A load and traffic flow distribution-based electric vehicle charging station planning method

Abstract

The application relates to a kind of electric vehicle charging station planning methods based on load and traffic distribution, which considers the superimposed influence of topographic data, historical load data and historical traffic data in planning, and the fixed load distribution of charging station and electric vehicle charging load are predicted by historical load data, and the prediction result is combined with the analysis of topographic data, and the grid map of distribution network is optimized by using distribution optimization model, then the grid area load division is carried out, to more accurately describe the influence of topographic data and historical load data on the grid planning layout of charging station, and the grid area traffic heat distribution is analyzed by combining the grid optimization map of distribution network and historical traffic data, finally, the superimposed influence of grid area load division and grid area traffic heat division result is considered, to determine the capacity planning scheme of electric vehicle charging pile, to better guarantee the accuracy and rationality of electric vehicle charging station planning.

Description

Technical Field

[0001] This invention relates to the fields of power system analysis technology and distribution network technology, specifically to a method for planning electric vehicle charging stations based on load and traffic flow distribution. Background Technology

[0002] Charging infrastructure planning refers to the systematic arrangement and design process of providing charging services for electric vehicles within a specific area. This process involves considering factors such as determining the location of charging stations, the number of charging piles, and the construction of related infrastructure. The goal of charging infrastructure planning is usually to meet the charging needs of electric vehicles, increase the coverage of charging facilities, and promote the development of electric transportation.

[0003] Charging station planning is a systematic process that requires comprehensive consideration of multiple factors, including demand, location, power supply, and user experience, to ensure that charging facilities meet user needs, achieve sustainable development, and comply with local environmental and regulatory requirements. The primary goal of charging station facility planning is to ensure that electric vehicles have access to effective, convenient, and sustainable charging services, establishing a reliable, efficient, and sustainable charging infrastructure to promote the development of electric transportation and meet the growing market demand for electric vehicles. However, charging station facility planning is highly uncertain, including uncertainties in user-side load demand, power supply, and the location of charging stations, which further complicates the problem of solving charging station facility planning issues.

[0004] Charging station facility planning technology is constantly evolving, involving multiple fields and typically considering geographical information factors, historical and forecast data, and power system analysis. However, the coupling degree between transportation networks and power transmission networks is still not high enough, leading to problems such as excessively long planning and construction cycles and insufficient rationality in charging station facility planning. Therefore, how to design a planning scheme for the location and capacity of charging piles in urban electric vehicle charging stations that considers multiple factors to improve the efficiency and rationality of its implementation planning and solve problems such as difficulty in planning solutions, long planning cycles, and insufficient planning rationality has become an urgent research direction. Summary of the Invention

[0005] To address the shortcomings of the existing technologies, this invention provides a method for planning electric vehicle charging stations based on load and traffic flow distribution, thereby improving the efficiency and rationality of electric vehicle charging station planning and solving the problems of difficulty in solving electric vehicle charging station planning in urban areas with complex terrain, long planning cycles, and insufficient planning rationality in the existing technologies.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] A method for planning electric vehicle charging stations based on load and traffic flow distribution includes the following steps:

[0008] S1. Obtain terrain data, historical load data, and historical traffic flow data for the area to be planned; the terrain data includes slope information and a planar power grid map; the historical load data includes fixed load demand data and electric vehicle charging load demand data.

[0009] S2. Based on the historical load data, predict the fixed load and electric vehicle charging load to obtain the fixed load distribution prediction results and the electric vehicle charging load prediction results.

[0010] S3. Based on the slope information, the power distribution grid map, and the obtained fixed load distribution prediction results and electric vehicle charging load prediction results, the preset power distribution optimization model is used to optimize the power distribution grid map with the total cost minimization as the objective function.

[0011] S4. Based on the fixed load distribution prediction results and the electric vehicle charging load prediction results, the grid areas in the power distribution network grid optimization map are divided into load levels to obtain the grid area load division results.

[0012] S5. Based on the historical traffic flow data, the grid areas in the power distribution network grid optimization map are classified into traffic heat levels to obtain the grid area traffic heat classification results.

[0013] S6. Based on the grid area load division results and grid area traffic heat division results, obtain the electric vehicle charging pile capacity planning scheme for each different grid area in the area to be planned.

[0014] S7. Based on the obtained electric vehicle charging pile capacity planning scheme, plan the capacity of electric vehicle charging piles distributed in different grid areas within the planning area.

[0015] As a preferred option, step S2 specifically involves:

[0016] S201. Decompose the fixed load demand data and electric vehicle charging load demand data in the historical load data into internal model functions to obtain the fixed load internal model function components and the electric vehicle charging load internal model function components respectively.

[0017] S202. Using the ensemble empirical mode decomposition method, the internal mode function components of each fixed load and the internal mode function components of the electric vehicle charging load are decomposed to obtain the decomposed fixed load time series and electric vehicle charging load time series.

[0018] S203. Based on the decomposed fixed load time series and electric vehicle charging load time series, determine the information entropy (IE) of each fixed load internal model function component and electric vehicle charging load internal model function component, and reconstruct the fixed load extended internal model function component and electric vehicle charging load extended internal model function component with similar information entropy (IE); the similar information entropy (IE) means that the difference between the two information entropy (IE) is within a preset difference range.

[0019] S204. The information entropy IE of the fixed load internal model function component and the fixed load extended internal model function component is used as the input of the preset LSTM model for prediction to obtain the first fixed load distribution prediction result; the information entropy IE of the electric vehicle charging load internal model function component and the electric vehicle charging load extended internal model function component is used as the input of the LSTM model for prediction to obtain the first electric vehicle charging load prediction result.

[0020] S205. The information entropy IE of the fixed load internal model function component and the fixed load extended internal model function component is used as the input of the preset GMDH model for prediction to obtain the second fixed load distribution prediction result; the information entropy IE of the electric vehicle charging load internal model function component and the electric vehicle charging load extended internal model function component is used as the input of the GMDH model for prediction to obtain the second electric vehicle charging load prediction result.

[0021] S206. The first fixed load distribution prediction result and the second fixed load distribution prediction result are weighted and summed to obtain the final fixed load distribution prediction result; the first electric vehicle charging load prediction result and the second electric vehicle charging load prediction result are weighted and summed to obtain the final electric vehicle charging load prediction result.

[0022] As a preferred option, step S206 specifically involves:

[0023] First, the first weight value of the weighted sum of the first fixed load distribution prediction result and the second fixed load distribution prediction result, and the second weight value of the weighted sum of the first electric vehicle charging load prediction result and the second electric vehicle charging load prediction result are optimized by the bitter fish optimization algorithm. Then, the optimized first weight value and second weight value are used to calculate the fixed load distribution prediction result and the electric vehicle charging load prediction result by weighted summation.

