, until all load samples are divided into one kind, algorithm knot Beam；

Step 1.4: Cluster Validity is examined；By establishing Validity Index, cluster result is evaluated, and determine best cluster Number, obtains final short term curve；

Using Dai Weisenbaoding index, that is, DBI index, for determining preferable clustering number and evaluating clustering result quality.DBI index It is one of common Validity Index in existing clustering method, the purpose of load curve cluster is to obtain different typical load songs Line makes every class curve have similar mode, reflects the user power utilization feature of same type.

DBI index comprehensive considers the compactedness in dispersibility and class between class, DBI index I_DBISmaller expression Clustering Effect It is better, calculation formula are as follows:

Wherein, K is cluster numbers, R_kThe sum of inter- object distance average distance for any two classification divided by two cluster centres away from From d (X_k) and d (X_j) be k and j inter- object distance average distance, X_k、X_jRespectively indicate k and j class data, d (c_k, c_j) it is two poly- Class centre distance, c_k、c_jRespectively indicate the center of k Yu j class；

Clustering is carried out using AHPK-means, is repeated 20 times, wherein I is chosen_DBICorresponding solution is most when minimum Good cluster result；

Obtained Clustering Effect is as shown in Figure 3.

Step 2: electric car ideal charging load curve, including following step are calculated by obtained short term curve It is rapid:

Step 2.1: it is short at what there were significant differences that the class class center N being obtained by step 1.4 processing Power system load data matrix A Phase load curve；

Step 2.2: the behavior and composition of user being obtained by obtained N class short term curve, and combines cluster result The behavior and composition of user are analyzed, N kind class of subscriber is obtained；

4 kinds of class of subscribers are obtained in the present embodiment, are respectively as follows:

(1) evening peak type user group: the load variations trend of its user meets the form of common daily load curve, i.e., from When morning 5, load constantly rises, and starts to be maintained at a higher level when to morning 9, and lunch break when noon 12 causes Electric power is reduced on a small quantity, later from afternoon 3 when at night 8 when electricity consumption persistently rise, and reach the top of whole day electricity consumption, it Electricity consumption has beginning slowly decline afterwards；The electricity consumption active power variation that such user generates in one day is as shown in Figure 4；

(2) single peak type user group: there is a peaks of power consumption for the load variations trend of its user, and take advantage of a favourable situation integrally On, peak and the amplitude difference of electric power corresponding to low ebb are little, when morning 5, at night 20 points continue relatively to put down Steady electricity consumption；The electricity consumption active power variation that such user generates in one day is as shown in Figure 5；

(3) leveling style user group: the load variations trend of its user is more gentle, and one day load variations is little, but flat The power load active power level of slow class user group is very high, and load is constantly in higher level；Such user is one The electricity consumption active power variation generated in it is as shown in Figure 6；

(4) peak type user group: the load variations trend of its user and the difference highly significant of other users is kept away, electricity consumption is special Property with other four classes there is significant difference, show two head heights, it is intermediate it is low keep away peak type feature, 6:00 AM or so stops Electricity consumption, electricity consumption quickly reduce and are maintained at a lower level, until 6 points of night, remaining time is then in peak times of power consumption, Peak value, peak-valley difference are also highest in all classes；The electricity consumption active power variation that such user generates in one day is such as Shown in Fig. 7；

Step 2.3: calculating separately the maximum power value P of this N kind class of subscriber_max, minimal power values P_min, peak occur Time period t_h, low ebb occur time period t_lAnd peak-valley difference Δ p, wherein maximum power value P_max, minimal power values P_min, peak The time period t of appearance_h, low ebb occur time period t_lIt is directly obtained by obtained short term curve, peak-valley difference Δ p is by following formula It is calculated:

Δ p=P_max-P_min；

Step 2.4: electric car ideal charging load curve, i.e., the period occurred in low ebb are obtained by information above T1 charges and with maximum power value P_max, minimal power values P_minAs restrictive condition.

Step 3: bent with charge requirement constraint, charging time constraint, the limitation of electric car charging quantity, grid power load Line, power distribution network capacity are as boundary condition, with load curve and the minimum constraint condition of ideal charging load curve error of charging, With the minimum objective function of user power utilization cost, optimization method is established, and solve, obtain charging strategy；Specifically include following step It is rapid:

Step 3.1: initialization same day power distribution network load information；

Step 3.2: charging station system judges whether there is new electric car and drives into charging station；If so, reading all new Access electric car useful data information, comprising: obtain batteries of electric automobile capacity, the current state-of-charge of battery of electric vehicle, The charge power curve and electric vehicle of residence time, batteries of electric automobile expected from electric car desired electric car when leaving State-of-charge is horizontal；If not provided, continuing to use the charge mode of a period；

