CN117974221A - Electric vehicle charging station location selection method and system based on artificial intelligence - Google Patents
Electric vehicle charging station location selection method and system based on artificial intelligence Download PDFInfo
- Publication number
- CN117974221A CN117974221A CN202410381566.7A CN202410381566A CN117974221A CN 117974221 A CN117974221 A CN 117974221A CN 202410381566 A CN202410381566 A CN 202410381566A CN 117974221 A CN117974221 A CN 117974221A
- Authority
- CN
- China
- Prior art keywords
- data
- electric vehicle
- function
- vehicle charging
- individual
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 22
- 238000010187 selection method Methods 0.000 title claims abstract description 7
- 238000000034 method Methods 0.000 claims abstract description 62
- 238000005457 optimization Methods 0.000 claims abstract description 36
- 238000013528 artificial neural network Methods 0.000 claims abstract description 35
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 35
- 230000009467 reduction Effects 0.000 claims abstract description 28
- 238000012549 training Methods 0.000 claims abstract description 27
- 238000000605 extraction Methods 0.000 claims abstract description 9
- 230000006870 function Effects 0.000 claims description 85
- 230000008569 process Effects 0.000 claims description 30
- 230000007613 environmental effect Effects 0.000 claims description 24
- 230000003993 interaction Effects 0.000 claims description 23
- 238000004364 calculation method Methods 0.000 claims description 19
- 230000002068 genetic effect Effects 0.000 claims description 16
- 238000011156 evaluation Methods 0.000 claims description 14
- 210000002569 neuron Anatomy 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000007246 mechanism Effects 0.000 claims description 10
- 230000004913 activation Effects 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 7
- 238000007635 classification algorithm Methods 0.000 claims description 7
- 230000007614 genetic variation Effects 0.000 claims description 7
- 230000036961 partial effect Effects 0.000 claims description 7
- 239000002131 composite material Substances 0.000 claims description 6
- 238000001303 quality assessment method Methods 0.000 claims description 6
- 230000002860 competitive effect Effects 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 4
- 230000006978 adaptation Effects 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 238000002790 cross-validation Methods 0.000 claims description 3
- 230000008846 dynamic interplay Effects 0.000 claims description 3
- 230000001172 regenerating effect Effects 0.000 claims description 3
- 238000013441 quality evaluation Methods 0.000 claims description 2
- 238000012795 verification Methods 0.000 claims description 2
- 238000011946 reduction process Methods 0.000 claims 1
- 238000012545 processing Methods 0.000 description 10
- 230000000694 effects Effects 0.000 description 6
- 230000008859 change Effects 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 238000011161 development Methods 0.000 description 4
- 230000018109 developmental process Effects 0.000 description 4
- 238000002372 labelling Methods 0.000 description 4
- 230000035772 mutation Effects 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 230000006399 behavior Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000002028 premature Effects 0.000 description 2
- 230000001737 promoting effect Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000035899 viability Effects 0.000 description 2
- 241000283153 Cetacea Species 0.000 description 1
- 241000669618 Nothes Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000006353 environmental stress Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 230000031068 symbiosis, encompassing mutualism through parasitism Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an electric vehicle charging station location selection method and system based on artificial intelligence, wherein the method collects public transportation data, social and economic data, electric vehicle use data and geographic information system data, performs data annotation and data expansion, and obtains an electric vehicle charging pile location selection factor data set; training a neural network by adopting an ecological system optimization algorithm, and extracting features of an electric vehicle charging pile site selection factor dataset by adopting the trained neural network; performing dimension reduction treatment on the extracted features; and inputting the feature subjected to dimension reduction into a classifier to obtain an electric vehicle charging station site selection classification result. The invention improves generalization capability, improves the precision and efficiency of feature extraction, and effectively supports the complex site selection decision requirement.
Description
Technical Field
The invention belongs to the technical field of charging stations, and particularly relates to an electric vehicle charging station site selection method and system based on artificial intelligence.
Background
Under the current challenges of energy transformation and climate change, electric Vehicles (EV) are becoming important for popularization in the world as a clean energy vehicle. With the rapid increase in the number of electric vehicles, building a sufficient charging infrastructure becomes one of the key factors supporting the rapid development of electric vehicles. However, the location process of electric vehicle charging stations presents a number of challenges, including how to integrate various factors such as geographic location, traffic flow, user demand, grid capacity, and how to balance cost and efficiency. These challenges require the development of more efficient, intelligent site selection methods to support the rapid expansion and optimization of the layout of the electric vehicle charging network.
The traditional electric vehicle charging station location method depends on expert experience and simple qualitative analysis, and a large amount of Geographic Information System (GIS) data, traffic data and user behavior data are difficult to fully utilize. This approach is inefficient in processing large-scale complex data and is difficult to accommodate with rapidly changing market demands and technological developments. With the development of information technology and artificial intelligence technology, a data-driven decision support system becomes possible, and a new solution is provided for electric vehicle charging station site selection.
The Chinese patent application No. CN202310276165.0 proposes a method and a system for selecting the address of an electric vehicle charging station based on game. The Chinese patent application No. CN2020110167740. X proposes an electric vehicle charging station site selection modeling method based on an improved whale algorithm. The Chinese patent application No. CN202311114831.7 proposes an electric vehicle public charging station location and volume-fixing method based on charging behavior characteristics. These methods each have advantages.
However, the prior art has the following disadvantages:
1. The traditional method may not effectively solve the problem of data unbalance, so that sample diversity is insufficient during model training, and generalization capability of a model and accuracy of an address selection result are affected.
2. The parameter optimization and feature extraction methods used in the prior art may not be accurate enough, and the model cannot be effectively prevented from being converged to a local optimal solution too early, or the feature extraction is not accurate enough, so that the performance of the site selection model is affected.
3. Traditional addressing methods may not efficiently support complex decision making requirements, and are inefficient in data processing, particularly when processing large-scale data sets, and may not provide fast and accurate decision support.
4. The classification algorithm in the prior art may not solve the problems of data noise and outlier well, so that the generalization capability and robustness of the classifier are insufficient, and the quality of the site selection decision is affected.
Disclosure of Invention
In order to solve the defects, the invention provides an electric vehicle charging station location method and system based on artificial intelligence.