[0024] As a preferred option, in step S206, for any one of the first weight value and the second weight value, the specific method of optimization using the bitter fish optimization algorithm is as follows:

[0025] S2061. Randomly generate n initial weight values ​​as individuals of the bitter fish population, i = 1, 2, ..., n, to obtain the initial bitter fish population Pop, i.e.:

[0026] Pop = [F i |i=1,2,…,n],F i =[f i,k |k=1,2,…,m];

[0027] Among them, F i ∈Pop represents any i-th individual bitterfish in the bitterfish population, and the individual bitterfish population is the weight value to be optimized; f i,k For any i-th individual bitterfish in the bitterfish population, F i The value in the k-th dimension; n is the number of bitter fish individuals in the bitter fish population in the bitter fish optimization algorithm, and m is the dimension of each bitter fish individual in the bitter fish population;

[0028] Then, initialize the iteration count t=1 and proceed to step S2062;

[0029] S2062. Randomly select half of the bitter fish individuals in the bitter fish population as positive individuals and label them as F. i (+), the other half of the bitter fish individuals are negative individuals and are marked as F. j (-);

[0030] S2063, For each positive individual F i (+) is updated iteratively in the following manner:

[0031]

[0032] In the formula, and They are positive individuals F i (+) represents the position of (+) in the t-th and t+1-th iterations; It is a randomly selected oyster location; is the optimal solution for the individual bitterfish in the t-th iteration, that is, the bitterfish individual closest to the oyster in the t-th iteration; δ and r are random numbers between 0 and 1; J(t) is the number of steps or speed at which the bitterfish individual moves to escape or approach the oyster, and:

[0033]

[0034] J(1) is the initial step size and jump value for each bitter fish; t max This is the preset maximum number of iterations; U(t) is a random function, and:

[0035] U(t+1)=cos(t×cos -1(U(t)));

[0036] U(1) equals 1; the probability parameter P is:

[0037]

[0038] Where t is the number of iterations; a represents the power of restoration; rand is a random number;

[0039] S2064, For each negative individual F j (-) is updated iteratively in the following manner:

[0040]

[0041] Where [l,u] represents the range of individual bitter fish values ​​in the initial bitter fish population, with l and u being the lower and upper limits of the individual bitter fish values, respectively; M represents the average position of the fish population.

[0042]

[0043] S2065, each positive individual F i The newly created bitter fish (+) is represented as:

[0044]

[0045] Indicates a positive individual The newly created bitter fish; R represents the distribution radius of the bitter fish individual around the oyster shell, with a value between [0,2].

[0046] The probability of a newly hatched bitterfish being hunted is:

[0047]

[0048] Let f(·) be the probability that a newly hatched bitter fish individual will be hunted, and let f(·) be the objective function solved by the bitter fish optimization algorithm.

[0049] Count the number K of newly hatched bitter fish individuals that survived after being hunted, and then use K of the remaining surviving newly hatched bitter fish individuals Replace the oyster location K negative individuals This results in the formation of a new population of bitter fish in the (t+1)th iteration;

[0050] S2066. Let the iteration number t = t + 1, and then return to step S2062;

[0051] S2067. Repeat steps S2062 to S2066 until the preset maximum number of iterations t = t is reached. maxThe optimal solution for the bitter fish individual in the final iteration is obtained, which is the bitter fish individual closest to the oyster position after the last iteration, and is used as the optimized weight value.

[0052] As a preferred embodiment, in step S3, the objective function expression of the power distribution optimization model is:

[0053]

[0054] Among them, C totle C represents the total cost; grid,i C represents the construction cost of the distribution network for the i-th grid; install,i C represents the installation cost of charging stations in the i-th grid. trans,i Let be the power transmission loss cost of the i-th grid; M is the number of grids contained in the distribution network grid map;

[0055] The construction cost C of the distribution network in the i-th grid grid,i and power transmission loss cost C trans,i Calculate using the following formulas respectively:

[0056] C grid,i =α1·L grid,i +α2·θ i ;

[0057] Among them, L grid,i θ represents the length of the power distribution line in the i-th grid. i Let α1 and α2 be the average slope angle of the i-th grid, and α1 and α2 be coefficients used to represent the weight of the distribution network construction and the slope angle.

[0058] C trans,i =γ·P load,i ·L grid,i ;

[0059] Among them, P load,i L represents the load of the i-th grid. grid,i γ represents the length of the power distribution line in the i-th grid, and γ represents the proportional coefficient of power transmission loss.

[0060] The constraints of the power distribution optimization model include:

[0061] 1) Load balance constraints:

[0062]

[0063] Among them, P load,i P represents the load of the i-th grid. totle This indicates the total load demand.

[0064] 2) Slope angle constraint:

[0065] θi ≥θ max ;

[0066] Where, θ i Let θ be the average slope angle of the i-th grid. max This represents the maximum permissible slope angle for the planned area.

[0067] 3) Traffic flow constraints:

[0068] T grid,i ≥T min ;

[0069] Among them, T grid,i T represents the minimum traffic flow in the i-th grid. min This indicates the minimum traffic flow set within the planned area.

[0070] As a preferred option, step S4 specifically involves:

[0071] S401. Based on the fixed load distribution prediction results and the electric vehicle charging load prediction results, the grid areas in the distribution network grid optimization map are divided into load levels; the load levels include light load areas, medium load areas, even load areas, heavy load areas, and overload areas, and the load level division method is as follows:

[0072] The grid area is divided into light load areas, with the percentage of fixed load ranging from [0% to 30%] and the percentage of electric vehicle charging load ranging from [0% to 10%].

[0073] The grid area is divided into medium load areas, with the percentage of fixed load ranging from (30% to 50%) and the percentage of electric vehicle charging load ranging from (10% to 20%).

[0074] The grid area is divided into uniform load areas, with the percentage of fixed load ranging from (50%, 70%) and the percentage of electric vehicle charging load ranging from (20%, 30%).

[0075] The grid area is divided into heavy load areas, with the percentage of fixed load ranging from (70% to 100%) and the percentage of electric vehicle charging load ranging from (30% to 40%).

[0076] The grid area is divided into overload zones, with the percentage of fixed load ranging from >100% and the percentage of electric vehicle charging load ranging from >40%.

[0077] S402. According to the load classification standard, fill the grid of the grid division result with grid colors. The chromaticity for filling the overload area is: R: 90%, G: 10%, B: 10%; the chromaticity for filling the heavy load area is: R: 100%, G: 0%, B: 0%; the chromaticity for filling the evenly loaded area is: R: 80%, G: 20%, B: 20%; the chromaticity for filling the medium load area is: R: 70%, G: 30%, B: 30%; and the chromaticity for filling the light load area is: R: 100%, G: 30%, B: 0%.

[0078] S403. Based on the above steps, the grid area load division results are obtained.

[0079] As a preferred option, step S5 specifically involves:

[0080] S501. Based on the historical traffic flow data, classify the grid areas in the power distribution network grid optimization map into traffic heat maps; the traffic heat maps include completely unobstructed areas, basically unobstructed areas, normally operating areas, slow-moving areas, congested areas, and heavily congested areas, and the method for classifying traffic heat maps is as follows:

[0081] The grid area is divided into a completely unobstructed area, and the traffic heat value ranges from [0,50] vehicles / hour.

[0082] The grid area is divided into a basically unobstructed area, and the traffic heat value ranges from (50, 200) vehicles / hour.

[0083] The grid area is divided into normal operating zones, and the traffic heat value ranges from (200, 500) vehicles / hour.

[0084] The grid area is divided into slow-moving zones, and the traffic heat value ranges from (500, 1000) vehicles / hour.

[0085] The grid area is divided into congested areas, and the traffic heat value ranges from (1000, 2000) vehicles / hour.

[0086] For grid areas classified as heavily congested areas, the traffic heat value range is >2000 vehicles / hour;

[0087] S502. According to the traffic heat map classification standard, fill the grid division results with grid colors. The color values ​​for filling heavily congested areas are: R: 43%, G: 0%, B: 100%; for congested areas, R: 29%, G: 0%, B: 51%; for slow-moving areas, R: 0%, G: 25%, B: 50%; for normally moving areas, R: 0%, G: 50%, B: 100%; for basically unobstructed areas, R: 0%, G: 100%, B: 70%; and for completely unobstructed areas, R: 0%, G: 40%, B: 60%.

[0088] S503. Based on the above steps, the traffic heat map results of the grid area are obtained.

[0089] As a preferred option, step S6 specifically involves:

[0090] S601. Based on the grid area load division results and the grid area traffic heat division results, and taking the division results of each grid area in the area to be planned as the basis, the chromaticity of the grid area load division results and the grid area traffic heat division results corresponding to each grid area is superimposed.

[0091] S602. Based on the chromaticity overlay results of each grid area, plan the location and capacity of charging piles according to the following standards:

[0092] .