Step 3.3: according to the expected downtime of charging station electric car in the period, when obtaining all stoppage of vehicle Between maximum value t_M, and this suboptimization time span T is calculated,| x | for the smallest positive integral less than x；

Step 3.4: turning to target so that electric car charging is optimal, according to system information, formulate new charge control strategy； With charge requirement constraint, charging time constraint, electric car charging quantity limitation, grid power load curve, power distribution network capacity As boundary condition, with load curve and the minimum constraint condition of ideal charging load curve error of charging, with user power utilization at This minimum objective function, establishes optimization method；

The charge requirement is constrained to

Wherein, x_{N, t}T moment point active power when being accessed for n-th automobile, S_{N, S}For initial state-of-charge, b_nTo fill Increasable battery SOC numerical value of the electric period, within T period, the battery charge state for the electric car being electrically charged is answered The required final state-of-charge S when at least up to charging starts_{N, E}, while should stop charging in the case where fully charged；

The charging time is constrained to

t_{N, E}≤T_{N, E}；

The electric car being electrically charged needs the charging complete in expected downtime set by user；Wherein, t_{N, E}It is n-th Electric car charges the end time；T_{N, E}For the expected charging end time of automobile user setting；

The electric car charging quantity is limited to

Wherein, n_tFor the active power of n-th electric car t moment, charging pile limited amount in charging station, each time Rechargeable vehicle quantity is limited by charging pile quantity in section；X is charging pile quantity in charging station；

The peak- valley load constraint is

|P_max1-P_min1| < Δ P；

Wherein, P_max1And P_min1It is negative to terminate system in this period to the current optimization period since morning on the same day respectively Lotus maximum value and minimum value；In conjunction with the peak valley difference value of the period in past 7 days, Δ P initial value is set to the period peak valley in this seven days Poor minimum value, but the value be likely due to it is less than normal cause optimisation strategy without solution, if Δ P initial value progressively increases 1% without solution, until There is solution；

Step 3.5: the selection for time started of charging uses nonlinear optimization method, solves non-linear side by MATLAB Journey finds the suitable charging time started, fluctuates total load curve minimum, in this, as the objective function of optimization, specifically Method are as follows:

Step 3.5.1:T time internal loading variance is minimum；After a charging load accesses, if at a time starting to fill Electricity, the total load curve after making to be added the load is minimum in T period internal variance, then the moment be the charging time started selected； Its 2 rank center is calculated away from variance is indicated for continuous function, can be obtained

Wherein, P_varFor power distribution network daily load variance, P_avFor the per day load before adjustment, P_EtFor t period electric car Total charge power, P_modelFor the typical load for the t period that step 1 obtains, the load is without electric car charging load；

Step 3.5.2: in conjunction with the case where tou power price, with the minimum objective function of user power utilization cost, i.e.,

Wherein, x_{N, t}T moment point active power when being accessed for n-th automobile, S_{N, s}For initial state-of-charge, b_nTo fill Increasable battery SOC numerical value of the electric period, S_jPower grid is represented in the electricity price at j moment, indicates that electric car fills for positive value Electricity price then represents the subsidy electricity price that automobile user is fed to power grid for negative value；

Step 3.5.3: above-mentioned 2 objective functions be it is interactional, for realize both synthesis it is optimal, using linear weighted function It is handled with method, i.e., converts single object optimization for multi-objective optimization question；Simultaneously as the dimension of two objective functions is not Together, each function is normalized respectively, the objective function after conversion is shown below:

min T₁=ω₁(T₃/T_3max)+ω₂(T₄/T_4max)；

In formula, T_3maxFor the mean square deviation of original grid load curve；T_4maxIt is accustomed to being charged to day for lower car owner with vehicle for tradition This, i.e., be full of required cost by minimum electricity for battery in down time；T₃、T₄Respectively represent step 3.3.1 and step 3.3.2 In two objective functions, ω₁、ω₂Respectively T₃、T₄Weight coefficient, and meet ω₁+ω₂=1；

Step 3.6: the maximum power value P in N kind class of subscriber for respectively obtaining step 2.3_max, minimal power values P_min, peak occur time period t_h, low ebb occur time period t_lAnd peak-valley difference Δ p is brought into the optimization of step 3.5 foundation In equation, similarity judgement is carried out with obtained N kind typical load curve data by real-time data of power grid respectively, if judgement When having electric car access, it should which kind of ideal charging load curve used, in the hope of the optimal charging plan of every kind of class of subscriber Slightly；

Similarity judgment formula is shown below:

Wherein, K (x)^newFor the real time data on the same day,For q kind typical load data, q=1,2 ..., N.