The invention is realized by the following technical scheme. An electric vehicle charging station location method based on artificial intelligence comprises the following steps:
S1, collecting public transportation data, social and economic data, electric vehicle use data and geographic information system data, and performing data annotation and data expansion to obtain an electric vehicle charging pile site selection factor data set;
s2, training a neural network by adopting an ecological system optimization algorithm, and extracting features of an electric vehicle charging pile site selection factor dataset by adopting the trained neural network;
s3, performing dimension reduction treatment on the extracted features;
s4, inputting the feature subjected to dimension reduction into a classifier, and training the classifier;
S5, collecting site selection factor data of the electric vehicle charging piles to be classified, extracting features and performing dimension reduction, and inputting the obtained dimension reduced features into a trained classifier for classification to obtain site selection classification results of the electric vehicle charging stations;
The training process of the neural network by adopting the ecological system optimization algorithm comprises the following steps:
s21, generating an initial ecological system, and initializing the ecological system;
S22, calculating the fitness of each individual; the fitness updating function of the individual is:
;
Wherein, Is the firstThe initial fitness of the individual person(s),For individualsAndThe strength of the interaction between the two,To take into account fitness of the ith individual after ecological interaction; Is a weight adjustment coefficient; To adapt the competition strength between the ith individual and the jth individual, The symbiotic intensity of the ith individual and the jth individual after self-adaptive adjustment;
And Is based on a performance feedback functionAdjusting; performance feedback functionEvaluating the overall state of the current ecological network based on the performance of the neural network on the verification set;
s23, simulating a natural selection process, selecting individuals with high fitness for reproduction, and eliminating individuals with low fitness at the same time;
s24, regenerating and reproducing, simulating the genetic variation process of organisms, and generating offspring by crossing and variation operations in a genetic algorithm for individuals with high adaptability;
S25, simulating ecological interaction among individuals, and updating the fitness of the individuals;
S26, simulating the influence of environmental changes on an ecological system, and adjusting the fitness;
And S27, updating parameters of the neural network according to the fitness and ecological dynamics of the final individual in the ecological system.
Further preferably, the data expansion is performed by adopting an improved quadratic programming SMOTE algorithm, and the steps are as follows:
S11, carrying out sample statistics of each category on an original data set, determining a few types of samples, evaluating the unbalance degree of the original data set, and initializing the iteration times and the number of synthesized samples in each iteration according to a statistical result;
S12, in the first layer iteration, for each minority sample, finding out the nearest neighbor sample set in the minority sample sets, and generating a synthesized sample according to each nearest neighbor sample and the random number;
s13, based on the distance between the synthesized sample and the nearest original sample, evaluating the quality of the synthesized sample through a quality evaluation function, and adjusting the generation strategy of the synthesized sample according to the evaluation result;
s14, in subsequent iterations, continuing to refine the generation process of the synthetic sample, and gradually optimizing the representativeness and diversity of the synthetic sample by continuously evaluating the quality of the synthetic sample and adjusting parameters;
and S15, the iteration process is continuously carried out until the preset iteration times are reached, and a final expansion data set is output.
Further preferably, in step S12, a synthetic sample is generatedThe formula of (2) is as follows: ; wherein, Representing the i-th minority class of samples,Representing the sample of the nearest neighbor of the j,Is a random number, and the code is a random number,For controlling atAndThe degree of interpolation between.
Further preferably, the specific procedure of step S13 is as follows: defining a quality assessment function of the composite sample asAn adaptive adjustment mechanism is introduced to adjust random numbersValues of (2) such thatMaximization, updating random numbersThe strategy of (2) is as follows:
;
Wherein, In order to update the random number after the update,Is the learning rate of the random number update,Is thatWith respect to random numbersIs a gradient of (2);
Quality assessment function Expressed as:
;
Wherein, Is thatNearest neighbor samples in the original minority sample set,The euclidean distance is represented as,Is a parameter controlling the shape of the function;
Calculated according to the chain law, expressed as:
;
Wherein, Is thatWith respect toIs used for the partial derivative of (a),Is thatWith respect toIs a partial derivative of (c).
Further preferably, the process of generating an initial ecosystem and initializing the ecosystem includes:
Constructing an ecosystem comprising N individuals, the first Initial fitness of individualDepending on its performance on a particular task,; Wherein,Representing a variable of the environment,Represents a genetic factor of the human body,Is the firstA set of parameters for the individual person,WhereinAndRespectively representing the weight and the bias of the neural network;
The initial fitness calculation function is expressed as follows:
;
Wherein, As a function of the genetic factors,As a function of the performance evaluation,Is an environmental interaction function that describes the dynamic interaction between individual parameters and environmental variables. Sig () is a Sigmoid activation function,The weight coefficients of performance evaluation, environment variables, genetic factors, individual parameters and environment interactions are respectively;
Genetic factor function The calculation mode is expressed as follows:
;
Wherein, AndThe first weight coefficient and the second weight coefficient are respectively; Representing the variance of the weights and biases, for measuring the diversity of the parameters, n represents the total number of neurons, Representing the weight of the jth neuron in the ith individual parameter,Representing the bias of the jth neuron in the ith individual parameter;
environmental interaction function Expressed as:
;
Wherein, Is the third weight coefficient of the weight,Represents the first of the environmental variablesThe number of elements to be added to the composition,Representing and environment variablesThe input characteristics of the data are associated with,Is a hyperbolic tangent function;
Performance evaluation function Expressed as:
;
Wherein, Representing a loss function and,Is a real label; is a predictive tag derived from a preset Softmax function.
Further preferably, in step S22, the competition strength is calculatedAnd symbiotic strengthThe manner of making the adaptation is expressed as:
;
;
Wherein, The competitive strength of the ith individual and the jth individual is given, S ij is the symbiotic strength of the ith individual and the jth individual,AndThe adaptive functions that adjust the competition and symbiotic strengths respectively,Representing a performance feedback functionA variation amount;
Performance feedback function The calculation mode of (a) is expressed as follows:
;
Wherein, The weight coefficients of accuracy, complexity, diversity and time, respectively;
the functions of each component are respectively as follows:
Accuracy function :;
Complexity function:;
Diversity function:;
Time function:;
Wherein P is the accuracy rate, L is the average loss value; c is network complexity and is in direct proportion to parameter quantity; h is a data diversity index proportional to the variance of the data features; t is training time and is in direct proportion to the iteration times;
;
;
Wherein, AndRepresents the ith individual and the jth individual pair, respectivelyThe need for seed resources is addressed,Is the total category number of the resource.
Further preferably, the following is a process of performing dimension reduction processing on the extracted features by using an improved self-encoder algorithm:
S31, initializing the structure and parameters of a self-encoder;
S32, data are transmitted forwards through the encoder part, and the output of each layer is the input of the next layer until the last encoding layer is reached;
s33, reconstructing the coded features through a decoder part;
S34, adopting a combined loss function of the reconstruction error and the regularization term, and updating model parameters according to the gradient of the loss function by using an Adam optimization algorithm;
s35, dynamically adjusting coding parameters according to model performance and multi-target requirements in the training process;
S36, iterating the steps S32 to S35 until the preset maximum iteration times are met.
Further preferably, a classifier is constructed by adopting a low-rank sparse-based extreme learning machine classification algorithm, and the flow of the low-rank sparse-based extreme learning machine classification algorithm is as follows:
S41, initializing input weight and bias of an extreme learning machine;
S42, data are transmitted forwards through a hidden layer of the extreme learning machine;
S43, calculating output weight by solving an optimization problem by using a low-rank sparse constraint and a robust optimization strategy; output weight The solution of (2) translates into the following optimization problem:
;
Wherein, Is the output matrix of the object and,AndIs a regularization parameter controlling sparsity and low rank,Is the Frobenius norm,Is the L1 norm of the sample,Representing the rank of the matrix;
S44, evaluating the performance of the classifier in a cross-validation mode, and adjusting the parameters of the classifier.