[0093] Compared with the prior art, the present invention has the following beneficial effects:

[0094] 1. This invention provides a planning method for electric vehicle charging stations based on load and traffic flow distribution. The planning process considers the combined effects of terrain data, historical load data, and historical traffic flow data. Historical load data is used to predict the fixed load distribution of charging stations and the charging load of electric vehicles. Combining the prediction results with terrain data analysis, a distribution optimization model is used to optimize the distribution network grid map and then divide the grid areas into load zones. This more accurately describes the impact of terrain data and historical load data on the gridded planning layout of charging stations. Simultaneously, the optimized distribution network grid map and historical traffic flow data are used to analyze the traffic heat distribution in the grid areas. Finally, considering the combined effects of the grid area load division and grid area traffic heat division results, a capacity planning scheme for electric vehicle charging piles is determined to better ensure the accuracy and rationality of electric vehicle charging station planning.

[0095] 2. The method of the present invention adopts a grid-based planning approach based on the grid map of the power distribution network. It not only fully considers the grid layout characteristics of charging stations in urban areas, but also simplifies the distribution information of charging stations in urban areas by using grid-based methods. This makes it easier to use the power distribution optimization model and the grid area division and superposition operation for effective calculation, thereby improving the effectiveness of the grid layout of charging stations in urban areas and the efficiency of charging station planning.

[0096] 3. In the method of the present invention, both the fixed load demand data and the electric vehicle charging load demand data are decomposed into multiple internal model function components. After decomposition by the ensemble empirical mode decomposition method, a neural network is used for prediction, which helps to solve the problems of high complexity, difficulty in solving and low prediction accuracy of electric vehicle charging station planning factors in urban areas with complex terrain. Attached Figure Description

[0097] To make the objectives, technical solutions, and advantages of the invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0098] Figure 1 This is a flowchart of the electric vehicle charging station planning method based on load and traffic flow distribution according to the present invention. Detailed Implementation

[0099] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but only to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0100] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0101] To address the challenges of solving electric vehicle charging station planning problems in complex urban terrain, such as long planning cycles and insufficient planning rationality, existing technologies disclose an electric vehicle charging station planning method based on load and traffic flow distribution, such as... Figure 1 As shown, it includes the following steps:

[0102] S1. Obtain terrain data, historical load data, and historical traffic flow data for the area to be planned; the terrain data includes slope information and a planar power distribution grid map; the historical load data includes fixed load demand data and electric vehicle charging load demand data.

[0103] In the specific implementation of this invention, the area to be planned can be an urban area with different terrains, including plains, mountains, and coastal areas. This invention is particularly suitable for planning electric vehicle charging stations in urban areas with complex terrain, such as mountainous and coastal areas, because it considers terrain data including slope information and a planar power grid map. In specific applications, terrain data can include elevation, slope angle, and Geographic Information System (GIS) data. This invention primarily uses slope angle data, while other terrain data can be used to correct the slope angle.

[0104] Unlike plains, mountainous and coastal areas, with their numerous gentle slopes and hills, require power transmission lines to be erected on a non-horizontal plane, making line layout constrained by the geographical environment. Line laying necessitates bending, crossing, and tilting based on building distribution and altitude variations. In the specific implementation of this invention, to obtain more accurate slope angle data, this embodiment combines spatial planning and specialized planning to extract elements such as ridgelines, slope angles, ecological protection zones, administrative boundaries, and existing medium-voltage distribution network lines. It also analyzes spatial planning and specialized planning to extract elements such as road networks, rivers, and functional clusters. Because the cost of medium-voltage distribution lines crossing mountains is relatively high (due to excessively high ridges and steep slopes), construction and maintenance are difficult, and the risk of natural disasters is also high. In the specific implementation of this invention, slope angles greater than 30° are directly used as grid boundaries; while for slope angles less than or equal to 30°, it is necessary to analyze the terrain slope angles to identify steep slope areas.

[0105] In the specific implementation of this invention, considering that in actual power distribution network scenarios, loads can be divided into fixed loads and dynamic loads; among them, fixed loads include residential loads and industrial loads, which are relatively stable; dynamic loads mainly refer to the dynamic loads of electric vehicle charging at various charging stations, which are relatively random. Therefore, this invention comprehensively considers both fixed and dynamic loads in the power distribution network scenario, and requires the collection of historical fixed load demand data and electric vehicle charging load demand data as historical load data to participate in prediction, in order to obtain the fixed load distribution prediction results and the electric vehicle charging load prediction results.

[0106] In the specific implementation of the present invention, the collected historical traffic flow data may include location information, speed and acceleration, vehicle remaining mileage, battery status and other data. In addition, other historical operating parameters of electric vehicles may be selected as historical traffic flow data according to actual needs and participate in the necessary planning considerations.

[0107] S2. Based on the historical load data, predict the fixed load and electric vehicle charging load to obtain the fixed load distribution prediction results and the electric vehicle charging load prediction results.

[0108] In specific implementation, step S2 includes:

[0109] S201. The fixed load demand data and electric vehicle charging load demand data in the historical load data are decomposed into internal model functions to obtain the fixed load internal model function components and the electric vehicle charging load internal model function components, respectively.

[0110] In the specific implementation of the present invention, both the fixed load demand data and the electric vehicle charging load demand data can be decomposed into multiple internal model function components, and then predicted separately, thereby solving the problems of high complexity, difficulty in solving, and low prediction accuracy of electric vehicle charging station planning factors in urban areas with complex terrain.

[0111] S202. Using the ensemble empirical mode decomposition method, the internal mode function components of each fixed load and the internal mode function components of the electric vehicle charging load are decomposed to obtain the decomposed fixed load time series and electric vehicle charging load time series.

[0112] This embodiment employs Ensemble Empirical Mode Decomposition (EEMD), which addresses the mode mixing problem inherent in traditional empirical mode decomposition. The decomposition principle of EEMD is as follows: when added white noise is uniformly distributed throughout the time-frequency space, this space is composed of components at different scales segmented by the filter bank. When a signal is added to a uniformly distributed white noise background, signal regions at different scales are automatically mapped to appropriate scales related to the background white noise. Each independent test may produce very noisy results because each added noise component includes both the signal and the added white noise. Since the noise is different in each independent test, the noise will be eliminated when the overall mean of enough tests is used. The overall mean will eventually be considered the true result. As more and more tests are conducted, the added noise is eliminated, and the only persistent and stable part is the signal itself. Therefore, by directly adding the inner mode function component of Gaussian white noise during the adaptive noise process of EEMD, the reconstruction error in the traditional empirical mode decomposition process is eliminated, forming a signal analysis and processing method for fully integrated empirical mode decomposition adaptive noise.

[0113] S203. Based on the decomposed fixed load time series and electric vehicle charging load time series, determine the information entropy (IE) of each fixed load internal model function component and electric vehicle charging load internal model function component, and reconstruct the fixed load extended internal model function component and electric vehicle charging load extended internal model function component with similar information entropy (IE); the similar information entropy (IE) means that the difference between the two information entropy (IE) is within a preset difference range.

[0114] In specific implementation, for any load time series in the constant load time series and the electric vehicle charging load time series, the specific process of constructing the extended internal model function components with similar information entropy IE through step S203 is explained as follows:

[0115] S2031, the white noise sequence W i Adding this to the decomposed load time series, a new load time series is constructed, represented as:

[0116] P (i) =P+β0E(W) (i) (1)

[0117] Where P is the load time series; E k (·) represents the k-th order IMF component generated after the internal model function component decomposition; N(·) represents the local average value of the solved signal, W (i) This represents a white noise sequence with a mean of 0; β is the standard deviation of the white noise.