Step 4: judging whether to charge to electric car by the result of decision of dynamic game, i.e. judgement obtains most Whether excellent charging strategy has controllability, and the specific method is as follows:

Step 4.1: when in a certain grid branch with there is new electric car access in any one charging pile carried, filling System information is updated to next 15min automatically according to charging pile operating status in standing and controls time point by power station operation system, and Calculate first the corresponding SOC state of this electric car, it is fully charged when required power, required charging time number of segment J_i, parking Time hop counts T_iThe T left in the vehicle is counted from present period with system_iCharging load margin M in a period_t, M_t=A_tS_T λ, wherein t=1,2 ..., T_i, T_i=96；S_TIndication transformer rated capacity；A_tWith indicating in one day to permit in t-th of period Perhaps charging station accounts for the ratio of transformer capacity to the power that electric car charges；λ is the power factor of charging load；

Step 4.2: before formulating the preferential tou power price period towards the user, can system is prejudged meet this The charge requirement of electric car counts the T left in the vehicle from present period by computing system_iCharging in a period is negative Lotus nargin M_tIt realizes:

H_i=| A_i|, A_i=t | M_t≥P_i, t=1,2 ..., T_i}；

Wherein, | A_i| it is set A_iInterior element number；

Step 4.3: if the M of corresponding period_t≥P_i+P₀, i.e., all electric cars can be arranged to charge in the period, otherwise adopted Reasonably arrange whether each automobile charges with optimizing algorithm,；P_iIt is required active when having an electric car access for each charging pile The summation of power, P₀For current real-time active power, M_tIt is bigger to be worth the bigger charging load margin for indicating the corresponding period；

Step 4.4: if H_i< J_i, J_iIndicate the time hop counts that electricity is full of, then explanation is in the electric car berthing time Interior, system is unable to satisfy the charge requirement of all user's inputs, thus arrange electric car to charge using optimizing algorithm, with Charge load margin M_tIt is excellent with χ if charge power needed for certain electric car is more than the 80% of SOC as constraint condition First parameter is allocated, i.e. P_i=χ P_i, later with max M_t=α P₁+βP₂+...+φP_nIt is solved as target equation, Middle P_iBy the charge power of electric car configured, α, β ..., φ be specific gravity factor, ask and be satisfied optimal charge demand Electric car quantity, and learn which charging pile can charge for electric car, while showing that user should be with great function Rate charges；

Step 4.5: the result of step 4.4 being analyzed, if cannot charge for electric car, prompts user It leaves, if can charge for electric car, reminds user that can charge for its electric car, and be calculated When user leaves, the batteries of electric automobile state-of-charge of system maximum satisfaction:

Wherein,Indicate that initial battery capacity, η are power parameter, Δ t is the when length from starting to charge to leaving, B_i For electronic vapour battery capacity；

Since the electric car needs to charge always in berthing time, charging station prompts custom system maximum can meet Battery charge stateAnd prompt user by peak electricity tariff p_hIt charges, independently selected to receive charging service by user or puts Abandon charging service；If user receives charging service, arrange electric car in charging load margin M_tGreater than P_i+P₀Period into Row charging, and will charging load margin A in the corresponding period_tS_Tλ is updated to A_tS_Tλ-P_i；

Step 4.6: if H_i≥J_i, then explanation system in the electric car berthing time can meet all electric cars Charge requirement, and on the basis of meeting user's charge requirement, peak load shifting is realized to the maximum extent；Therefore, charging station has The sequence charging continuous J of coordination system initial option_iThe minimum value of the maximum start periods of the sum of a load margin is low ebb electricity price Start periods；Specifically, according to the start periods of following expression primary Calculation low ebb electricity price:

Consider fromPeriod starts when the electric car leaves, and there are J_iA period corresponding charging load Nargin M_t≥P_i+P₀The case where, then it needs to adjust low ebb electricity price start periods to there is J thereafter_iPeriod is abundant to inductive charging load Spend M_t≥P_i+P₀Position, required for ensuring that the SOC of electric car can be charged to by user's system after responding low ebb electricity price It is horizontal；Specifically, it is adjusted according to start periods of the following expression to low ebb electricity price:

Wherein,N is set of integers；

Step 4.7: the low ebb electricity price start periods adjusted are being calculatedAfterwards, the period is determined The charging electricity price of interior user is paddy electricity electricity price p_l, remaining period charging electricity price is peak electricity tariff p_hIt is autonomous by user to prompt user Selection immediately begins to charge, or is delayed to paddy electricity electricity price and starts to charge；Wherein, D is time parameter, takes 0~1；

If user's selection immediately begins to charge, arrange the electric car since the next period one by one negative to inductive charging Lotus nargin M_t≥P_iPeriod charging, until arrange J_iUntil a period；Simultaneously by the charging load margin of corresponding charge period A_tS_Tλ is updated to A_tS_Tλ-P_i；