The invention provides an electric vehicle charging station site selection system based on artificial intelligence, comprising:
The data acquisition device is used for acquiring public transportation data, social and economic data, electric vehicle use data and geographic information system data, marking the data and expanding the data to obtain an electric vehicle charging pile site selection factor data set;
The feature extraction module is used for training the neural network by adopting an ecological system optimization algorithm, and extracting features of the electric vehicle charging pile site selection factor dataset by adopting the trained neural network;
The dimension reduction module is used for carrying out dimension reduction treatment on the extracted features;
And the classifier is used for classifying according to the feature after dimension reduction to obtain the address selection classification result of the electric vehicle charging station.
The invention provides a non-volatile computer storage medium, which stores computer executable instructions capable of executing an electric vehicle charging station location selection method based on artificial intelligence.
The present invention provides an electronic device including: the system comprises at least one processor and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform an artificial intelligence based electric vehicle charging station locating method.
The invention has the following advantages:
1. Aiming at the particularity of the problem of electric vehicle charging station site selection, the invention improves the traditional SMOTE algorithm, adopts an optimization method based on quadratic programming and a hierarchical iteration mechanism, effectively solves the problems of unbalanced data and insufficient sample diversity, and improves the generalization capability of the method.
2. The neural network parameters are optimized by simulating natural selection and genetic variation processes in the ecological system, and the problems of premature convergence and local optimization are effectively prevented by adopting a diversity maintaining mechanism based on the ecological niche theory, so that the performance of the neural network model is improved, and the precision and the efficiency of feature extraction are improved.
3. By adopting quantum mapping characteristics and a dynamic coding adjustment mechanism, the method optimizes the data characteristics of the electric vehicle charging station site selection problem, improves the efficiency and the precision of feature dimension reduction, introduces a multi-objective optimization strategy, ensures that the dimension-reduced features can effectively support decision requirements, effectively supports complex site selection decision requirements, and improves the efficiency of data processing and decision making.
4. By adopting a low-rank sparse constraint and robust optimization strategy, the stability of the model in the face of data noise and abnormal values is improved, the generalization capability and the interpretation of the classifier are improved, and the model is kept efficient and accurate in the process of large-scale data.
Drawings
FIG. 1 is a flow chart of the present invention.
Detailed Description
The invention is further elucidated in the following in connection with the accompanying drawings and examples.
Example 1
As shown in fig. 1, the electric vehicle charging station location method based on artificial intelligence comprises the following steps:
S1, collecting public transportation data, social and economic data, electric vehicle use data and geographic information system data, and performing data annotation and data expansion to obtain an electric vehicle charging pile site selection factor data set;
s2, training a neural network by adopting an ecological system optimization algorithm, and extracting features of an electric vehicle charging pile site selection factor dataset by adopting the trained neural network;
s3, performing dimension reduction treatment on the extracted features;
s4, inputting the feature subjected to dimension reduction into a classifier, and training the classifier;
s5, collecting address selection factor data of the electric vehicle charging piles to be classified, extracting features and performing dimension reduction processing, and inputting the obtained dimension reduced features into a trained classifier for classification to obtain address selection classification results of the electric vehicle charging stations. The electric vehicle charging station address selection classification result is 1 of recommended address selection, alternative recommended address selection and non-recommended address selection, and the charging station address selection decision is made according to the electric vehicle charging station address selection classification result.
In step S1 of the present embodiment, public transportation data includes a station position, route information, and an operation schedule of public transportation means such as buses and subways; the socioeconomic data comprise indexes such as population density, economic activity intensity, electric vehicle possession and the like; the electric vehicle use data comprise user behavior data such as electric vehicle charging frequency, charging duration, driving mileage and the like; the geographical information data includes geographical space data such as road network, building location, current situation of charging station, etc. The data labeling mode is manual labeling, and the labeling category comprises 3 categories of recommended site selection, alternative recommended site selection and non-recommended site selection.
The traditional neural network training adopts a gradient descent mode, so that the local optimum condition is easily trapped in the training process, and meanwhile, the gradient disappearance and gradient explosion phenomena are also easily caused. Inspired by the mechanism of species living and reproducing in the ecosystem through competition with each other, symbiosis and adaptation to environmental changes, step S2 of this embodiment proposes training the neural network with an ecosystem optimization algorithm in which each parameter (weight w and bias b) of the neural network is treated as an individual organism in the ecosystem, whose ability to survive and reproduce is determined by its fitness (i.e. the difference from the expected output), and optimizing these parameters by simulating natural selection and genetic variation processes in the ecosystem, thereby improving the performance of the neural network. In addition, the ecological system optimization algorithm introduces a diversity maintaining mechanism based on the ecological niche theory, aims at preventing premature convergence and local optimization, and explores a wider solution space by maintaining the diversity of parameter populations.
Example 2
It can be understood that in the task of the invention, acquisition, labeling and preprocessing of training data are time-consuming and labor-consuming, and insufficient training samples easily result in poor generalization capability of the model, and simultaneously influence the accuracy of the model. The embodiment proposes to perform data expansion based on an improved quadratic programming SMOTE (SYNTHETIC MINORITY OVER-sampling Technique) algorithm. The SMOTE algorithm generates new composite samples by interpolating between a few classes of samples to solve the data imbalance problem. But the SMOTE algorithm does not work well when processing highly unbalanced data sets or data sets with complex distribution characteristics, the invention improves the SMOTE algorithm, introduces a quadratic programming-based optimization method to more precisely control the generation process of the synthesized samples, and considers not only the linear relationship among the samples but also the nonlinear characteristics of the sample distribution. In addition, in order to further improve the efficiency and effect of data expansion, a hierarchical iteration mechanism is introduced, and the targets and the ranges of data expansion are gradually refined through a multi-level iteration process. In each iteration, the algorithm evaluates the quality of the currently generated synthetic sample and adjusts the parameter setting of the subsequent iteration according to the evaluation result, so that the algorithm can dynamically adjust according to the specific characteristics of the data set to generate more accurate and diversified synthetic samples.
Specifically, the data expansion based on the improved quadratic programming SMOTE algorithm comprises the following steps:
S11, carrying out sample statistics of each category on an original data set, determining a few types of samples, evaluating the unbalance degree of the original data set, and initializing the iteration times and the number of synthesized samples in each iteration according to a statistical result;
raw data set ,Representing the 1 st, 2 nd, … th, n data points of the original dataset, respectively, wherein each data point contains characteristics related to the charging requirement of the electric vehicle, in one embodiment, including geographic location, charging time, charging frequency, etc. Further, a minority class sample set is set asAnd calculating the unbalanced proportion of the original data setCan be expressed as:。
S12, in the first layer iteration, for the ith minority class sample Find it in a minority class sample setIn (a) and (b)Sets of nearest neighbor samplesFor the j nearest neighbor sampleGenerating a synthetic sampleThe following are provided: ; wherein, Is a random number, and the code is a random number,For controlling atAndThe degree of interpolation between; the generated composite samples are combined with the original data set to form an expanded data set. In the present embodiment, parametersIs set to 5. Step S12 focuses on increasing the sample size to alleviate the data imbalance problem.