[0118] S2032, the new load time series is processed using the ensemble empirical mode decomposition method to obtain the first set of residuals R1, expressed as:

[0119] R1=(N(P) (i) (2)

[0120] S2033, the first internal model function component d1 is obtained by subtracting the residual of the original data sequence, and is expressed as:

[0121] d1=P-R1 (3)

[0122] S2034, according to the order of the internal mode components in the load time series, select the internal mode components of white noise and add them to the residuals obtained in step S2032 to obtain the second set of residuals R2. Then, use ensemble empirical mode decomposition to process R1+βE(W) in the second set of residuals. (i) The decomposition is performed, where the second set of residuals R2 is expressed as:

[0123] R2=(N(R1+βE(W (i) (4)

[0124] d2=k1-k2 (5)

[0125] Where d2 represents the second component of the internal mold component;

[0126] S2035, Repeat steps S2031 to S2034 until the remaining residue R of the load time series. n Without further decomposition, we obtain a new decomposed load time series, represented as:

[0127]

[0128] S2036, Determine the IE of each IMF component based on the new load time series to reconstruct IMF components with similar IEs, using the following calculation formula:

[0129]

[0130] Where IE is the information entropy value, X is the load dataset; N is the number of data sets of the loaded load time series; P k It is the percentage of the k-th load data in the load time series dataset.

[0131] By reconstructing the extended internal model function components of fixed load and electric vehicle charging load with similar information entropy (IE), the basic data used for fixed load distribution prediction and electric vehicle charging load prediction can be reasonably expanded, the data prediction granularity can be refined, and the prediction accuracy can be improved.

[0132] S204. The information entropy IE of the fixed load internal model function component and the fixed load extended internal model function component is used as the input of the preset LSTM model for prediction to obtain the first fixed load distribution prediction result; the information entropy IE of the electric vehicle charging load internal model function component and the electric vehicle charging load extended internal model function component is used as the input of the LSTM model for prediction to obtain the first electric vehicle charging load prediction result.

[0133] LSTM (Long Short-Term Memory) neural networks are a special type of recurrent neural network (RNN), proposed by Hochreiter and Schmidhuber in 1997. They aim to solve the vanishing or exploding gradient problems encountered by traditional RNNs when processing long sequences of data. The core of an LSTM network is a three-gate mechanism: a forget gate, an input gate, and an output gate. These gates adaptively control the flow of information, thereby capturing long-term dependent information. Specifically, the forget gate forgets some information from the previous unit's output that might affect the load prediction result; the input gate determines the information stored in the unit with the new input; the storage unit stores useful information transferred from the previous LSTM unit; and the output gate outputs the processed information to the next LSTM unit. During training, the LSTM network continuously optimizes the weights in its units and shares weights across all units. Therefore, with fixed model parameters, the integral scale continuously changes at different times, solving the vanishing gradient problem while reducing the frequency of gradient explosion.

[0134] The expression calculated by the LSTM model, with IE as input, is as follows:

[0135]

[0136] In the formula, i t For input gate; x t The input at the current time step is the IE for each IMF component; σ is the tanh activation function, and the value of σ ranges from [-1, 1]; W xs W xi W xf W xo Both represent input X t The weight matrix H; t This is the output at the current moment; H t-1 It's the output from the previous moment; H xs H xi H xf H xo All represent Ht-1 The weight matrix; b i b f b o b s Both are deviation vectors; f t It's the Gate of Oblivion, s t This represents the input unit vector at the current time.

[0137] S205. The information entropy IE of the fixed load internal model function component and the fixed load extended internal model function component is used as the input of the preset GMDH model for prediction to obtain the second fixed load distribution prediction result; the information entropy IE of the electric vehicle charging load internal model function component and the electric vehicle charging load extended internal model function component is used as the input of the GMDH model for prediction to obtain the second electric vehicle charging load prediction result.

[0138] The GMDH (Group Method of Data Handling) model is a multi-layered feedforward neural network (up to eight layers) that uses polynomial iteration to obtain the non-linear relationship between input and output, hence it is also called a multinomial network. The GMDH neural network is hierarchical; its initialization does not require setting the number of neurons or network layers. It uses adaptive linear functions (polynomial functions) as neurons, and its training process is a continuous process of generating active neurons. These active neurons are selected and combined to generate neurons in the next layer, until an optimal prediction model is obtained. Neurons in the input layer are only responsible for transmitting the input signal to neurons in the intermediate layers; each neuron in the hidden layer and each neuron in the output layer is connected to only two neurons in the previous layer. Therefore, each neuron in each layer of the GMDH neural network is independent.

[0139] The expression calculated by the GDMH model, with IE as input, is as follows:

[0140] The input data is defined as P = (P1, P2, P3, ..., P...). N The predicted output is: The actual output is O i The relationships between the neurons in the GDMH are shown in the following formula:

[0141]

[0142] In the formula, O k,l This represents the output of the l-th neuron in the k-th layer of the network; O k-1,i This represents the output of the i-th neuron in the k-th layer of the network; O k-1,j This represents the output of the j-th neuron in the k-th layer of the network; O k-1i Indicates input (P) i), for each IMF component's IE, a k,1 b k,1 c k,1 d k,1 e k,1 All are polynomial coefficients.

[0143] The quadratic polynomial O of the GMDH model is expressed as:

[0144] O=Pδ (10)

[0145]

[0146] Where σ is the polynomial coefficient of the binomial and P is the matrix of the load time series;

[0147] To minimize the error between the predicted and actual outputs, the GMDH neural network obtains the coefficients of the quadratic polynomial in the above equation through regression analysis, and then optimizes the polynomial coefficients of the binomial using the least squares method. The optimization formula is as follows:

[0148]

[0149] in, For the desired output, O i This represents the actual predicted output; R represents the prediction error. The optimal quadratic polynomial is obtained when R is minimized. The polynomial coefficients β of the binomial are calculated using the least squares formula:

[0150] σπ=(P T P) -1 P T O (13)

[0151] Among them, P T This is the transpose of the matrix representing the load time series.

[0152] In the neural network of the GMDH model, any two neurons in a neural layer are used as input to generate neurons for a new neural layer. If the number of neurons m in the current neural layer is less than or equal to the size M of the preset elite pool, then the neurons in the current neural layer are not processed. If the number of neurons m in the current neural layer is greater than the size M of the preset elite pool, then the mean square error D of the neurons in the next layer is calculated. i The calculation formula is as follows:

[0153]

[0154] Among them, C N This represents the total number of neural layers, and the GMDH model is used to calculate the mean square error D of the output. iThe size of the set is used to sort each neuron in ascending order, and according to the preset elite pool size M, the top M neurons generated in the current neural layer are selected as input neurons for the next layer, and the other neurons in the current neural layer except for the M neurons are discarded, so as to simplify the structure of the neural network.

[0155] S206. The first fixed load distribution prediction result and the second fixed load distribution prediction result are weighted and summed to obtain the final fixed load distribution prediction result; the first electric vehicle charging load prediction result and the second electric vehicle charging load prediction result are weighted and summed to obtain the final electric vehicle charging load prediction result.

[0156] In specific implementation, step S206 is as follows: First, the first weight value of the weighted sum of the first fixed load distribution prediction result and the second fixed load distribution prediction result, and the second weight value of the weighted sum of the first electric vehicle charging load prediction result and the second electric vehicle charging load prediction result are optimized by the bitter fish optimization algorithm. Then, the optimized first weight value and second weight value are used to calculate the fixed load distribution prediction result and the electric vehicle charging load prediction result by weighted summation.

[0157] Bitterling Fish Optimization (BFO) is a swarm intelligence optimization algorithm inspired by nature. Bitterling fish are a typical example of intelligent behaviors employed by fish in nature to ensure survival. This algorithm models the Bitterling Fish Optimization algorithm based on the mating behavior of these fish. Experiments and implementations on various benchmark functions show that the Bitterling Fish Optimization algorithm is more accurate and achieves better optimization results than commonly used optimization algorithms such as the Gray Wolf Optimization Algorithm (Q-algorithm), Whale Optimization Algorithm, Butterfly Optimization Algorithm, and Harris Eagle Optimization Algorithm.