If user's selection is delayed to paddy electricity electricity price and starts to charge, arrange the electric car fromStart, is filled one by one in correspondence Electric load nargin M_t≥P_i+P₀Period charging, until arrange J_iUntil a period；Simultaneously by the charging of corresponding charge period Load margin A_tS_Tλ is updated to A_tS_Tλ-P_i。

After the method for the present embodiment obtains charging strategy, it is completed by charging system for electric automobile as shown in Figure 8 and is filled Electric strategy, the system include transformer (current transformer and voltage transformer), intelligent electric meter, Intelligent charge control device, fill Electric stake and distribution transformer monitoring terminal.Transformer (current transformer and voltage transformer) refers to that being mounted on residential area matches The current transformer and voltage transformer of piezoelectric transformer low-pressure side, the input of the output end and distribution transformer monitoring terminal of the two End is connected；Intelligent charge control device is communicated in a manner of RS485 or RS232 with distribution transformer monitoring terminal, and is made The optimization charging plan of fixed each charging pile is simultaneously assigned to each charging pile, monitors the charged state of each charging pile, Mei Geyi The charging load of the charging pile for storing the charging load of the charging pile and being obtained of fixing time is sent to distributing transformer monitoring Terminal；Charging pile is communicated in a manner of power line carrier with electric car Intelligent charge control device；Distribution transformer prison The voltage and current etc. that terminal obtains somewhere distribution transformer low-pressure side is surveyed, whole use other than electric car are calculated Electric load and load curve, and it is sent to Intelligent charge control device.

Method of the invention in specific implementation, carries out following reasonable simplified and assumes:

(1) assume in this region it is all it is for electric vehicle select to rest on electric car in charging station per family to charge, The place of parking and charge of electric car is fixed；

(2) it is furnished with scheduling system in charging station, whenever thering is new electric car to drive into charging station, and access charging pile, System can read this batteries of electric automobile automatically and hold b_nState-of-charge S is originated with electric car_{N, S}Deng, meanwhile, in order to formulate Corresponding optimisation strategy, automobile user need to input the expected downtime T of this electric car to system_{N, E}With it is electronic When automobile leaves charging station, user it is expected the electric car state-of-charge S reached_{N, E}；

(3) based on the simulating analysis of Monte Carlo simulation；In practice, for electric car enter the charging station time, Many useful informations such as starting state-of-charge, expected downtime are all unforeseen, calculating for convenience, the present embodiment utilization Monte carlo method, is randomly generated how required electric car charge data, and Monte Carlo Analogue Method is as shown in Figure 9.

Monte Carlo method is a kind of method that physics and mathematical problem are solved using duplicate statistical experiment.Use Meng Teka When Lip river method handles problem, answer is often configured to the mathematic expectaion of some stochastic variable, passes through certain number on computers It carries out imagination test and obtains the arithmetic mean of instantaneous value of the specific value of stochastic variable, and use it as the approximate solution of problem.

1. the determination of random number generates random number using mixed congruence method, and carries out randomness test, to obtain more True random number, the recurrence formula of mixed congruence method are x_i=mod (Ax_i-1+ C, M)；In formula, mod is remainder function, and A is to multiply Son, M are mould, C is increment, and three is positive integer, initial value x₀When, then can recursion go out x₁, x₂..., x_n.By this numerical value sequence Column can obtain the equally distributed random number sequence r in [0,1] section divided by M_i.Randomness test for random number sequence, will successively The N number of random number occurred, is divided into two classes or k class by its size, checks whether the appearance of all kinds of numbers has coherent phenomena, so that it is determined that True random number.

The Load Calculation Method 2. electric car charges, first has to the charging load for acquiring all kinds of electric cars, and succession is tired Count available total charging load curve.The unit of charging carry calculation is day, one day is divided into 1440 minutes, when i minutes When, total charging load is whole all vehicles, and at this moment, charging load can be expressed as follows:

In formula, L_iFor i-th minute total charge power, i=1,2 ..., 1440；N is electric car total amount number；P_{N, i}It is Charge power of the n vehicle at i-th minute.

The load model 3. electric car charges, the where the shoe pinches for calculating electric car charging load are the initiation of charge time With the randomness of starting SOC.Assuming that power grid can not play the role of decision for the charging behavior of electric car, when being linked into electricity It begins to charge after net, extracts each automobile initiation of charge time, starting SOC.The initiation of charge time is according to difference Vehicle different from, starting SOC then meet normal distribution.After determining different automobile types charging load model and its parameter, need Quick charge and normal charge are modeled respectively, it includes the following contents that system, which inputs information, such as electric car total amount, different chargings The incidence of behavior, possible charging time section, constraint of charging duration etc., including the corresponding starting SOC of different type charging behavior Probability distribution.

Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used To modify to technical solution documented by previous embodiment, or some or all of the technical features are equal Replacement；And these are modified or replaceed, model defined by the claims in the present invention that it does not separate the essence of the corresponding technical solution It encloses.