S13, based on the distance between the synthesized sample and the nearest original sample, the quality of the synthesized sample is estimated through a quality estimation function, and the generation strategy of the synthesized sample is adjusted according to the estimation result. The generation strategy of the synthetic sample is to adjust the interpolation weight and select different sample pairs for interpolation.
Specifically, a quality assessment function of the synthesized sample is defined asAn adaptive adjustment mechanism is introduced to adjust random numbersValues of (2) such thatMaximization, updating random numbersThe strategy of (2) is as follows:
;
Wherein, In order to update the random number after the update,Is the learning rate of the random number update,Is thatWith respect to random numbersIs a gradient of (a). In one embodiment of the present invention, in one embodiment,Is set to 0.01.
Quality assessment functionFor evaluating synthetic samplesWhich is defined in terms of the distance between the synthesized sample and its nearest neighbor original sample, can be expressed as:
;
Wherein, Is thatNearest neighbor samples in the original minority sample set,The euclidean distance is represented as,Is a parameter that controls the shape of the function.
According to the chain law calculation, it can be expressed as:
;
Wherein, Is thatWith respect toIs used for the partial derivative of (a),Is thatWith respect toThe partial derivative of (2) is calculated as follows:
;
;
S14, in the subsequent iteration, continuing to refine the generation process of the synthetic sample, and gradually optimizing the representativeness and diversity of the synthetic sample by continuously evaluating the quality of the synthetic sample and adjusting parameters. In one embodiment, the method comprises the steps of introducing a polynomial function Instead of linear interpolation, a new formula for generating a composite sample is as follows:
;
Wherein, Is a polynomial function. Meanwhile, the dynamic adjustment is carried out according to the distribution condition of the synthesized sample of the current iterationValues and nearest neighbor selection criteria to generate a more diverse and higher quality synthetic sample;
Polynomial function The definition is as follows:
;
Wherein, Is a coefficient of a polynomial, and K is the highest degree of the polynomial. In one embodiment, the highest degree K of the polynomial is set to 2.
And S15, the iteration process is continuously carried out until the preset iteration times are reached, and a final expansion data set is output. In one embodiment, the preset number of iterations is 500.
Example 3
The present embodiment adopts a multilayer neural network for feature extraction, the neural network including:
Input layer: setting the number of neurons according to the characteristic number of the input data;
hidden layer: comprising 2 hidden layers, each hidden layer having 50 and 80 neurons, respectively;
output layer: the number of neurons of the output layer was set to 100.
In this embodiment, the process of training the neural network by using the ecosystem optimization algorithm is as follows:
S21, generating an initial ecological system which comprises a plurality of populations, wherein each population represents a parameter set in the neural network, the fitness of each individual is determined by the performance of each individual on a specific task, and initializing the ecological system. Specifically, an ecosystem comprising N individuals is constructed, item Initial fitness of individualDepending on its performance on a particular task,; Wherein,Representing a variable of the environment,Represents a genetic factor of the human body,Is the firstA set of parameters for the individual person,WhereinAndRespectively representing the weight and the bias of the neural network;
further, the initial fitness calculation function is expressed as follows:
;
Wherein, As a function of the genetic factors,As a function of the performance evaluation,Is an environmental interaction function that describes the dynamic interaction between individual parameters and environmental variables. Sig () is a Sigmoid activation function,The performance evaluation, the environment variable, the genetic factor and the weight coefficient of the interaction between the individual parameter and the environment are respectively used for balancing the contribution of each factor to the adaptability.
Further, genetic factor functionConsidering the diversity and complexity of the parameters, the calculation mode can be expressed as:
;
Wherein, AndThe first weight coefficient and the second weight coefficient are respectively used for adjusting the contribution of each item; The variance of the weights and biases is expressed for measuring the diversity of the parameters, n is the total number of neurons, w ij is the weight of the jth neuron in the ith individual parameter, and b ij is the bias of the jth neuron in the ith individual parameter.
Environmental interaction functionCan be expressed as:
;
Wherein, Is the third weight coefficient of the weight,Represents the first of the environmental variablesThe number of elements to be added to the composition,Representing and environment variablesThe input characteristics of the data are associated with,Is a hyperbolic tangent function.
Performance evaluation functionBased on the performance of the neural network on a specific task, considering that the better the performance is in general, the smaller the loss function value is, the performance is expressed by using the negative value or the inverse of the loss function, and can be expressed as:
;
Wherein, Representing a loss function and,Is a real label; is a predictive tag derived from a preset Softmax function.
S22, calculating the fitness of each individual to determine the viability of each individual in the physiological system, wherein individuals with higher fitness have a greater chance to reproduce offspring. Specifically, the fitness is calculated based on the current state of the individual and its interactions with other individuals in the ecosystemFor individualsAndThe strength of interaction between the two, the fitness updating function of the individual can be expressed as:
;
Wherein, To take into account fitness of the ith individual after ecological interaction; Is a weight adjustment coefficient representing the degree of influence of interactions between individuals; To adapt the competition strength between the ith individual and the jth individual, Is the symbiotic intensity of the ith individual and the jth individual after self-adaption adjustment. In one embodiment of the present invention, in one embodiment,Is set to 0.5.
Further, the method comprises the steps of,AndIs based on a performance feedback functionAdjustment, i.e. to the competitive strengthAnd symbiotic strengthThe adaptive adjustment is performed in a manner expressed as:
;
;
Wherein, For the competition strength of the ith individual and the jth individual, For the symbiotic intensity of the ith individual and the jth individual,AndThe adaptive functions that adjust the competition and symbiotic strengths respectively,Representing a performance feedback functionThe amount of change, i.e. the difference between the current evaluation and the last evaluation.
Further, performance feedback functionsThe overall state of the current ecological network is evaluated based on the performance of the neural network on the validation set. In one embodiment, the performance feedback functionThe calculation of (2) can be expressed as:
;
Wherein, The weight coefficients of accuracy, complexity, diversity and time are used to balance the contributions of different indexes.
And, each component function is respectively:
Accuracy function :;
Complexity function:;
Diversity function:;
Time function:;
Wherein P is the accuracy rate, L is the average loss value; c is the network complexity, and in this embodiment, is proportional to the parameter; h is a data diversity index proportional to the variance of the data features; t is training time, which is proportional to the number of iterations.
Further, the competition strength is calculated according to the resource requirement and the overlapping degree, and the calculation mode can be expressed as follows:
;
Wherein, AndRepresents the ith individual and the jth individual pair, respectivelyThe need for seed resources is addressed,Is the total category number of the resource and an exponential functionThe effect of (a) is to smooth out the effects of inter-individual demand variation, ensuring that even small demand variation can be reflected in competitive strength.
And, the symbiotic strength reflects the degree of reciprocal cooperation between two individuals, measured by their efficiency of sharing resources, can be expressed as:
。
S23, simulating a natural selection process, selecting individuals with higher fitness to reproduce, and simultaneously eliminating individuals with lower fitness, wherein the probability of the ith individual being selected The calculation of (2) can be expressed as:
;
where p i is the probability that the ith individual is selected.