[0158] The BFO algorithm is inspired by the unique reproductive mechanism of the bitterling. This fish feeds on oysters, and finding suitable oyster shells for spawning is the male's responsibility. They need to find larger oysters with more space to accommodate their eggs. When these fish find one or more oysters, they will fight off their predators. The aggressive behavior of the male species stems from the fact that other male species also tend to claim oysters. In combat mode, the male's body color darkens, and it defends its territory to the best of its ability. During mating, a phenomenon called sexual selection occurs. In sexual selection, the female typically chooses a larger male with more vibrant colors. The female then lays her eggs inside the oyster, and the male fertilizes the eggs with sperm, resulting in new fry.

[0159] Specifically, in step S206, for any one of the first weight value and the second weight value, the optimization method using the bitter fish optimization algorithm is as follows:

[0160] S2061. Initialize the population;

[0161] Randomly generate n initial weight values ​​as individuals of the bitter fish population, i = 1, 2, ..., n, to obtain the initial bitter fish population Pop, i.e.:

[0162] Pop = [F i |i=1,2,…,n],F i =[f i,k |k=1,2,…,m];(15)

[0163] Among them, F i ∈Pop represents any i-th individual bitterfish in the bitterfish population, and the individual bitterfish population is the weight value to be optimized; f i,k For any i-th individual bitterfish in the bitterfish population, F i The value in the k-th dimension represents a decision variable; n is the number of bitter fish individuals in the bitter fish population in the bitter fish optimization algorithm, and m is the dimension of each bitter fish individual in the bitter fish population.

[0164] Then, initialize the iteration count t=0, proceed to step S2062, and start iterating the bitter fish population.

[0165] S2062, Classification of Individual Behavior in a Population;

[0166] Half of the individuals in the bitter fish population were randomly selected as positive individuals and labeled as F. i (+), the other half of the bitter fish individuals are negative individuals and are marked as F. j (-). Positive individual F i (+)∈Pop are the bitter fish individuals that are searching for and trying to catch oysters; negative individuals F j(-)∈Pop represents the bitter fish individuals that escaped or failed to catch oysters.

[0167] S2063, Search for and capture oysters;

[0168] Any male bitterling is eligible to be homed in areas with more oyster shells; bitterling individuals can roam in search of oysters, in which case, positive individuals F i (+) Target the oyster and move toward it; in this case, the target oyster is not noticed by the other fish and is F'd by this fish. i (+) Capture; Oysters are captured by positively influenced individuals F. i The state of capture and possession is represented by the following formula (16):

[0169]

[0170] In the formula, and They are positive individuals F i (+) represents the position of (+) in the t-th and t+1-th iterations; It is a randomly selected oyster location; δ is the optimal solution for the individual bitter fish in the t-th iteration, that is, the bitter fish individual closest to the oyster in the t-th iteration; δ and r are random numbers between 0 and 1. J(t) is the number of steps or speed at which the bitter fish moves to escape or approach the oyster. J(t) decreases as the number of iterations t increases. This is because after a period of time, the activity of the male fish decreases because they have successfully mated. As the number of iterations increases, decreasing J will cause the global search to decrease to a local search. The reduction of parameter J can be expressed by formula (17):

[0171]

[0172] J(1) represents the initial step size and jump value for each bitter fish. In practice, J(t) can be set to 1; t max The maximum number of iterations is preset; J(t) is the number of steps each fish takes to find an oyster in repetition t. A random sequence is generated using the random function U(t), denoted by (18):

[0173] U(t+1)=cos(t×cos -1 (U(t))) (18)

[0174] U(1) can be considered equal to 1; in summary, (17) and (18) can be represented by (19):

[0175]

[0176] In the bitter fish optimization algorithm, the optimal solution is repeatedly searched over time and in more space around the optimal solution. Therefore, it is necessary to reduce the probability parameter P over time to the second criterion of formula (16), where r > P. At this time, the probability parameter P can be expressed by formula (20):

[0177]

[0178] Where t is the number of iterations; a represents the power of restoration, which can be set to 0.9 in practice; and rand is a random number.

[0179] S2064, Escape behavior;

[0180] For negative individuals F j (-) To perform a mathematical simulation of escape, formula (21) can be used to describe individual bitter fish that escape or fail to catch oysters:

[0181]

[0182] Wherein, [l,u] represents the range of values ​​for individual bitter fish in the initial bitter fish population, and l and u are the lower and upper limits of individual bitter fish values, respectively.

[0183] M represents the average position of the fish school; that is, after the oysters escape, the bitter fish can only search for oysters in the average space, which can be expressed as a random optimal space search problem, represented by formula (22):

[0184]

[0185] S2065, Oviposition behavior;

[0186] After finding a suitable oyster, the male fish can attract the female fish to lay eggs in the oyster. The male fish fertilizes the eggs, thus creating new fish in the surrounding area. This process is represented by formula (23):

[0187]

[0188] Indicates a positive individual The newly created bitter fish; R represents the distribution radius of the bitter fish individual around the oyster shell, which takes a value between [0,2]. Usually, the initial value is set to R=2, and then reduced as the number of iterations increases.

[0189] Hunting behavior;

[0190] At the same time, an oyster can hold several small fish at the same time, but the weaker small fish will be hunted by the oyster or predators. The probability of losing a small fish is:

[0191]

[0192] Let f(·) be the probability that a newly hatched bitter fish individual will be hunted, and let f(·) be the objective function solved by the bitter fish optimization algorithm.

[0193] Count the number K of newly hatched bitter fish individuals that survived after being hunted, and then use K of the remaining surviving newly hatched bitter fish individuals Replace the oyster location K negative individuals This results in the formation of a new population of bitter fish in the (t+1)th iteration.

[0194] S2066, Iteration:

[0195] Let the iteration number t = t + 1, and then return to step S2062; in this way, the bitter fish population enters the next iteration.

[0196] S2067, Convergence:

[0197] Repeat steps S2062 to S2066 until the preset maximum number of iterations t = t is reached. max The optimal solution for the bitter fish individual in the final iteration is obtained, which is the bitter fish individual closest to the oyster position after the last iteration, and is used as the optimized weight value.

[0198] This invention employs the bitter fish optimization algorithm to optimize the first and second weight values, thereby finding the optimal weight values. This makes the fixed load distribution prediction results and electric vehicle charging load prediction results obtained by weighted summation more accurate, which in turn facilitates the accurate optimization of the distribution network grid optimization map based on the fixed load distribution prediction results and electric vehicle charging load prediction results in subsequent processes.

[0199] S3. Based on the slope information, the distribution network grid map, and the obtained fixed load distribution prediction results and electric vehicle charging load prediction results, the preset distribution optimization model is used to optimize the distribution network grid map with the total cost minimization as the objective function.

[0200] In the specific implementation of this invention, a grid area in the distribution network grid map is regarded as a charging station planning area (regardless of the actual number of charging stations in the area, they are all considered as the same grid area for unified charging station planning). The goal is to minimize the total cost, and optimization is performed using a pre-defined distribution optimization model. Assuming the distribution network grid map contains *n* grids, the objective function expression of the distribution optimization model is:

[0201]

[0202] Among them, C totle C represents the total cost;grid,i C represents the construction cost of the distribution network for the i-th grid; install,i C represents the installation cost of charging stations in the i-th grid. trans,i is the power transmission loss cost of the i-th grid; M is the number of grids contained in the distribution network grid map.