Claims (6)

1. a kind of city electric car charging decision method based on typical load dynamic game, it is characterised in that: this method by One transformer branch is installed a total intelligent scheduling device and is realized, the n of the intelligent scheduling device control transformer branch A charging pile, when there is electric car access, charging pile uploads electric car information and is sent to intelligent scheduling device, intelligent scheduling Device sends charging strategy after analysis and charges to charging pile, between intelligent scheduling device and charging pile by TCP into Row communication；The intelligent scheduling device obtain the specific method of charging strategy by analysis the following steps are included:

Step 1: short-term electric load being clustered by improved AHPK-means algorithm, predicts the typical case of certain regional power grid Load curve, the typical load curve of power grid are the load curve removed except electric car charging load, are included the following steps:

Step 1.1: the load data that terminal user is chosen from the database of Utilities Electric Co.'s SCADA system in certain city, which is constituted, to be born Lotus matrix A；The load data includes the user type of a branch of some transformer of city, active power, idle The measured value of power and voltage and current, every 15min sampling is primary, samples 96 data points altogether daily；

Step 1.2: data prediction, identification and amendment abnormal data being carried out to matrix of loadings A, and are normalized；

Step 1.3: clustering is carried out to step 1.2 treated matrix of loadings A using AHPK-means algorithm, with bilayer plus Power Euclidean distance is that similarity criteria is clustered；

Step 1.4: Cluster Validity is examined；By establishing Validity Index, evaluation cluster result, and determine preferable clustering number, Obtain final short term curve；

Step 2: electric car ideal charging load curve is calculated by obtained short term curve, comprising the following steps:

Step 2.1: it is short-term negative at what there were significant differences that the class class center N being obtained by step 1.4 processing Power system load data matrix A Lotus curve；

Step 2.2: the behavior and composition of user is obtained by obtained N class short term curve, and combine cluster result to The behavior at family is analyzed with composition, obtains N kind class of subscriber；

Step 2.3: calculating separately the maximum power value P of this N kind class of subscriber_max, minimal power values P_min, peak occur time Section t_h, low ebb occur time period t₁And peak-valley difference Δ p, wherein maximum power value P_max, minimal power values P_min, peak occur Time period t_h, low ebb occur time period t₁It is directly obtained by obtained short term curve, peak-valley difference Δ p is calculated by following formula It obtains: Δ p=P_max-P_min；

Step 2.4: electric car ideal charging load curve, i.e., the time period t occurred in low ebb are obtained by information above₁It fills Electricity and with maximum power value P_max, minimal power values P_minAs restrictive condition；

Step 3: with charge requirement constraint, the charging time constraint, electric car charging quantity limitation, grid power load curve, Power distribution network capacity is as boundary condition, with load curve and the minimum constraint condition of ideal charging load curve error of charging, with The minimum objective function of user power utilization cost, establishes optimization method, and solve, obtains charging strategy；

Step 4: judging whether to charge to electric car by the result of decision of dynamic game, that is, what is judged optimal fills Whether electric strategy has controllability, and the specific method is as follows:

Step 4.1: when in a certain grid branch with there is new electric car access in any one charging pile carried, charging station System information is updated to next 15min automatically according to charging pile operating status in standing and controls time point by operation system, and first Calculate the corresponding SOC state of this electric car, it is fully charged when required power, required charging time number of segment J_i, down time Number of segment T_iThe T left in the vehicle is counted from present period with system_iCharging load margin M in a period_t, M_t=A_tS_Tλ, In, t=1,2 ..., T_i, T_i=96；S_TIndication transformer rated capacity；A_tWith indicating to allow to fill in t-th of period in one day Power station accounts for the ratio of transformer capacity to the power that electric car charges；λ is the power factor of charging load；

Step 4.2: before formulating the preferential tou power price period towards the user, system prejudges that can meet this electronic The charge requirement of automobile counts the T left in the vehicle from present period by computing system_iCharging load in a period is abundant Spend M_tIt realizes: H_i=| A_i|, A_i=t | M_t≥P_i, t=1,2 ..., T_i}；Wherein, | A_i| it is set A_iInterior element number；

Step 4.3: if the M of corresponding period_t≥P_i+P₀, i.e., all electric cars can be arranged to charge in the period, otherwise use and seek Excellent algorithm reasonably arranges whether each automobile charges,；P_iRequired active power when having an electric car access for each charging pile Summation, P₀For current real-time active power, M_tIt is bigger to be worth the bigger charging load margin for indicating the corresponding period；