S24, regenerating and reproducing, simulating the genetic variation process of organisms, and generating offspring and offspring by crossing and variation operations in genetic algorithm of individuals with higher adaptabilityGenerated by the following crossover formula:
;
Wherein, Is a crossover coefficient, and determines the fusion proportion of the genetic information of the father; And Parameter sets for the ith and jth parent individuals, respectively. In one embodiment of the present invention, in one embodiment,Set to 0.7.
Further, the mutation operation is performed by adding a genetic mutation disturbanceRealized by the method that the size is subject to variation rateCan be expressed as:
;
Wherein, A set of parameters for offspring individuals; for the mutation rate, the embodiment is set to 0.1, and the amplitude of mutation operation is controlled; is a disturbance of genetic variation.
The role of the perturbation of genetic variation is to increase the diversity of the population, which can be expressed as:
;
Wherein, Is the disturbance amplitude control coefficient,Is one and withRandom vectors of the same dimension, the elements of which are distributed uniformlyAnd (3) extracting. In one embodiment of the present invention, in one embodiment,Is set to 0.05 to control the magnitude of the variation.
S25, simulating ecological interactions between individuals, such as competition and symbiotic relationships, which affect the fitness and viability of individuals. Specifically, an ecological interaction matrix I is set, where the element I ij represents a relationship type between the ith individual and the jth individual, and the relationship type affects fitness update of the individuals and may be expressed as:
;
Wherein, To take into account the fitness after ecological interactions.
S26, simulating the influence of environmental changes on the ecological system, such as resource changes, new adaptability challenges and the like, forcing the individual to adapt to the new environment, and further optimizing parameters in the ecological system. In particular, the environmental change is generated by modifying an environmental complexity factorTo simulate, thereby affecting the fitness calculation of the individual, the fitness adjustment caused by the environmental stress can be expressed as:
;
Wherein, In order to finally adjust the degree of adaptability,Is an environmental complexity coefficient and is preset manually; Is an environmental change factor that describes the extent of environmental change from the previous generation to the current generation, and can be expressed as:
;
Wherein, Is an individualThe degree of adaptation in the current environment,Is the adaptability in the previous generation environment.
And S27, updating parameters of the neural network according to the fitness and ecological dynamics of the individuals in the ecological system so as to reflect the evolution of the biological individuals in the ecological system. Specifically, the fitness after adjustment based on ecological interactions and environmental pressureUpdating parameters of an individual can be expressed as:
;
Wherein, In order to update the set of individual parameters,The learning rate of the individual parameter set update is set to be 0.01 in the embodiment, and the step length of parameter update is represented; Is the gradient pointing in the direction of maximum increase in fitness.
Example 4
The embodiment adopts an improved self-encoder algorithm to carry out dimension reduction processing on the extracted characteristics, and optimizes the data characteristics aiming at the problem of electric vehicle charging station site selection. The traditional self-encoder realizes the compression and reconstruction of data through a series of linear or nonlinear transformation, the embodiment introduces quantum mapping characteristics on the basis, and the density and diversity of the encoding are increased by utilizing the concept of quantum bits, so that the efficiency and the precision of feature dimension reduction are improved. Meanwhile, in order to adapt to the complexity of the electric vehicle charging station site selection problem, the self-encoder of the embodiment adopts a dynamic encoding adjustment mechanism, and parameters in the encoding process are adjusted in real time according to data characteristics and model performance so as to optimize characteristic representation. In addition, while the feature reduces the dimension, the embodiment considers multiple targets of the addressing problem, such as cost minimization, coverage rate maximization and the like, and ensures that the feature after dimension reduction can effectively support complex decision requirements by introducing a multi-target optimization strategy.
Specifically, the process of performing dimension reduction processing on the extracted features by adopting the improved self-encoder algorithm is as follows:
S31, setting the feature data set after feature extraction to be expressed as Initializing the structure and parameters of the self-encoder, including the following parameters:
Encoder layer number Set to 3;
decoder layer number In generalSet to 3;
number of neurons per layer Set to 128;
Activation function Specifically, sigmoid activation functions.
S32, data are transmitted forwards through the encoder part, and the output of each layer is the input of the next layer until the encoded layer is reached; at the coding layer, the data is compressed into a low-dimensional representation of the features. Specifically, the output of each layer of the encoder can be expressed as:
;
Wherein, AndEncoder respectively NoThe weight and bias of the layers are such that,Encoder respectively NoLayer (a)Layer output, for encoderLayer output,Elements representing a feature dataset;
and, the encoded low-dimensional features are expressed as ,Is the encoder ofAnd (3) outputting the layer.
S33, reconstructing the coded features through a decoder part, wherein the aim is to restore the input data as much as possible, and the reconstructed quality directly influences the performance of the model. Specifically, the way in which the output of each layer of the decoder is calculated can be expressed as:
;
Wherein, AndRespectively decoder thThe weight and bias of the layers are such that,Respectively decoder thLayer (a)The output of the layers, for the decoder first layer,。
And, the reconstructed data is represented as,Is decoder NoAnd (3) outputting the layer.
S34, adopting a combined loss function of a reconstruction error and a regularization termModel parameters are updated according to gradients of the loss functions by using an Adam optimization algorithm, and the loss functions are combinedThe calculation of (2) can be expressed as:
;
Wherein, AndRespectively representing the Frobenius norm and the L2 norm,Is a regularization coefficient.
Updating model parameters through Adam optimization algorithm,,,To minimize the loss function。
During the training of the self-encoder, the weightsBias and method of making sameThe updating of (2) is realized by a back propagation and gradient descent algorithm, and the weight updating mode can be expressed as follows:
;
;
Wherein, Is the learning rate updated from the encoder weights,AndRespectively, combined loss functionFor weightBias and method of making sameIs a partial derivative of (c).Is the updated encoder weight that is to be used,Is the pre-update encoder weight.Is the updated encoder bias that is to be used,Is the pre-update encoder bias. In one embodiment, the learning rate of the self-encoder weight updateSet to 0.001.
Regularization termThe calculation of (2) can be expressed as:
;
Wherein, Representation encoder NoLayer weight matrix of the first layerLine 1The elements of the column are arranged such that,Representation encoder NoLayer offset vector of the first layerThe elements. In one embodiment, the regularization coefficientsIs set to 0.01.
S35, in the training process, according to the model performance and the multi-objective requirements (such as cost minimization objectiveAnd coverage maximizing goal) The encoding parameters are dynamically adjusted. In one embodiment, this is achieved by introducing additional penalty terms, which can be expressed as:
;
Wherein, Is the goal of cost minimizationIs used for the weight of the (c),Is coverage maximizationIs a weight of (2).
Further, cost minimization objectivesThe calculation of (2) can be expressed as:
;
Wherein, Represent the firstThe cost of construction and operation of the individual electric vehicle charging stations,Is a decision variable indicating whether or not to be in the electric vehicle charging station positionAnd (5) constructing a charging station.