[0203] Cost function definition:

[0204] The cost of each grid cell is defined as follows:

[0205] The construction cost C of the distribution network in the i-th grid grid,i and power transmission loss cost C trans,i Calculate using the following formulas respectively:

[0206] C grid,i =α1·L grid,i +α2·θ i ;

[0207] Among them, L grid,i θ represents the length of the power distribution line in the i-th grid. i Let α1 and α2 be the average slope angle of the i-th grid, and α1 and α2 be coefficients used to represent the weight of the distribution network construction and the slope angle.

[0208] C trans,i =γ·P load,i ·L grid,i ;

[0209] Among them, P load,i L represents the load of the i-th grid. grid,i γ represents the length of the power distribution line in the i-th grid, and γ represents the proportional coefficient of power transmission loss.

[0210] The installation cost C of the charging pile in the i-th grid install,i The cost can be calculated based on the actual installation and construction cost of the charging pile.

[0211] Constraints of the power distribution optimization model:

[0212] To ensure the rationality of the mesh generation, the following constraints need to be considered:

[0213] 1) Load balance constraints

[0214]

[0215] Among them, P load,i P represents the load of the i-th grid. totle This indicates the total load demand.

[0216] 2) Slope angle constraint

[0217] θ i≥θ max ;

[0218] Where, θ i Let θ be the average slope angle of the i-th grid. max This represents the maximum permissible slope angle for the planned area.

[0219] 3) Traffic flow constraints

[0220] T grid,i ≥T min ;

[0221] Among them, T grid,i T represents the minimum traffic flow in the i-th grid. min This indicates the minimum traffic flow set within the planned area.

[0222] Through the optimization of the above power distribution optimization model, an optimized power distribution grid map can be obtained. This map will then be used to classify the load level and traffic heat of each grid area, determine the capacity layout of electric vehicle charging piles in each grid area, and ultimately obtain the final planning scheme.

[0223] The grid-based planning approach based on the distribution network grid map not only fully considers the grid layout characteristics of charging stations in urban areas, but also simplifies the distribution information of charging stations in urban areas through gridding. This makes it easier to use the distribution optimization model and the grid area division and overlay operation for effective calculation, thereby improving the effectiveness of the grid layout of charging stations in urban areas and the efficiency of charging station planning.

[0224] S4. Based on the fixed load distribution prediction results and the electric vehicle charging load prediction results, the grid areas in the power distribution network grid optimization map are divided into load levels to obtain the grid area load division results.

[0225] In practice, step S4 is as follows:

[0226] S401. Based on the fixed load distribution prediction results and the electric vehicle charging load prediction results, the grid areas in the distribution network grid optimization map are divided into load levels; the load levels include light load areas, medium load areas, even load areas, heavy load areas, and overload areas, and the load level division method is as follows:

[0227] The grid area is divided into light load areas, with the percentage of fixed load ranging from [0% to 30%] and the percentage of electric vehicle charging load ranging from [0% to 10%]. This means that the load in the grid area is much lower than the planned capacity, and it is mostly an area with low electricity demand (such as suburbs and industrial areas during off-peak hours).

[0228] The grid area is divided into medium load zones, with the percentage of fixed load ranging from 30% to 50% and the percentage of electric vehicle charging load ranging from 10% to 20%. This indicates that the load level in the grid area is moderate and the grid load is within the normal working range, such as ordinary residential areas or light industrial areas.

[0229] The grid area is divided into a uniform load zone, with the percentage of fixed load ranging from (50% to 70%) and the percentage of electric vehicle charging load ranging from (20% to 30%). This indicates that the load in the grid area is close to a balanced operating state, which is within the range of good utilization of the planned capacity and the system has high safety.

[0230] In grid areas designated as heavy-load zones, the percentage of fixed load ranges from 70% to 100%, and the percentage of electric vehicle charging load ranges from 30% to 40%. This indicates that the load within the grid area is close to the upper limit of the planned capacity, which may be a high-density residential area or a concentrated commercial area.

[0231] In grid areas designated as overload zones, the percentage of fixed load is >100%, and the percentage of electric vehicle charging load is >40%. This indicates that the load in the grid area exceeds the planned capacity, usually due to short-term peak loads or excessive concentrated charging of electric vehicles.

[0232] S402. According to the load classification standard, fill the grid of the grid division result with grid colors. The chromaticity of the overload area is: R:90%, G:10%, B:10% (dark red); the chromaticity of the heavy load area is: R:100%, G:0%, B:0% (red); the chromaticity of the evenly loaded area is: R:80%, G:20%, B:20% (medium red); the chromaticity of the medium load area is: R:70%, G:30%, B:30% (light red); and the chromaticity of the light load area is: R:100%, G:30%, B:0% (orange).

[0233] S403. Based on the above steps, the grid area load division results are obtained.

[0234] S5. Based on the historical traffic flow data, the grid areas in the power distribution network grid optimization map are classified into traffic heat levels to obtain the grid area traffic heat classification results.

[0235] In specific implementation, step S5 is as follows:

[0236] S501. Based on the historical traffic flow data, classify the grid areas in the power distribution network grid optimization map into traffic heat maps; the traffic heat maps include completely unobstructed areas, basically unobstructed areas, normally operating areas, slow-moving areas, congested areas, and heavily congested areas, and the method for classifying traffic heat maps is as follows:

[0237] The grid area is divided into a completely unobstructed area, and the traffic heat value ranges from [0,50] vehicles / hour; this means that the traffic flow in the grid area is extremely low, there is almost no traffic congestion, and vehicles can move freely.

[0238] The grid area is divided into a basically smooth traffic zone, and the traffic heat value ranges from (50, 200) vehicles / hour; this means that the traffic flow in the grid area is low, the vehicles are moving smoothly, and there are occasional slight fluctuations in vehicle speed.

[0239] The grid area is designated as the normal operating zone, and the traffic heat value ranges from (200, 500) vehicles / hour. This indicates that the traffic flow in the grid area is at a moderate level, the traffic is basically smooth, and there may be a slight decrease in vehicle speed on some road sections.

[0240] The grid area is designated as a slow-moving zone, with a traffic heat value ranging from (500, 1000) vehicles / hour. This indicates that the traffic flow within the grid area is high, and traffic on some road sections is starting to slow down, with small-scale queuing occurring.

[0241] The grid area is divided into congested areas, and the traffic heat value ranges from (1000, 2000) vehicles / hour. This means that the traffic flow in the grid area is close to road saturation, vehicles are moving slowly, and most road sections are under heavy traffic pressure.

[0242] The grid area designated as a severely congested area has a traffic heat value range of >2000 vehicles / hour; this indicates that the traffic flow in the grid area seriously exceeds the road's carrying capacity, vehicles are almost at a standstill, and traffic is paralyzed.

[0243] S502. According to the traffic heat map classification standard, fill the grid division results with grid colors. The color of the heavily congested area is: R: 43%, G: 0%, B: 100% (blue); the color of the congested area is: R: 29%, G: 0%, B: 51% (blue); the color of the slow-moving area is: R: 0%, G: 25%, B: 50% (glass); the color of the normal-moving area is: R: 0%, G: 50%, B: 100% (light blue); the color of the basically smooth area is: R: 0%, G: 100%, B: 70% (cyan); the color of the completely smooth area is: R: 0%, G: 40%, B: 60% (empty).

[0244] S503. Based on the above steps, the traffic heat map results of the grid area are obtained.

[0245] S6. Based on the grid area load division results and grid area traffic heat division results, obtain the electric vehicle charging pile capacity planning scheme for each different grid area in the area to be planned.

[0246] In the specific implementation of the present invention, since step S4 has already yielded the grid area load division result, and step S5 has already yielded the grid area traffic heat division result, and different grid areas are distinguished by color according to their load level and traffic heat division, with darker colors used for heavier loads and higher traffic heat divisions. In this step, the colors for load level division and traffic heat division can be further combined to visually reflect the electric vehicle charging pile capacity layout in each grid area, thus facilitating the final planning scheme. Therefore, in step S6, the electric vehicle charging pile capacity planning for different grid areas in the area to be planned is implemented in the following manner:

[0247] S601. Based on the grid area load division results and the grid area traffic heat division results, and taking the division results of each grid area in the area to be planned as the basis, the chromaticity of the grid area load division results and the grid area traffic heat division results corresponding to each grid area is superimposed.