Step 4.4: if H_i< J_i, J_iIndicate the time hop counts that electricity is full of, then explanation is in the electric car berthing time System is unable to satisfy the charge requirement of all user's inputs, to arrange electric car to charge using optimizing algorithm, with charging Load margin M_tAs constraint condition, if charge power needed for certain electric car is more than the 80% of SOC, preferentially joined with χ Number is allocated, i.e. P_i=χ P_i, later with maxM_t=α P₁+βP₂+...+φP_nIt is solved as target equation, wherein P_iFor The charge power of electric car configured, α, β ..., φ be specific gravity factor, ask and be satisfied the electronic of optimal charge demand Automobile quantity, and learn which charging pile can charge for electric car, while showing that user should be carried out with great power Charging；

Step 4.5: the result of step 4.4 is analyzed, if cannot charge for electric car, prompt user from Open, if can charge for electric car, remind user that can charge for its electric car, and be calculated with When family is left, the batteries of electric automobile state-of-charge of system maximum satisfaction:Wherein,Table Show that initial battery capacity, η are power parameter, Δ t is the when length from starting to charge to leaving, B_iFor electronic vapour battery capacity；

Since the electric car needs to charge always in berthing time, charging station prompts custom system can the maximum battery met State-of-chargeAnd prompt user by peak electricity tariff p_hIt charges, independently selected to receive charging service by user or abandons filling Electricity service；If user receives charging service, arrange electric car in charging load margin M_tGreater than P_i+P₀Period filled Electricity, and will charging load margin A in the corresponding period_tS_Tλ is updated to A_tS_Tλ-P_i；

Step 4.6: if H_i≥J_i, then explanation system in the electric car berthing time can meet the charging of all electric cars Demand, and on the basis of meeting user's charge requirement, peak load shifting is realized to the maximum extent；Therefore, charging station orderly fills The electric continuous J of coordination system initial option_iThe minimum value of the maximum start periods of the sum of a load margin is the starting of low ebb electricity price Period；Specifically, according to the start periods of following expression primary Calculation low ebb electricity price:

Consider fromPeriod starts when the electric car leaves, and there are J_iA period corresponding charging load margin M_t≥P_i+P₀The case where, then it needs to adjust low ebb electricity price start periods to there is J thereafter_iPeriod is to inductive charging load margin M_t ≥P_i+P₀Position, to ensure that the SOC of electric car can be charged to required level by user's system after responding low ebb electricity price； Specifically, it is adjusted according to start periods of the following expression to low ebb electricity price:

Wherein,N is set of integers；

Step 4.7: the low ebb electricity price start periods adjusted are being calculatedAfterwards, the period is determinedInterior use The charging electricity price at family is paddy electricity electricity price p_l, remaining period charging electricity price is peak electricity tariff p_h, to prompt user, independently selected by user It selects and immediately begins to charge, or be delayed to paddy electricity electricity price and start to charge；Wherein, D is time parameter, takes 0~1；

If user's selection immediately begins to charge, arrange the electric car since the next period one by one abundant to inductive charging load Spend M_t≥P_i+P₀Period charging, until arrange J_iUntil a period；Simultaneously by the charging load margin of corresponding charge period A_tS_Tλ is updated to A_tS_Tλ-P_i；

If user's selection is delayed to paddy electricity electricity price and starts to charge, arrange the electric car fromStart, one by one negative to inductive charging Lotus nargin M_t≥P_i+P₀Period charging, until arrange J_iUntil a period；Simultaneously by the charging load of corresponding charge period Nargin A_tS_Tλ is updated to A_tS_Tλ-P_i。

2. the city electric car charging decision method according to claim 1 based on typical load dynamic game, special Sign is: the step 1.2 further includes steps of

Step 1.2.1: the abnormal data in matrix of loadings A is identified and is corrected；Actual load data can be due to some survey It measures the prolonged failure of unit or the failure of other related elements lacks load data in data acquisition, for 96 Data point, the record that the active record data of record, load of missing wherein 30 and above data point are negative, directly kicks and removes；

Step 1.2.2: for rising sharply in load curve, rapid drawdown data, being judged by calculated load change rate, and load becomes Rate calculation formula are as follows:

ρ=| (p_d-p_d-1)/p_d|；

Wherein, ρ indicates load changing rate of the customer charge sequence P in d point, P={ p_d, d=1 ..., 96 }, p_dFor the active of d point Power；When load changing rate ρ is more than preset threshold ε, then it is considered as abnormal data；

Load data for abnormal data and containing missing is filled up as the following formula using smooth window:

Wherein, p_d' it is the active power for filling up rear d point, a, b, which are respectively indicated, forwardly and rearwardly to be taken a little, and maximum gets a respectively₁、b₁；

Step 1.2.3: being normalized to by identification with modified matrix of loadings A using maximum value method for normalizing, Emphasize that the similitude of load tendency, normalization expression formula are as follows in the base lotus part of elimination capacity curve:

x_d=p_d′/max(P)；

Wherein, x_dFor the active power of d point after normalization, max (P) is the maximum power in 96 points.