And, coverage maximizing the targetThe calculation of (2) can be expressed as:
;
Wherein, Is the total number of points of demand,Is a set of selected charging station locations,Representing a demand pointTo electric vehicle charging station locationIs a distance of (3).
S36, iterating the steps S32 to S35 until the preset maximum iteration times are met.
Example 5
According to the embodiment, the classifier is constructed by adopting the extreme learning machine classification algorithm based on low-rank sparsity, and the generalization capability and the interpretation of the classifier are improved by introducing low-rank sparsity constraint, so that the classifier can still keep high efficiency and accuracy when processing large-scale electric vehicle charging station data. At the same time, the combination of robust optimization strategies makes the classifier more robust against data noise and outliers. Specifically, the flow of the extreme learning machine classification algorithm based on low-rank sparsity is as follows:
S41, representing the feature data set after dimension reduction as . Initializing input weights for extreme learning machinesBias and method of making sameIn one embodiment, the manner of initialization follows a gaussian distribution random generation, which can be expressed as:
;
;
Wherein, Is the variance of the distribution.
S42, data are transmitted forwards through a hidden layer of the extreme learning machine, hidden layer nodes adopt Sigmoid nonlinear activation functions, and the hidden layer outputsThe calculation of (2) can be expressed as:
;
Wherein, The ith feature vector of the feature data set after the dimension reduction,Is a Sigmoid nonlinear activation function, and is specifically expressed as follows:
;
Wherein the method comprises the steps of The output of the Sigmoid nonlinear activation function is represented, x is the input, and e is the natural constant.
Thus, for the input feature vector, the output of the hidden layer is calculated as:
;
Wherein, Representing the output of the hidden layer.
S43, calculating output weight by solving an optimization problem by using low-rank sparse constraint and a robust optimization strategy. The optimization problem aims to ensure low rank and sparsity of the output weights while minimizing training errors. In one embodiment, the weights are output taking into account low rank sparsity constraintsThe solution of (2) translates into the following optimization problem:
;
Wherein, Is the output matrix of the object and,AndIs a regularization parameter controlling sparsity and low rank,Is the Frobenius norm,Is the L1 norm for promoting sparsity.The rank of the matrix is represented for facilitating low rank performance.
Further, L1 normFor promoting sparsity, the calculation method can be expressed as:
;
And, rank of matrix It is generally more difficult to optimize directly, so the kernel norms (sum of matrix singular values) can be used as their convex relaxation, which can be expressed as:
;
Wherein, Representation matrixIs characterized by a convex relaxation of (1),Representation matrixIs the first of (2)Singular values.
S44, evaluating the performance of the classifier in a cross-validation mode, and adjusting classifier parameters including the number of hidden layer nodes, regularization parameters and the like so as to achieve the optimal classification effect. Classifier performance is evaluated by comparing the difference of the predicted output to the actual target output. In case of poor performance, parameters can be adjusted, such as increasing or decreasing the number of hidden layer nodes, or regularization parametersAnd。
Example 6
The embodiment provides an electric vehicle charging station site selection system based on artificial intelligence, including:
The data acquisition device is used for acquiring public transportation data, social and economic data, electric vehicle use data and geographic information system data, marking the data and expanding the data to obtain an electric vehicle charging pile site selection factor data set;
The feature extraction module is used for training the neural network by adopting an ecological system optimization algorithm, and extracting features of the electric vehicle charging pile site selection factor dataset by adopting the trained neural network;
The dimension reduction module is used for carrying out dimension reduction treatment on the extracted features;
And the classifier is used for classifying according to the feature after dimension reduction to obtain the address selection classification result of the electric vehicle charging station.
Example 7
In another embodiment, a non-volatile computer storage medium is provided, the computer storage medium having stored thereon computer-executable instructions for performing the artificial intelligence based electric vehicle charging station location method of embodiments 1-6.
Example 8
The present embodiment provides an electronic device including: the system comprises at least one processor and a memory communicatively coupled to the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the artificial intelligence based electric vehicle charging station location method of embodiments 1-6.
The above-described specific embodiments further illustrate the objects, technical solutions and technical effects of the present invention in detail. It should be understood that the foregoing is only illustrative of the present invention and is not intended to limit the scope of the invention, and that all equivalent changes and modifications that may be made by those skilled in the art without departing from the spirit and principles of the invention shall fall within the scope of the invention.
Claims (10)
1. The electric vehicle charging station location selection method based on artificial intelligence is characterized by comprising the following steps of:
S1, collecting public transportation data, social and economic data, electric vehicle use data and geographic information system data, and performing data annotation and data expansion to obtain an electric vehicle charging pile site selection factor data set;
s2, training a neural network by adopting an ecological system optimization algorithm, and extracting features of an electric vehicle charging pile site selection factor dataset by adopting the trained neural network;
s3, performing dimension reduction treatment on the extracted features;
s4, inputting the feature subjected to dimension reduction into a classifier, and training the classifier;
S5, collecting site selection factor data of the electric vehicle charging piles to be classified, extracting features and performing dimension reduction, and inputting the obtained dimension reduced features into a trained classifier for classification to obtain site selection classification results of the electric vehicle charging stations;
The training process of the neural network by adopting the ecological system optimization algorithm comprises the following steps:
s21, generating an initial ecological system, and initializing the ecological system;
S22, calculating the fitness of each individual; the fitness updating function of the individual is:
;
Wherein, For individuals/>And/>Intensity of interaction between,/>To take into account fitness of the ith individual after ecological interaction; /(I)Is a weight adjustment coefficient; /(I)Is the competition strength of the ith individual and the jth individual after self-adaption adjustment,/>The symbiotic intensity of the ith individual and the jth individual after self-adaptive adjustment;
And/> Is based on the performance feedback function/>Adjusting; performance feedback function/>Evaluating the overall state of the current ecological network based on the performance of the neural network on the verification set;
s23, simulating a natural selection process, selecting individuals with high fitness for reproduction, and eliminating individuals with low fitness at the same time;
s24, regenerating and reproducing, simulating the genetic variation process of organisms, and generating offspring by crossing and variation operations in a genetic algorithm for individuals with high adaptability;
S25, simulating ecological interaction among individuals, and updating the fitness of the individuals;
S26, simulating the influence of environmental changes on an ecological system, and adjusting the fitness;
And S27, updating parameters of the neural network according to the fitness and ecological dynamics of the final individual in the ecological system.
2. The method for locating an electric vehicle charging station based on artificial intelligence of claim 1, wherein the data expansion is performed by adopting an improved quadratic programming SMOTE algorithm, comprising the steps of:
S11, carrying out sample statistics of each category on an original data set, determining a few types of samples, evaluating the unbalance degree of the original data set, and initializing the iteration times and the number of synthesized samples in each iteration according to a statistical result;
S12, in the first layer iteration, for each minority sample, finding out the nearest neighbor sample set in the minority sample sets, and generating a synthesized sample according to each nearest neighbor sample and the random number;
s13, based on the distance between the synthesized sample and the nearest original sample, evaluating the quality of the synthesized sample through a quality evaluation function, and adjusting the generation strategy of the synthesized sample according to the evaluation result;
s14, in subsequent iterations, continuing to refine the generation process of the synthetic sample, and gradually optimizing the representativeness and diversity of the synthetic sample by continuously evaluating the quality of the synthetic sample and adjusting parameters;
and S15, the iteration process is continuously carried out until the preset iteration times are reached, and a final expansion data set is output.