[0248] S602. Based on the chromaticity overlay results of each grid area, plan the location and capacity of charging piles according to the following standards:

[0249]

[0250] Therefore, the power-traffic heat maps of different grid areas obtained by color overlay comprehensively consider the combined effects of the grid area load division results and the grid area traffic heat map results. The resulting electric vehicle charging pile capacity planning schemes for different grid areas have the following technical advantages:

[0251] 1. Dynamically match power load with traffic demand. Take power load and traffic flow into account to ensure that charging stations are rationally distributed within the power system's capacity. Reduce or limit charging station planning in overloaded or heavily loaded areas, and increase charging station deployment in areas with even or light loads to avoid further exacerbating grid pressure. Prevent grid overload due to improper charging station planning, while simultaneously meeting vehicle charging needs in areas with high traffic volume.

[0252] 2. Clear regional classification and targeted planning. Through a well-defined zoning strategy (light load, medium load, even load, heavy load, overload) and corresponding traffic heat levels, the capacity and functional distribution of charging stations are more precise. For example:

[0253] Lightly loaded areas: With low power pressure and moderate traffic demand, 100% charging capacity can be planned, and these areas can even be designated as key areas for promoting electric vehicle charging.

[0254] Medium-load areas: Traffic volume is moderate, power load is stable, and charging piles are deployed at 80% of the planned capacity to ensure a balance between supply and demand.

[0255] Evenly loaded areas: The load is close to the design capacity, suitable for 60% to 80% of the planned deployment of charging piles, while taking into account the stability of the power grid.

[0256] Heavy load and overload areas: The power grid is under great pressure. Load transfer is guided by limiting the construction of charging piles or by using only discharge mode.

[0257] 3. Make full use of resources in areas with even load and light load. Plan a higher proportion of charging piles (up to 100%) in areas with even load and light load to avoid idle power resources, and at the same time alleviate the power pressure in other areas by guiding traffic flow to these areas. Improve the utilization rate of the power grid, optimize the overall layout of charging piles, and alleviate the charging pressure in heavily loaded or overloaded areas.

[0258] 4. Balancing traffic flow and service convenience. In areas with high traffic flow (such as heavily congested areas), even with a high grid load, service convenience can be improved by appropriately planning discharge modes or using a limited number of charging stations. This ensures the efficiency of urban traffic operations and prevents further traffic deterioration due to a shortage of charging stations.

[0259] 5. Alleviate urban congestion and improve resource allocation efficiency. In heavily congested areas (high traffic heat values), reduce the number of charging stations or only provide discharge functions to reduce vehicle dwell time. Simultaneously, guide vehicles to charge in areas with even or light loads. Optimize urban traffic flow distribution and reduce the dual pressure of congestion and electricity load.

[0260] S7. Based on the obtained electric vehicle charging pile capacity planning scheme, plan the capacity of electric vehicle charging piles distributed in different grid areas within the planning area.

[0261] In the specific implementation of the present invention, based on the obtained electric vehicle charging pile capacity planning scheme, a corresponding electric vehicle charging pile capacity planning strategy is implemented for the charging stations in each grid area to achieve reasonable planning of the charging pile capacity in all charging stations in the planning area.

[0262] This invention presents an electric vehicle charging station planning method based on load and traffic flow distribution. The planning process considers the combined effects of terrain data, historical load data, and historical traffic flow data. Historical load data is used to predict the fixed load distribution of charging stations and the charging load of electric vehicles. Combining the prediction results with terrain data analysis, a distribution optimization model is employed to optimize the distribution network grid map and then divide the grid areas into load zones. This more accurately describes the impact of terrain data and historical load data on the gridded planning layout of charging stations. Furthermore, the optimized distribution network grid map and historical traffic flow data are used to analyze the traffic heat distribution within the grid areas. Finally, considering the combined effects of the grid area load division and grid area traffic heat distribution results, a capacity planning scheme for electric vehicle charging piles is determined to better ensure the accuracy and rationality of electric vehicle charging station planning.

[0263] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include ROM, RAM, disk, or optical disk, etc.

[0264] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit the technical solutions. Those skilled in the art should understand that any modifications or equivalent substitutions to the technical solutions of the present invention without departing from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A method for planning electric vehicle charging stations based on load and traffic flow distribution, characterized in that, Includes the following steps: S1. Obtain terrain data, historical load data, and historical traffic flow data for the area to be planned; the terrain data includes slope information and a planar power grid map; the historical load data includes fixed load demand data and electric vehicle charging load demand data. S2. Based on the historical load data, predict the fixed load and electric vehicle charging load to obtain the fixed load distribution prediction results and the electric vehicle charging load prediction results; this step specifically involves: S201. Decompose the fixed load demand data and electric vehicle charging load demand data in the historical load data into internal model functions to obtain the fixed load internal model function components and the electric vehicle charging load internal model function components respectively. S202. Using the ensemble empirical mode decomposition method, the internal mode function components of each fixed load and the internal mode function components of the electric vehicle charging load are decomposed to obtain the decomposed fixed load time series and electric vehicle charging load time series. S203. Based on the decomposed fixed load time series and electric vehicle charging load time series, determine the information entropy (IE) of each fixed load internal model function component and electric vehicle charging load internal model function component, and reconstruct the fixed load extended internal model function component and electric vehicle charging load extended internal model function component with similar information entropy (IE); the similar information entropy (IE) means that the difference between the two information entropy (IE) is within a preset difference range. S204. The information entropy IE of the fixed load internal model function component and the fixed load extended internal model function component is used as the input of the preset LSTM model for prediction to obtain the first fixed load distribution prediction result; the information entropy IE of the electric vehicle charging load internal model function component and the electric vehicle charging load extended internal model function component is used as the input of the LSTM model for prediction to obtain the first electric vehicle charging load prediction result. S205. The information entropy IE of the fixed load internal model function component and the fixed load extended internal model function component is used as the input of the preset GMDH model for prediction to obtain the second fixed load distribution prediction result; the information entropy IE of the electric vehicle charging load internal model function component and the electric vehicle charging load extended internal model function component is used as the input of the GMDH model for prediction to obtain the second electric vehicle charging load prediction result. S206. The first fixed load distribution prediction result and the second fixed load distribution prediction result are weighted and summed to obtain the final fixed load distribution prediction result; the first electric vehicle charging load prediction result and the second electric vehicle charging load prediction result are weighted and summed to obtain the final electric vehicle charging load prediction result. S3. Based on the slope information, the distribution network grid map, and the obtained fixed load distribution prediction results and electric vehicle charging load prediction results, a preset distribution optimization model is used to optimize the distribution network grid map with the objective function of minimizing total cost; wherein, the objective function expression of the distribution optimization model is: ； in, Total cost; For the first The construction cost of the distribution network for each grid; For the first The installation cost of charging piles per grid; For the first The cost of power transmission loss per grid; This represents the number of grids contained in the distribution network grid map. No. Construction cost of distribution network per grid and power transmission loss cost Calculate using the following formulas respectively: ； in, Indicates the first The length of the power distribution lines in each grid, For the first The average slope angle of each grid, and , where is a coefficient used to represent the weight of the distribution network construction and the slope angle; ； in, Indicates the first The load of each grid, Indicates the first The length of the power distribution lines in each grid, A proportionality coefficient representing power transmission loss; The constraints of the power distribution optimization model include: 1) Load balance constraints: ； in, Indicates the first The load of each grid, Indicates total load demand; 2) Slope angle constraint: ； in, For the first The average slope angle of each grid, The maximum allowable slope angle for the planned area; 3) Traffic flow constraints: ； in, Indicates the first The minimum traffic flow within each grid, This indicates the minimum traffic flow rate set within the planned area; S4. Based on the fixed load distribution prediction results and the electric vehicle charging load prediction results, the grid areas in the power distribution network grid optimization map are divided into load levels to obtain the grid area load division results. S5. Based on the historical traffic flow data, the grid areas in the power distribution network grid optimization map are classified into traffic heat levels to obtain the grid area traffic heat classification results. S6. Based on the grid area load division results and grid area traffic heat division results, obtain the electric vehicle charging pile capacity planning scheme for each different grid area in the area to be planned. S7. Based on the obtained electric vehicle charging pile capacity planning scheme, plan the capacity of electric vehicle charging piles distributed in different grid areas within the planning area.