3. the city electric car charging decision method according to claim 2 based on typical load dynamic game, special Sign is: the step 1.3 further includes steps of

Step 1.3.1: the class of subscriber and quantity of the branch, including industrial user's number h are obtained₁, resident living user's number h₂, school user number h₃, commercial user's number h₄, public become user's number h₅；It is calculate by the following formula optimization cluster numbers k:

Wherein, f () is normal function, c_iFor constant,For the transposition of triangular matrix, h_iFor i-th kind of class of subscriber quantity, i= 1,2 ..., n, n=5, ε are measurement error；

Step 1.3.2: the K-means algorithm based on double-weighing Euclidean distance, method particularly includes:

Step 1.3.2.1: the optimization cluster numbers k that step 1.3.1 is obtained is as initial cluster center；

Step 1.3.2.2: sample is sorted out；All center of a sample of k class are divided into the nearest class center of its weighted euclidean distance, sample This A_kTo j-th of cluster centre m_j={ m_{J, 1}, m_{J, 2}..., m_{J, q}Weighted distance calculated by following formula:

Step 1.3.2.3: cluster centre updates；According to step 1.3.2.2's as a result, calculating the average value of every class as all kinds of new Cluster centre；

Step 1.3.2.4: iterative calculation；Judge whether cluster centre restrains, if not converged, return step 1.3.2.2, otherwise It performs the next step；

Step 1.3.2.5: all cluster centres are combined into a new data set, each cluster centre itself is considered as one Class calculates all between class distances, i.e. similarity between sample；

Step 1.3.2.6: by established rule choose wherein between class distance reach requirement classification carry out class between union operation；

Step 1.3.2.7: for class C_iAnd C_j, distance of the average distance at two class centers as the two classes is selected, is then chosen The smallest class of class spacing merges:Wherein, D_ijFor C_iAnd C_jIt is after merging as a result, n_i、n_j Respectively class C_iAnd C_jNumber of samples, x, x ' be class center, d (x, x ') be two class centers average distance；

Step 1.3.2.8: the new class of calculating previous step generation and the before similarity between class；

Step 1.3.2.9: repeating step 1.3.2.6-1.3.2.8, until all load samples are divided into one kind, algorithm terminates.

4. the city electric car charging decision method according to claim 3 based on typical load dynamic game, special Sign is: the step 1.4 uses Dai Weisenbaoding index i.e. DBI index, for determining preferable clustering number and evaluating cluster matter Amount, DBI index I_DBISmaller expression Clustering Effect is better, calculation formula are as follows:

Wherein, K is cluster numbers, R_kThe sum of inter- object distance average distance for any two classification is divided by two cluster centre distances, d (X_k) and d (X_j) be k and j inter- object distance average distance, X_k、X_jRespectively indicate k and j class data, d (c_k, c_j) it is in two clusters Heart distance, c_k、c_jRespectively indicate the center of k Yu j class；

Clustering is carried out using AHPK-means, is repeated 20 times, wherein I is chosen_DBICorresponding solution is best poly- when minimum Class result.

5. the city electric car charging decision method according to claim 4 based on typical load dynamic game, special Sign is: class of subscriber number N=4 obtained in the step 2.2, and 4 kinds of class of subscribers are respectively as follows:

(1) evening peak type user group: the load variations trend of its user meets the form of common daily load curve, i.e., from morning When 5, load constantly rises, and starts to be maintained at a higher level when to morning 9, and lunch break when noon 12 leads to electricity consumption Power is reduced on a small quantity, later from afternoon 3 when at night 8 when electricity consumption persistently rise, and reach the top of whole day electricity consumption, Zhi Houyong Electricity has beginning slowly decline；

(2) single peak type user group: there is a peaks of power consumption for the load variations trend of its user, and on integrally taking advantage of a favourable situation, Peak and the amplitude difference of electric power corresponding to low ebb are little, when morning 5, at night 20 points continue relatively smoothly Electricity consumption；

(3) leveling style user group: the load variations trend of its user is more gentle, and one day load variations is little, but gentle class The power load active power level of user group is very high, and load is constantly in higher level；

(4) keep away peak type user group: the load variations trend of its user and the difference highly significant of other users, with electrical characteristics with Other four classes have significant difference, show two head heights, it is intermediate it is low keep away peak type feature, 6:00 AM or so stops electricity consumption, Electricity consumption quickly reduces and is maintained at a lower level, until 6 points of night, remaining time is then in peak times of power consumption, peak value, Peak-valley difference is also highest in all classes.

6. the city electric car charging decision method according to claim 4 based on typical load dynamic game, special Sign is: the step 3 further comprises following steps:

Step 3.1: initialization same day power distribution network load information；

Step 3.2: charging station system judges whether there is new electric car and drives into charging station；If so, reading all new accesses The useful data information of electric car, comprising: obtain batteries of electric automobile capacity, the current state-of-charge of battery of electric vehicle, electronic The charge power curve and electric vehicle of residence time, batteries of electric automobile expected from automobile when leaving desired electric car it is charged Status level；If not provided, continuing to use the charge mode of a period；

Step 3.3: according to the expected downtime of charging station electric car in the period, obtaining all vehicle dwell times Maximum value t_M, and this suboptimization time span T is calculated,| x | for the smallest positive integral less than x；

Step 3.4: turning to target so that electric car charging is optimal, according to system information, formulate new charge control strategy；To fill Electricity demanding constraint, charging time constraint, electric car charging quantity limitation, grid power load curve, power distribution network capacity conduct Boundary condition, with load curve and the minimum constraint condition of ideal charging load curve error of charging, most with user power utilization cost Low is objective function, establishes optimization method；

The charge requirement is constrained toWherein, x_{N, t}T moment when being accessed for n-th automobile Point active power, S_{N, S}For initial state-of-charge, b_nFor charge a period increasable battery SOC numerical value, in T period Interior, the battery charge state for the electric car being electrically charged should be at least up to required final state-of-charge when charging starts S_{N, E}, while should stop charging in the case where fully charged；

The charging time is constrained to t_{N, E}≤T_{N, E}, the electric car needs being electrically charged are in expected downtime set by user Charging complete；Wherein t_{N, E}It charges the end time for n-th electric car；T_{N, E}For the expected charging of automobile user setting End time；

The electric car charging quantity is limited toWherein, n_tFor the active power of n-th electric car t moment, Charging pile limited amount in charging station, Rechargeable vehicle quantity is limited by charging pile quantity in each period；X is in charging station Charging pile quantity；

The peak- valley load constraint is | P_max1-P_min1| < Δ P；Wherein, P_max1And P_min1Respectively to current excellent since morning on the same day Changing the period terminates system loading maximum value and minimum value in this period；In conjunction with the peak valley difference value of the period in past 7 days, Δ P initial value is set to the period peak-valley difference minimum value in this seven days, if the value causes optimization aim without solution due to less than normal, at the beginning of Δ P Value progressively increases 1%, until there is solution；

Step 3.5: the selection for time started of charging uses nonlinear optimization method, solves nonlinear equation by MATLAB, seeks The suitable charging time started is looked for, fluctuates total load curve minimum, in this, as the objective function of optimization, specific method Are as follows:

Step 3.5.1:T time internal loading variance is minimum；After a charging load accesses, if at a time starting to charge, Total load curve after making to be added the load is minimum in T period internal variance, then the moment be the charging time started selected；For Continuous function calculates its 2 rank center away from variance is indicated, can obtain

Wherein, P_varFor power distribution network daily load variance, P_avFor the per day load before adjustment, P_EtFor always filling for t period electric car Electrical power, P_LtThe conventional load of t period when for power distribution network without electric car charging load, that is, obtained by load prediction It arrives；

Step 3.5.2: in conjunction with the case where tou power price, with the minimum objective function of user power utilization cost, i.e.,Wherein, S_jPower grid is represented in the electricity price at j moment, indicates electric car charging electricity price for positive value, The subsidy electricity price that automobile user is fed to power grid is then represented for negative value；P_ijNeeded for indicating i platform automobile within the j period Power；

Step 3.5.3: linear weight sum method is used, converts single object optimization for the optimization problem of above-mentioned 2 objective functions, together When, since the dimension of two objective functions is different, each function is normalized respectively, such as following formula institute of the objective function after conversion Show:

minT₁=ω₁(T₃/T_3max)+ω₂(T₄/T_4max)；

In formula, T_3maxFor the mean square deviation of original grid load curve；T_4maxIt is accustomed to the day charging cost of lower car owner for tradition vehicle, Battery is full of required cost by minimum electricity in down time；T₃、T₄It respectively represents in step 3.3.1 and step 3.3.2 Two objective functions, ω₁、ω₂Respectively T₃、T₄Weight coefficient, and meet ω₁+ω₂=1；

Step 3.6: the maximum power value P in N kind class of subscriber for respectively obtaining step 2.3_max, minimal power values P_min, it is high The time period t that peak occurs_h, low ebb occur time period t_lAnd peak-valley difference Δ p is brought into the optimization method of step 3.5 foundation, Similarity judgement is carried out respectively with obtained N kind typical load curve data by real-time data of power grid, is judged if there is electronic When automobile accesses, it should which kind of ideal charging load curve used, in the hope of the optimal charging strategy of every kind of class of subscriber；

Similarity judgment formula is shown below:

Wherein, K (x)^newFor the real time data on the same day,For q kind typical load data, q=1,2 ..., N.