3. The method for locating an electric vehicle charging station based on artificial intelligence according to claim 1, wherein the dimension reduction process of the extracted features using the modified self-encoder algorithm is as follows:
S31, initializing the structure and parameters of a self-encoder;
S32, data are transmitted forwards through the encoder part, and the output of each layer is the input of the next layer until the last encoding layer is reached;
s33, reconstructing the coded features through a decoder part;
S34, adopting a combined loss function of the reconstruction error and the regularization term, and updating model parameters according to the gradient of the loss function by using an Adam optimization algorithm;
s35, dynamically adjusting coding parameters according to model performance and multi-target requirements in the training process;
S36, iterating the steps S32 to S35 until the preset maximum iteration times are met.
4. The method for locating an electric vehicle charging station based on artificial intelligence according to claim 1, wherein a classifier is constructed by adopting a low-rank sparse-based extreme learning machine classification algorithm, and the flow of the low-rank sparse-based extreme learning machine classification algorithm is as follows:
S41, initializing input weight and bias of an extreme learning machine;
S42, data are transmitted forwards through a hidden layer of the extreme learning machine;
S43, calculating output weight by solving an optimization problem by using a low-rank sparse constraint and a robust optimization strategy; output weight The solution of (2) translates into the following optimization problem:
;
Wherein, Is the target output matrix,/>And/>Is a regularization parameter controlling sparsity and low rank,/>Is the Frobenius norm,/>Is L1 norm,/>Representing the rank of the matrix;
S44, evaluating the performance of the classifier in a cross-validation mode, and adjusting the parameters of the classifier.
5. The method for locating an electric vehicle charging station based on artificial intelligence of claim 2, wherein in step S12, a composite sample is generatedThe formula of (2) is as follows: /(I); Wherein/>Representing the i-th minority class sample,/>Represents the j nearest neighbor sample,/>Is a random number,/>For controlling at/>And/>The degree of interpolation between;
the specific process of step S13 is as follows: defining a quality assessment function of the composite sample as An adaptive adjustment mechanism is introduced to adjust the random number/>Values of (5) such that/>Maximization, updating random number/>The strategy of (2) is as follows:
;
Wherein, Is updated random number,/>Is the learning rate of random number update,/>Is/>With respect to random number/>Is a gradient of (2);
Quality assessment function Expressed as:
;
Wherein, Is/>Nearest neighbor samples in the original minority class sample set,/>Representing Euclidean distance,/>Is a parameter controlling the shape of the function;
Calculated according to the chain law, expressed as:
;
Wherein, Is/>Concerning/>Partial derivative of/(I)Is/>Concerning/>Is a partial derivative of (c).
6. The method of claim 1, wherein generating an initial ecosystem and initializing the ecosystem comprises:
Constructing an ecosystem comprising N individuals, the first Initial fitness/>, of individualDepending on its performance on a particular task,/>; Wherein/>Representing an environmental variable,/>Representing the genetic factor,/>For/>A set of parameters for the individual person,Wherein/>And/>Respectively representing the weight and the bias of the neural network;
The initial fitness calculation function is expressed as follows:
;
Wherein, As a function of genetic factors,/>As a performance evaluation function,/>Is an environmental interaction function describing the dynamic interaction between individual parameters and environmental variables; sig () is a Sigmoid activation function,The weight coefficients of performance evaluation, environment variables, genetic factors, individual parameters and environment interactions are respectively;
Genetic factor function The calculation mode is expressed as follows:
;
Wherein, And/>The first weight coefficient and the second weight coefficient are respectively; /(I)Representing the variance of the weights and biases for measuring the diversity of the parameters, n representing the total number of neurons,/>Representing the weight of the jth neuron in the ith individual parameter,/>Representing the bias of the jth neuron in the ith individual parameter;
environmental interaction function Expressed as:
;
Wherein, Is a third weight coefficient,/>/>, Representing environmental variablesElement,/>Represents and environmental variable/>Associated input features,/>Is a hyperbolic tangent function;
Performance evaluation function Expressed as:
;
Wherein, Representing a loss function,/>Is a real label; /(I)Is a predictive tag derived from a preset Softmax function.
7. The method according to claim 6, wherein in step S22, the competition strength is determined byAnd symbiotic strength/>The manner of making the adaptation is expressed as:
;
;
Wherein, Is the competitive strength of the ith individual and the jth individual, S ij is the symbiotic strength of the ith individual and the jth individual,/>And/>Adaptive functions for adjusting the competitive and symbiotic strengths,/>, respectivelyRepresenting a performance feedback function/>A variation amount;
Performance feedback function The calculation mode of (a) is expressed as follows:
;
Wherein, The weight coefficients of accuracy, complexity, diversity and time, respectively;
the functions of each component are respectively as follows:
Accuracy function :/>;
Complexity function:/>;
Diversity function:/>;
Time function:/>;
Wherein P is the accuracy rate, L is the average loss value; c is network complexity and is in direct proportion to parameter quantity; h is a data diversity index proportional to the variance of the data features; t is training time and is in direct proportion to the iteration times;
;
;
Wherein, And/>Represents the ith and jth individual pairs/>, respectivelyResource requirement,/>Is the total category number of the resource.
8. An artificial intelligence based electric vehicle charging station location system, comprising:
The data acquisition device is used for acquiring public transportation data, social and economic data, electric vehicle use data and geographic information system data, marking the data and expanding the data to obtain an electric vehicle charging pile site selection factor data set;
The feature extraction module is used for training the neural network by adopting an ecological system optimization algorithm, and extracting features of the electric vehicle charging pile site selection factor dataset by adopting the trained neural network;
The dimension reduction module is used for carrying out dimension reduction treatment on the extracted features;
And the classifier is used for classifying according to the feature after dimension reduction to obtain the address selection classification result of the electric vehicle charging station.
9. A non-transitory computer storage medium storing computer executable instructions for performing the artificial intelligence based electric vehicle charging station location method of any one of claims 1-7.