2. The electric vehicle charging station planning method based on load and traffic flow distribution according to claim 1, characterized in that, Step S206 is as follows: First, the first weight value of the weighted sum of the first fixed load distribution prediction result and the second fixed load distribution prediction result, and the second weight value of the weighted sum of the first electric vehicle charging load prediction result and the second electric vehicle charging load prediction result are optimized by the bitter fish optimization algorithm. Then, the optimized first weight value and second weight value are used to calculate the fixed load distribution prediction result and the electric vehicle charging load prediction result by weighted summation.

3. The electric vehicle charging station planning method based on load and traffic flow distribution according to claim 2, characterized in that, In step S206, the specific method for optimizing any one of the first weight value and the second weight value using the bitter fish optimization algorithm is as follows: S2061, Randomly generated Each initial weight is used as an individual weight for the bitter fish population. The initial population of bitter fish was obtained. ,Right now: ， ； in, For any number of bitter fish in the population The number of individual bitter fish and the number of individuals in the bitter fish population are the weight values ​​to be optimized. For any number of bitter fish in the population A single bitter fish In the The value of the dimension; Let represent the number of individual bitter fish in the bitter fish population in the bitter fish optimization algorithm. The dimension of each individual bitter fish in the bitter fish population; Then, initialize the number of iterations. =1, proceed to step S2062; S2062. Randomly select half of the bitterling individuals in the bitterling population as positive individuals and mark them as such. The other half of the bitter fish individuals were negative individuals and marked as... ; S2063, For each positive individual The iterative updates are performed in the following manner: ； In the formula, and They are positive individuals In the Second and third Position in the next iteration; It is a randomly selected oyster location; It is in the The optimal solution for the individual bitter fish in the nth iteration, i.e. the th The bitter fish individual closest to the oyster in the next iteration; and It is a random number between 0 and 1; It is the number of steps or speed at which an individual bitterfish moves in order to escape from or approach an oyster, and: ； These are the initial step size and jump value for each bitter fish; This is the preset maximum number of iterations; It is a random function, and: ； U(1) equals 1; probability parameter for: ； in, It is the number of iterations; Indicates the power of restoring; It is a random number; S2064, For each negative individual The iterative updates are performed in the following manner: ； in, The range of values ​​for individual bitter fish in the initial bitter fish population. , These represent the lower and upper limits of the values ​​for an individual bitterfish; Indicates the average position of the fish school: ； S2065, Each Positive Individual The newly created bitter fish is represented as: ； Indicates a positive individual The newly created bitter fish; This represents the distribution radius of individual bitterfish around the oyster shell, with values ​​between [0, 2]. The probability of a newly hatched bitterfish being hunted is: ； The probability of a newly hatched bitter fish being hunted. It is the objective function solved by the bitter fish optimization algorithm; Count the number of newborn bitter fish that survived after being hunted. Then use The remaining surviving newborn bitter fish individuals Replace the oyster location of A negative individual This forms a new first The next iteration of the bitter fish population; S2066, Let the number of iterations be... Then return to step S2062; S2067. Repeat steps S2062 to S2066 until the preset maximum number of iterations is reached. The optimal solution for the bitter fish individual in the final iteration is obtained, which is the bitter fish individual closest to the oyster position after the last iteration, and is used as the optimized weight value.

4. The electric vehicle charging station planning method based on load and traffic flow distribution according to claim 1, characterized in that, Step S4 is as follows: S401, Based on the fixed load distribution prediction results and the electric vehicle charging load prediction results, the grid areas in the distribution network grid optimization map are divided into load levels; the load levels include light load areas, medium load areas, even load areas, heavy load areas, and overload areas, and the load level division method is as follows: The grid area is divided into light load areas, with the percentage of fixed load ranging from [0%, 30%] and the percentage of electric vehicle charging load ranging from [0%, 10%]. The grid area is divided into medium load areas, with the percentage of fixed load ranging from (30%, 50%) and the percentage of electric vehicle charging load ranging from (10%, 20%). The grid area is divided into uniform load areas, with the percentage of fixed load ranging from (50%, 70%) and the percentage of electric vehicle charging load ranging from (20%, 30%). The grid area is divided into heavy load areas, with the percentage of fixed load ranging from (70%, 100%) and the percentage of electric vehicle charging load ranging from (30%, 40%). In the grid areas designated as overload zones, the percentage of fixed load is >100%, and the percentage of electric vehicle charging load is >40%. S402, according to the load classification standard, fill the grid of the grid division result with grid colors. The chromaticity for filling the overload area is: R:90%, G:10%, B:10%; the chromaticity for filling the heavy load area is: R:100%, G:0%, B:0%; the chromaticity for filling the evenly loaded area is: R:80%, G:20%, B:20%; the chromaticity for filling the medium load area is: R:70%, G:30%, B:30%; and the chromaticity for filling the light load area is: R:100%, G:30%, B:0%. S403, Based on the above steps, the grid area load division results are obtained.

5. The electric vehicle charging station planning method based on load and traffic flow distribution according to claim 1, characterized in that, Step S5 is as follows: S501. Based on the historical traffic flow data, classify the grid areas in the power distribution network grid optimization map into traffic heat maps; the traffic heat maps include completely unobstructed areas, basically unobstructed areas, normally operating areas, slow-moving areas, congested areas, and heavily congested areas, and the method for classifying traffic heat maps is as follows: The grid area is divided into a completely unobstructed area, and the traffic heat value ranges from [0,50] vehicles / hour. The grid area is divided into a basically unobstructed area, and the traffic heat value ranges from (50, 200) vehicles / hour. The grid area is divided into normal operating zones, and the traffic heat value ranges from (200, 500) vehicles / hour. The grid area is divided into slow-moving zones, and the traffic heat value ranges from (500, 1000) vehicles / hour. The grid area is divided into congested areas, and the traffic heat value ranges from (1000, 2000) vehicles / hour. For grid areas classified as heavily congested areas, the traffic heat value range is >2000 vehicles / hour; S502. According to the traffic heat map classification standard, fill the grid division results with grid colors. The color values ​​for filling heavily congested areas are: R: 43%, G: 0%, B: 100%; for congested areas, R: 29%, G: 0%, B: 51%; for slow-moving areas, R: 0%, G: 25%, B: 50%; for normally moving areas, R: 0%, G: 50%, B: 100%; for basically smooth areas, R: 0%, G: 100%, B: 70%; and for completely smooth areas, R: 0%, G: 40%, B: 60%. S503. Based on the above steps, the traffic heat map results of the grid area are obtained.

6. The electric vehicle charging station planning method based on load and traffic flow distribution according to claim 1, characterized in that, Step S6 is as follows: S601. Based on the grid area load division results and the grid area traffic heat division results, and taking the division results of each grid area in the area to be planned as the basis, the chromaticity of the grid area load division results and the grid area traffic heat division results corresponding to each grid area is superimposed. S602. Based on the chromaticity overlay results of each grid area, plan the location and capacity of charging piles according to the following standards: 。

Images