10. An electronic device, comprising: at least one processor, and a memory communicatively coupled to the at least one processor, wherein the memory stores instructions executable by the at least one processor to cause the at least one processor to perform the artificial intelligence based electric vehicle charging station location method of any one of claims 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410381566.7A CN117974221A (en) | 2024-04-01 | 2024-04-01 | Electric vehicle charging station location selection method and system based on artificial intelligence |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410381566.7A CN117974221A (en) | 2024-04-01 | 2024-04-01 | Electric vehicle charging station location selection method and system based on artificial intelligence |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117974221A true CN117974221A (en) | 2024-05-03 |
Family
ID=90866052
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410381566.7A Pending CN117974221A (en) | 2024-04-01 | 2024-04-01 | Electric vehicle charging station location selection method and system based on artificial intelligence |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117974221A (en) |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106326843A (en) * | 2016-08-15 | 2017-01-11 | 武汉工程大学 | Face recognition method |
CN109325844A (en) * | 2018-06-25 | 2019-02-12 | 南京工业大学 | Network loan borrower credit evaluation method under multidimensional data |
CN111178580A (en) * | 2019-11-28 | 2020-05-19 | 浙江工业大学 | Supermarket site selection method based on improved BP neural network |
CN111311122A (en) * | 2020-03-23 | 2020-06-19 | 广东工业大学 | Charging station construction method for smart city |
CN113642757A (en) * | 2021-06-01 | 2021-11-12 | 北京慧辰资道资讯股份有限公司 | Internet of things charging pile construction planning method and system based on artificial intelligence |
CN115271550A (en) * | 2022-09-26 | 2022-11-01 | 国网浙江慈溪市供电有限公司 | Electric vehicle public charging station site selection and volume fixing method, computing equipment and storage medium |
CN115761435A (en) * | 2022-11-07 | 2023-03-07 | 广东工业大学 | Active learning anomaly detection method and system based on graph propagation |
CN116013479A (en) * | 2022-11-08 | 2023-04-25 | 北京理工大学 | Multi-objective optimization-based intelligent distribution robot system resource dynamic planning method |
CN116307181A (en) * | 2023-03-17 | 2023-06-23 | 太原科技大学 | Intelligent site selection and volume determination method for electric vehicle charging station |
CN117374941A (en) * | 2023-10-13 | 2024-01-09 | 北方工业大学 | Photovoltaic power generation power prediction method based on neural network |
CN117557069A (en) * | 2024-01-10 | 2024-02-13 | 长峡数字能源科技(湖北)有限公司 | Charging pile site selection method and system |
CN117789038A (en) * | 2024-02-26 | 2024-03-29 | 聊城莱柯智能机器人有限公司 | Training method of data processing and recognition model based on machine learning |
-
2024
- 2024-04-01 CN CN202410381566.7A patent/CN117974221A/en active Pending
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106326843A (en) * | 2016-08-15 | 2017-01-11 | 武汉工程大学 | Face recognition method |
CN109325844A (en) * | 2018-06-25 | 2019-02-12 | 南京工业大学 | Network loan borrower credit evaluation method under multidimensional data |
CN111178580A (en) * | 2019-11-28 | 2020-05-19 | 浙江工业大学 | Supermarket site selection method based on improved BP neural network |
CN111311122A (en) * | 2020-03-23 | 2020-06-19 | 广东工业大学 | Charging station construction method for smart city |
CN113642757A (en) * | 2021-06-01 | 2021-11-12 | 北京慧辰资道资讯股份有限公司 | Internet of things charging pile construction planning method and system based on artificial intelligence |
CN115271550A (en) * | 2022-09-26 | 2022-11-01 | 国网浙江慈溪市供电有限公司 | Electric vehicle public charging station site selection and volume fixing method, computing equipment and storage medium |
CN115761435A (en) * | 2022-11-07 | 2023-03-07 | 广东工业大学 | Active learning anomaly detection method and system based on graph propagation |
CN116013479A (en) * | 2022-11-08 | 2023-04-25 | 北京理工大学 | Multi-objective optimization-based intelligent distribution robot system resource dynamic planning method |
CN116307181A (en) * | 2023-03-17 | 2023-06-23 | 太原科技大学 | Intelligent site selection and volume determination method for electric vehicle charging station |
CN117374941A (en) * | 2023-10-13 | 2024-01-09 | 北方工业大学 | Photovoltaic power generation power prediction method based on neural network |
CN117557069A (en) * | 2024-01-10 | 2024-02-13 | 长峡数字能源科技(湖北)有限公司 | Charging pile site selection method and system |
CN117789038A (en) * | 2024-02-26 | 2024-03-29 | 聊城莱柯智能机器人有限公司 | Training method of data processing and recognition model based on machine learning |
Non-Patent Citations (3)
Title |
---|
张洋: "SMOTE算法的改进与应用", 《中国优秀硕士学位论文全文数据库 (信息科技辑)》, 15 January 2021 (2021-01-15), pages 20 * |
李海峰;康中敏;: "基于混合遗传神经网络的电动汽车充电站最优选址的研究", 四川电力技术, no. 04, 20 August 2012 (2012-08-20), pages 49 * |
杨现青;刘法胜;杨玲玲;: "电动汽车充电站选址优化研究", 交通科技与经济, no. 03, 25 June 2016 (2016-06-25) * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Cui et al. | A new hyperparameters optimization method for convolutional neural networks | |
CN110084424A (en) | A kind of Methods of electric load forecasting based on LSTM and LGBM | |
Assareh et al. | Application of PSO (particle swarm optimization) and GA (genetic algorithm) techniques on demand estimation of oil in Iran | |
CN110929843A (en) | Abnormal electricity consumption behavior identification method based on improved deep self-coding network | |
CN103605909B (en) | It is a kind of based on gray theory and the water quality prediction method of support vector machine | |
CN106910337A (en) | A kind of traffic flow forecasting method based on glowworm swarm algorithm Yu RBF neural | |
CN109214449A (en) | A kind of electric grid investment needing forecasting method | |
CN103105246A (en) | Greenhouse environment forecasting feedback method of back propagation (BP) neural network based on improvement of genetic algorithm | |
CN117541026B (en) | Intelligent logistics transport vehicle dispatching method and system | |
Wang et al. | A regional pretraining-classification-selection forecasting system for wind power point forecasting and interval forecasting | |
CN104951847A (en) | Rainfall forecast method based on kernel principal component analysis and gene expression programming | |
CN114548591B (en) | Sequential data prediction method and system based on mixed deep learning model and Stacking | |
CN114492191A (en) | Heat station equipment residual life evaluation method based on DBN-SVR | |
CN115438833A (en) | Short-term power load hybrid prediction method | |
CN111311001B (en) | Bi-LSTM network short-term load prediction method based on DBSCAN algorithm and feature selection | |
CN116842354A (en) | Feature selection method based on quantum artificial jellyfish search mechanism | |
CN115907122A (en) | Regional electric vehicle charging load prediction method | |
CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
CN109408896B (en) | Multi-element intelligent real-time monitoring method for anaerobic sewage treatment gas production | |
CN108537581B (en) | Energy consumption time series prediction method and device based on GMDH selective combination | |
CN112819062B (en) | Fluorescence spectrum secondary feature selection method based on mixed particle swarm and continuous projection | |
CN114004341A (en) | Optical fiber preform preparation process optimization method based on genetic algorithm and BP neural network | |
CN112183721B (en) | Construction method of combined hydrological prediction model based on self-adaptive differential evolution | |
CN113537539B (en) | Multi-time-step heat and gas consumption prediction model based on attention mechanism | |
CN117291069A (en) | LSTM sewage water quality prediction method based on improved DE and attention mechanism |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |