Method and system for evaluating capacity of intelligent operation charging station for charging and battery swap of electric vehicle

Abstract

The invention discloses an intelligent operation charging station capability evaluation method and system for electric vehicle charging and battery swap, and relates to the technical field of electric vehicle intelligent operation, and the method comprises the steps: receiving the charging operation data of an electric vehicle charging station, and processing the charging operation data of the electric vehicle charging station, the processed charging operation data of the electric vehicle charging station are obtained, and the charging operation data of the electric vehicle charging station comprise active power, a power factor and an operation timestamp; the processed charging operation data of the electric vehicle charging station are input into a pre-established load decomposition model of a deep neural network algorithm based on unsupervised learning optimization, load data are output, and the model is trained based on a K-means clustering algorithm; based on the load data, the power utilization habit is predicted and the callable capability is evaluated, so that the charging station capability evaluation of the intelligent charging and battery swap operation of the electric vehicle is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent operation of electric vehicles, and specifically to a method and system for evaluating the capacity of a charging station for intelligent operation of charging and swapping of electric vehicles. Background Art

[0002] In the context of the carbon peaking action plan, the energy industry has high hopes, the power system has received much attention, and the low-carbon transformation of transportation needs to be accelerated. China is already the country with the largest number of charging facilities, the largest radiation area, and the most complete service vehicles in the world. When a large number of electric vehicles enter the power grid for charging, it will bring challenges to the stability of the power grid. In particular, once a large number of cluster vehicles enter the power grid for charging, these vehicles can be regarded as a kind of power storage system in the vehicle-grid interaction of vehicle-to-grid control. Under the control of vehicle-to-grid, electric vehicles can act as both loads and distributed power sources, becoming participants in the operation of auxiliary systems. However, in the traditional vehicle-to-grid solution, the charging model of electric vehicles is simplified to a known power model, while in reality, there are differences in the charging power and voltage harmonic characteristics of electric vehicles, and there is uncertainty in the charging time. How to sense the load of electric vehicles has become the key to the operation of the auxiliary power system of cluster electric vehicles.

[0003] Currently, a large number of published V2G algorithms for electric vehicle charging stations use parameter data of electric vehicle charging stations and electric vehicle models obtained by direct settings, rather than the characteristics of power and voltage under actual electric vehicle charging conditions, which reduces the practicality of the solution.

[0004] The patent with publication number CN117477593A is a method for evaluating the frequency modulation capability of a charging station based on electric vehicles. According to each initial state of charge, each expected state of charge, each expected end time, and each rated charging power, each electric vehicle is classified to obtain the charging category of each electric vehicle, determine the charging process of each electric vehicle from the start time to the expected end time, and determine the frequency modulation power of the charging station. Its data is not the characteristics of power and voltage under the actual charging of electric vehicles, resulting in reduced practicality. The patent with publication number CN114069604B considers the multi-period capacity evaluation method of electric vehicle charging station coupling. It is aimed at evaluating the available capacity of the operating charging station, but cannot effectively realize the evaluation of random loads and lacks mobility.

[0005] Most of the solutions that use non-invasive load monitoring to monitor electric vehicle charging stations still have the following problems. First, the charging power characteristics of each electric vehicle at an electric vehicle charging station are almost the same, which causes certain difficulties for identification and clustering. For example, there are small differences in the charging characteristics of electric vehicles caused by differences in electric vehicle models within each brand, or brand differences between brands, but due to the different battery capacities, they cannot be ignored, so it will cause certain troubles for modeling and identification. Secondly, there are regular charging loads and random charging loads in the charging conditions of each electric vehicle at an electric vehicle charging station. The randomness of the occurrence time and duration of random charging loads will cause certain deviations in the calculation method of the supervised learning algorithm based on user habits, which will cause errors in the evaluation of controllable capabilities. Summary of the invention

[0006] In order to solve the deficiencies mentioned in the above background technology, the purpose of the present invention is to provide a method and system for evaluating the capacity of charging stations for intelligent operation of electric vehicle charging and swapping, which can accurately evaluate the controllable capacity of electric vehicles.

[0007] In the first aspect, the purpose of the present invention can be achieved by the following technical solution: a method for evaluating the capacity of an intelligent charging station for electric vehicle charging and battery replacement, the method comprising the following steps:

[0008] Receiving charging operation data of an electric vehicle charging station, processing the charging operation data of the electric vehicle charging station, and obtaining processed charging operation data of the electric vehicle charging station, wherein the charging operation data of the electric vehicle charging station includes active power, power factor, and operation timestamp;

[0009] Inputting the processed charging operation data of the electric vehicle charging station into a pre-established load decomposition model based on a deep neural network algorithm optimized by unsupervised learning, and outputting load data, wherein the pre-established load decomposition model based on a deep neural network algorithm optimized by unsupervised learning is trained based on a K-means clustering algorithm;

[0010] Based on load data, electricity usage habits are predicted and available capacity is evaluated, thereby realizing charging station capacity assessment for intelligent operation of electric vehicle charging and swapping.

[0011] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: when evaluating the controllable capability of the electric vehicle, the electric vehicle needs to enter a controllable idle state, wherein the determination process of whether to enter the idle state is as follows:

[0012] The remaining power of electric vehicles entering the power grid is judged by the average daily travel distance, and the travel demand of the owner on the next day is used to determine how long it takes for the electric vehicle to enter the idle state after entering the power grid. The user's average daily travel distance obeys the log-normal distribution, and the probability density function is:

[0013]

[0014] Where d represents the average daily driving distance, μ d and σ d It represents the expectation and variance of the log-normal distribution, and sets the predicted remaining power of the vehicle when it enters the power grid and the conditions when the vehicle enters the idle state in the current electric vehicle charging station system according to probability.

[0015] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: calculating a charging time that is expected to be required for the electric vehicle to enter an idle controllable state:

[0016]

[0017] Among them, β refers to the amount of electricity required to travel one kilometer, d is the probability density of the average daily travel distance of electric vehicle users, and P is the minimum daily charging amount of this type of electric vehicle.

[0018] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: the construction of the unsupervised learning optimized deep neural network algorithm includes the construction of a basic neural network algorithm and the construction of a neural network based on non-invasive equipment monitoring NILM.

[0019] In combination with the first aspect, in some implementations of the first aspect, the method further includes: a construction process of the basic neural network algorithm is as follows:

[0020] Artificial Neurons:

[0021] x1~x i Represents all the inputs of this artificial neuron from other neurons or systems. Each input x i There is a corresponding weight w i , the direction of neural network learning is the direction of x transmission, x is the content transmitted from the previous neuron or system to the next neuron, then the input matrix X is described as:

[0022] X=[x0,x1,x2,…,x n ]

[0023] The expression of the weight vector W is:

[0024] W=[w i0 ,w i1 w i2 ,…,win ]

[0025] The connection between the input and output of an artificial neuron is shown below:

[0026]

[0027] And net i As the net activation of the neural network neurons, it can be simplified using vectors as follows:

[0028] ne t =XW

[0029] The final output y i The activation function expression is simplified as follows:

[0030] y i =f(net i )=f(XW)

[0031] For a three-layer feedforward neural network, X is used as the input vector of the neural network, W is used as the connection weight vector of each layer, and F is used as the activation function. Then the expression of the output Y1, Y2, and Y3 of each layer is:

[0032] Y1=F1(XW1)

[0033] Y2=F2(F1(XW1)W2)

[0034] Y3=F3(F2(F1(XW1)W2)W3.

[0035] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the neural network construction process based on non-invasive device monitoring NILM is as follows:

[0036] Select the differentiable activation function sigmoid as the activation function, as shown in the following formula:

[0037]

[0038] For the three-layer BP neural network model, when the input and output are not equal, there is an output error E, which is expressed as:

[0039]

[0040] When the output error is transferred to the hidden layer, the error expression is:

[0041]

[0042] If it is further transferred to the input layer, it becomes the input error, which is expressed as:

[0043]

[0044] The input error is related to the weights w of each layer jk and v ij Therefore, the adjustment of the weights of each layer will affect the change of the function error result E. The principle of weight adjustment is to reduce the error so that the weight is proportional to the gradient decrease of the error. The expression is:

[0045]

[0046] It is determined by three factors: the learning rate η, the error signal δ of the output signal of this layer, and the input signal Y of this layer. The difference between the expected output and the actual output of the network is directly expressed as the output error, which is fed back by the output layer after forward propagation.

[0047] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the process of processing the charging operation data of the electric vehicle charging station is as follows:

[0048] A part of the initial data of the charging operation data of the electric vehicle charging station is used to train the deep neural network. Part of the initial data is multiplexed, and the dimension of the neural network is increased by multi-layer training. The expression is:

[0049]

[0050] Where P is the matrix of all measured powers, P norm is the normalized power matrix of power, P min , P is the minimum and maximum power in the power matrix.

[0051] max

[0052] In combination with the first aspect, in some implementations of the first aspect, the method further includes: when the training is performed based on the K-means clustering algorithm, the process of the K-means clustering algorithm is as follows:

[0053] The first step of the K-means algorithm is to determine K initial centers. The expectation of each determined cluster center is the cluster center of each point. The Euclidean distance is the distance from each collected sample point to the initial cluster center. Each PQ power point is divided into each cluster center point according to the calculated Euclidean distance, and then the average distance is calculated. Repeat this process until the square error standard function meets the requirements. The square error is defined as:

[0054]

[0055] K is the number of cluster centers, and below is the formula for dist(m,x) to calculate the Euclidean distance between points m and x:

[0056]

[0057] When N types of vehicles are designed for clustering, multiple different combinations of cluster centers are obtained from the switching actions on M charging piles. The total number of cluster centers is:

[0058]

[0059] The number of cluster centers calculated by the above formula is used as K for clustering, and the obtained cluster centers can decompose the charging characteristics of each electric vehicle.

[0060] In the second aspect, in order to achieve the above-mentioned purpose, the present invention discloses a charging station capability evaluation system for intelligent operation of electric vehicle charging and swapping, comprising:

[0061] A data processing module, used for receiving charging operation data of the electric vehicle charging station, processing the charging operation data of the electric vehicle charging station, and obtaining the processed charging operation data of the electric vehicle charging station, wherein the charging operation data of the electric vehicle charging station includes active power, power factor and operation timestamp;

[0062] A load decomposition module, used for inputting the processed charging operation data of the electric vehicle charging station into a pre-established load decomposition model based on a deep neural network algorithm optimized by unsupervised learning, and outputting load data, wherein the pre-established load decomposition model based on a deep neural network algorithm optimized by unsupervised learning is trained based on a K-means clustering algorithm;

[0063] The capacity assessment module is used to predict electricity usage habits and evaluate available capacity based on load data, thereby realizing charging station capacity assessment for intelligent operation of electric vehicle charging and swapping.

[0064] In another aspect of the present invention, in order to achieve the above-mentioned purpose, a terminal device is disclosed, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the memory stores a computer program capable of running on the processor, and when the processor loads and executes the computer program, the above-mentioned method for evaluating the capacity of charging stations for intelligent operation of electric vehicle charging and swapping is adopted.

[0065] Beneficial effects of the present invention:

[0066] The present invention generates measured data based on the data characteristics of electric vehicle charging stations, and uses a deep neural network to perceive the load of each electric vehicle, and then uses unsupervised learning to optimize the neural network to perceive random loads and improve the neural network function. Finally, the charging speed and remaining power are used to evaluate the controllable capacity of the electric vehicle charging station. A practical solution for load perception and controllable capacity evaluation of electric vehicle charging stations is proposed, which has the advantages of reliable accuracy, good practicality and high scalability. This is used to solve the shortcomings of the V2G control solution that is not practical enough. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0068] Figure 1 It is a schematic diagram of the process of the present invention;

[0069] Figure 2 It is a schematic diagram of the working process of the present invention;

[0070] Figure 3 This is a power-time curve diagram of an electric vehicle charging station after adding electric vehicle load in an embodiment of the present invention;

[0071] Figure 4 This is a diagram of the deep neural network training results of an embodiment of the present invention;

[0072] Figure 5 This is a diagram showing the decomposition results of the deep neural network for the electric vehicle charging station EV1 without any new loads according to an embodiment of the present invention;

[0073] Figure 6 This is a diagram showing the decomposition result of the new load electric vehicle charging station EV6 by the deep neural network according to an embodiment of the present invention;

[0074] Figure 7 This is a K-means clustering result diagram of an electric vehicle charging station with newly added loads according to an embodiment of the present invention;

[0075] Figure 8 A new deep neural network training result diagram of EV11 is added to the embodiment of the present invention;

[0076] Fig. 9 This is a comparison chart of load perception results for electric vehicle charging stations;

[0077] Fig.10 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0078] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0079] Embodiment 1:

[0080] The following is an introduction to the relevant terms involved in the embodiments of the present application:

[0081] Electric vehicle: Electric vehicle (BEV) refers to a vehicle that uses an onboard power source as a power source, uses an electric motor to drive the wheels, and complies with various requirements of road traffic and safety regulations. As its impact on the environment is relatively small compared to traditional vehicles, its prospects are widely optimistic. Working principle: battery - current - power regulator - motor - power transmission system - drive the vehicle.

[0082] Electric vehicle charging station: Electric vehicle charging station is a station for charging electric vehicles. With the popularity of electric vehicles, electric vehicle charging stations will surely become the focus of the development of the automotive industry and the energy industry. Electric vehicle charging stations can better solve the problem of fast charging and save energy and reduce emissions.

[0083] like Figure 1 As shown, a charging station capability assessment method for intelligent operation of electric vehicle charging and swapping is provided, and the method comprises the following steps:

[0084] S101: receiving charging operation data of an electric vehicle charging station, processing the charging operation data of the electric vehicle charging station, and obtaining the processed charging operation data of the electric vehicle charging station, wherein the charging operation data of the electric vehicle charging station includes active power, power factor and operation timestamp;

[0085] The processing process is as follows:

[0086] Based on the current charging data of the electric vehicle charging station collected by the electric vehicle charging station data terminal, the data collected and processed are mainly in the form of active power, power factor and operation timestamp. Use part of the initial data to train the deep neural network, multiplex the initial part of the data, and use multi-layer training to increase the dimension of the neural network to improve the analysis accuracy of the neural network. The expression is:

[0087]

[0088] Where P is the matrix of all measured powers, P norm is the normalized power matrix of power.min , P max are the minimum and maximum powers in the power matrix;

[0089] S102: inputting the processed charging operation data of the electric vehicle charging station into a pre-established load decomposition model based on a deep neural network algorithm optimized by unsupervised learning, and outputting load data, wherein the pre-established load decomposition model based on a deep neural network algorithm optimized by unsupervised learning is trained based on a K-means clustering algorithm;

[0090] Design basic neural network algorithms and neural network modeling for non-intrusive load monitoring, input multi-layered data into the neural network and adjust parameters for training;

[0091] For the design of basic neural network: Artificial neuron is a basic element of neural network. MP model (McCulloch-Pitts model) is a threshold weighted sum model and a processing unit in neural network, also known as artificial neuron. i Represents all the inputs of this artificial neuron from other neurons or systems. Each input x i There is a corresponding weight w i , the weight corresponding to each input reflects the connection between the internal output of the neuron and the external input. It can be seen that the direction of neural network learning is the direction of x transmission, where x is the content transmitted from the previous neuron or system to the next neuron. Then its input matrix X can be described as:

[0092] X=[x0,x1,x2,…,x n ]

[0093] Each of the x mentioned above i The corresponding w i Weights are the key to the self-learning of neural networks. In the process of training neural networks, weights are always changing to reflect the connections between variables. Finally, after multiple trainings, a close connection between input and output is established. In fact, our training of neural networks is essentially multiple adjustments to weights. The expression of weight vector W is:

[0094] W=[w i0 ,w i1 ,w i2 ,…,w in ]

[0095] After we add the weight and the input product, the neuron also has its own threshold. When the threshold of the neuron is θ jWhen , the threshold is also often called the deviation, the connection between the input and output of our artificial neuron is shown as follows:

[0096]

[0097] And net i The net activation of a neural network neuron is the output of the weighted sum of its inputs minus the threshold, or deviation, which can be simplified using a vector as follows:

[0098] net i =XW

[0099] By applying a so-called activation function or transfer function Fare to the final output, the activation function or transfer function mentioned here actually converts the net activation obtained by our previous calculation into a number obtained by calculation, and uses our activation function to convert this weight result into a classification result. The final output y i Under the activation function expression, it can be simplified as:

[0100] y i =f(net i )=f(XW)

[0101] The above neurons can form a neural network. For a three-layer feedforward neural network, X is used as the input vector of the neural network, W is used as the connection weight vector of each layer, and F is used as the activation function. Then the expression of the output Y1, Y2, and Y3 of each layer is:

[0102] Y1=F1(XW1)

[0103] Y2=F2(F1(XW1)W2)

[0104] Y3=F3(F2(F1(XW1)W2)W3

[0105] Then, the construction of the neural network for non-intrusive equipment monitoring, because the NILM problem involving electric vehicle charging stations belongs to a high-order nonlinear function. Therefore, the linear BP neural network activation transfer function cannot meet the requirements of our non-intrusive load monitoring problem, so a suitable nonlinear function should be selected as the activation function, and the two-layer neural network with a nonlinear function as the activation function can approximate any nonlinear function. Since the BP neural network algorithm requires derivatives, we choose the differentiable activation function sigmoid. As shown in the following formula:

[0106]

[0107] For a three-layer BP neural network model, when the input and output are not equal, there is an output error E, which is expressed as:

[0108]

[0109] When the output error is transferred to the hidden layer, the error expression is:

[0110]

[0111] If it is further transferred to the input layer, it becomes the input error, which is expressed as:

[0112]

[0113] The input error is related to the weights w of each layer jk and v ij Therefore, the adjustment of the weights of each layer will affect the change of the function error result E. Obviously, the principle of weight adjustment is to gradually reduce the error so that the weight is proportional to the gradient of the error. Its expression is:

[0114]

[0115] So according to the weight adjustment formula, it is deduced layer by layer. It is determined by three factors, namely the learning rate η, the error signal δ of the output signal of this layer, and the input signal Y of this layer. The difference between the expected output and the actual output of the network can be directly expressed as the output error, which is fed back by the output layer after forward propagation;

[0116] At the same time, in the subsequent parameter training, the setting and adjustment of each parameter is also very important. We temporarily set the initial learning rate of the BP neural network to 0.01. The learning rate will affect the network training speed, etc. The initial number of hidden layers is 3 layers, and the maximum number of failures is 10 times. When this number is reached, the network will stop training. The BP neural network will use the gradient descent algorithm to adjust the error. However, since the setting of these parameters will not only affect the speed of network training, but also the accuracy of network training, the adjustment of these three parameters is very important for the training of the network model. Therefore, after each network training, we will use the fixed variable solution based on the training results, and then compare the constructed network model with the generated requirements to determine its effectiveness. If the ideal effectiveness is not achieved, use the coordinate axis gradient descent to optimize the network parameters, rebuild the network, and then input data for training, thereby repeating the above operations until the accuracy requirements are met.

[0117] However, when random vehicles are charging in the electric vehicle charging station, such as new vehicles and vehicles leaving, our neural network will not be able to monitor them. Therefore, the active and reactive power characteristics will be clustered by K-means to obtain the cluster center of the charging data of each electric vehicle. After comparing the cluster center at this time with the previous cluster center, different cluster centers will be found. These cluster centers have newly added electric vehicle data. Then, according to the characteristics of each electric vehicle charging data cluster center, we decompose the charging data of each electric vehicle collected from the intelligent terminal of the electric vehicle charging station. This decomposition scheme is to use the characteristics of the cluster center to set a suitable upper and lower threshold. The setting of this threshold requires multiple experimental discussions, and the gradient descent method is used to approximate it little by little to try to obtain a suitable range of upper and lower thresholds. Finally, according to the set threshold, the total data of the electric vehicle charging station is decomposed to obtain the data cluster of the newly added load. And update the training data of deep learning in this way, so as to train again to update the neural network model and complete the perception of the randomness of electric vehicle charging;

[0118] The K-means algorithm is a clustering algorithm that can cluster data types based on two data features. The first step of the K-means algorithm is to determine K initial centers. The expectation of each determined cluster center is the cluster center of each point. The Euclidean distance is the distance from each collected sample point to the initial cluster center. Each PQ power point is divided into each cluster center point according to the calculated Euclidean distance, and then the average distance is calculated. Repeat this process until the square error standard function meets the requirements. The square error is defined as:

[0119]

[0120] K is the number of cluster centers. K-means clustering is a clustering algorithm based on data spacing. Its working characteristic is that the closer the distance between two data points, the higher the similarity. Below is the formula for dist(m,x) to calculate the Euclidean distance between points m and x:

[0121]

[0122] When N types of vehicles are designed for clustering, how many different combinations of cluster centers can be obtained from the switching actions on M charging piles? The total number of cluster centers is:

[0123]

[0124] The number of cluster centers calculated by the above formula is used as K for clustering, and the obtained cluster centers can decompose the charging characteristics of each electric vehicle. After obtaining these cluster centers, they can be compared with the cluster center characteristics of the previous period, the changing load can be identified, and finally the training network can be updated to cope with the randomness of charging users at electric vehicle charging stations.

[0125] S103: Based on the load data, the electricity usage habits are predicted and the available capacity is evaluated, so as to realize the charging station capacity evaluation for the intelligent operation of electric vehicle charging and swapping.

[0126] Specifically, the present invention is further described below by way of embodiments:

[0127] like Figure 2-8 As shown, in order to discuss the effectiveness, reliability and operability of the scheme. We first need to use a public data set generated based on electric vehicle charging characteristics for case studies. In order to avoid excessive noise interference from the real data of electric vehicle charging stations, the huge increase in the complexity of the model caused by a large number of charging piles at electric vehicle charging stations, and the current security protection of information at domestic electric vehicle charging stations, we will generate the charging data of the electric vehicle charging stations used in our experiment based on the real electric vehicle charging characteristics. At present, it is known that the data frequency of the low-frequency data set of the target electric vehicle charging station is 1Hz. Based on the above conditions, we generated four types of electric vehicle charging characteristics, codenamed EV1, EV2, EV3, and EVn. There are certain differences in their active power and reactive power, and the differences are not large. This simulates the current small charging characteristics between electric vehicles of various brands. The apparent power and power factor of the three electric vehicles are different. At the same time, we also set the three electric vehicles to have charging power fluctuations of no more than 10 watts, which simulates the impact of power noise in electric vehicle charging stations on our modeling. At the same time, the charging habits of the three electric vehicle users are very different, and we keep their daily electricity habits and set habits that may have certain changes to simulate the random charging load in the electric vehicle charging station. The above rules set ten electric vehicle charging data, and the power time curve of the electric vehicle charging station is composed of them;

[0128] In order to simulate the randomness of electric vehicles in electric vehicle charging stations, we will add a new car data. The impact of the new car data on electric vehicle charging stations is as follows: Figure 3 shown.

[0129] In order to comprehensively evaluate the various properties of the research, such as accuracy, this chapter selects the four most important indicators in the non-intrusive load monitoring evaluation system, namely recall rate, F1 score, accuracy rate and mean absolute error, as evaluation indicators. The calculation formulas of each index are shown below.

[0130]

[0131]

[0132] The above formula is the calculation method of REC, F1 score, PRE and MAE. TP means that the actual working state of the load is consistent with the working state of the NILM analysis result and both are working data series. FP represents the data series where the actual working state of the load is inconsistent with the working state of the NILM analysis result, but in FP, the load is actually working, but the NILM analysis result shows that it is not working. FN represents the data series where the actual working state of the load is inconsistent with the working state of the NILM analysis result, but in FN, the load is actually not working, but the NILM analysis result shows that it is working. t is the actual power of the EV charging station at time t. is the decomposed power at time t, and MAE represents the mean absolute error from T0 to T1. PRE, REC, and F1 scores are basic indicators of NILM, which can be used to reflect the correctness of the analysis results judged by the NILM model and whether the load is really effective. MAE reflects the accuracy of the results across time periods. The lower the MAE result value, the higher the accuracy of its decomposition value.

[0133] When we get the experimental data, if we do not multi-layer the neural network, do not expand the depth of the neural network but directly use normalization, the pre-processed data will be used as the neural network training data to train the neural network, which will produce very poor training results;

[0134] After many experiments, we found that when the number of hidden layers is set to 16, the maximum number of failures is set to 10, and the learning rate is set to 0.01, the two requirements of speed and accuracy can be met. Figure 4 The training results of BP neural network of Matlab are shown;

[0135] The decomposition results of EV1 are basically consistent with the actual operating data after filtering, which shows that the neural network can effectively detect the user's power usage habits and the charging characteristics of the vehicle. Figure 5As shown in the figure, it can be clearly seen that although there are still some fluctuations in the decomposition of the power data of the electric vehicle charging station by the deep neural network, its waveform can already fit our actual power curve. We used the data from the first four days for training and the data from the last four days for testing, proving that the neural network deepened by this scheme can better handle the load perception problem of electric vehicle charging stations.

[0136] At the same time, we analyze the decomposed data using the various indicators in the non-intrusive load monitoring algorithm evaluation system mentioned earlier and calculate the accuracy, recall rate and F1 score of each load data. The analysis results and calculation indicators are shown in the following table.

[0137] Table 1-1 Neural network decomposition results of electric vehicle charging stations without the addition of EV11

[0138]

[0139] The calculated TP indicates that the actual working state of the load is consistent with the working state of the NILM analysis result and both are working data series. FP represents a data series in which the actual working state of the load is inconsistent with the working state of the NILM analysis result. However, in FP, the load is actually working, but the NILM analysis result shows that it is not working. FN represents a data series in which the actual working state of the load is inconsistent with the working state of the NILM analysis result. However, in FN, the load is actually not working, but the NILM analysis result shows that it is working. t is the actual power of the electric vehicle charging station at time t. is the decomposed power at time t, and MAE represents the mean absolute error from T0 to T1. Therefore, the results show that the accuracy of the scheme is still very good and the error is controlled within an acceptable level.

[0140] The purpose of using the training set of the deep learning neural network to continuously update the characteristics of the newly added load is to enable the neural network to update its parameters to complete the perception of the newly added electric vehicle load. So we added an electric vehicle to the 10 electric vehicles generated previously and added it to the bus;

[0141] After adding the load EV11, in addition to the fact that the current model cannot perceive EV11, our perception of other vehicles will also have more or less deviations. For example, in the perception of EV6, due to the influence of EV11, the fluctuations on the decomposed power curve will be more, which may affect the subsequent identification difficulties, such as Figure 6 As shown;

[0142] When we add EV11 data to our current electric vehicle charging station model, it will obviously have a great impact on the analytical ability of the deep neural network, so we will choose to continuously update the training set and use it to update the parameters in the deep neural network. First, we will enter the bus data of the electric vehicle charging station with the newly added load into the algorithm to calculate the initial K value in the K-means clustering algorithm. Since the active power of each charging pile in the electric vehicle charging station is consistent, we will get the size of K by comparing the active power with the reactive power. The basic steps are to use the algorithm to distinguish the data with different active powers and then use the difference in reactive power to roughly divide each area, and calculate the value of K by the number of areas. Then, based on the obtained K value, K-means clustering is performed on the electric vehicle charging station without the newly added EV11;

[0143] After obtaining the above cluster centers, we will also perform K calculation and K-means clustering on the newly added electric vehicle charging station data. The clustering results are as follows: Figure 7 As shown;

[0144] By comparing the cluster center characteristics of the newly added load and the non-added load, it is obvious that a new cluster center appears in the current electric vehicle charging station bus data after the new load is added. The appearance of this new cluster center is because the newly added EV11 load operation data overlaps with other loads during this time, which is a new feature. Therefore, we use time comparison and power data difference to generate EV11 operation data. The overlapping time is the operation time of the newly added load, and the operation power characteristics of the newly added load can be obtained by comparing with the previous cluster center.

[0145] After using the clustering and decomposition algorithm to obtain the EV11 running curve, we used it together with the original running data as training data, updated the deep neural network training set, and retrained the neural network. The number of hidden layers was set to 16, the maximum number of failures was 10, and the initial learning rate was set to 0.01. The training results are shown in Figure 2. Figure 8 As shown;

[0146] After that, we used the neural network with updated data to analyze the newly added load EV11 and decomposed the operating power-time curve of EV11;

[0147] It can be seen that although the accuracy of the neural network with updated parameters in the exploration of new loads is not that high, it can still clearly distinguish the running time of EV11 each time, so this is enough for subsequent identification using the identification algorithm. This solution is still very effective in dealing with the user randomness problem of electric vehicle charging stations. The following table shows the indicators obtained by running the non-intrusive load monitoring algorithm for electric vehicle charging stations after adding EV11;

[0148] Table 1-2 Results of deep neural network analysis of electric vehicle charging stations after adding EV11

[0149]

[0150] After adding EV11, the electric vehicle charging station runs the non-invasive load monitoring algorithm to obtain the indicators, and the electric vehicle deep neural network analysis results after adding EV11. We can clearly see the advantages of the non-invasive load monitoring algorithm proposed in this chapter over the traditional neural network algorithm in analyzing the non-invasive load monitoring problem. The comparison of the parameters finally calculated by the three can be obtained from the figure. It can be seen that the deep neural network is the most accurate in perceiving regular loads. Although the deep neural network optimized by unsupervised learning is not so accurate in perceiving random loads, it solves the perception ability of the non-invasive load monitoring algorithm for random loads, and the error is also within an acceptable range. For example, the comparison of the load perception results of electric vehicle charging stations is Fig. 9 shown.

[0151] Embodiment 2: In the second aspect, as Fig.10 As shown, in order to achieve the above-mentioned purpose, the present invention discloses a charging station capability evaluation system for intelligent operation of electric vehicle charging and swapping, comprising:

[0152] The data processing module 11 is used to receive the charging operation data of the electric vehicle charging station, process the charging operation data of the electric vehicle charging station, and obtain the processed charging operation data of the electric vehicle charging station, wherein the charging operation data of the electric vehicle charging station includes active power, power factor and operation timestamp;

[0153] A load decomposition module 12, used for inputting the processed charging operation data of the electric vehicle charging station into a pre-established load decomposition model based on a deep neural network algorithm optimized by unsupervised learning, and outputting load data, wherein the pre-established load decomposition model based on a deep neural network algorithm optimized by unsupervised learning is trained based on a K-means clustering algorithm;

[0154] The capacity evaluation module 13 is used to predict the power usage habits and evaluate the available capacity based on the load data, so as to realize the charging station capacity evaluation for the intelligent operation of electric vehicle charging and swapping.

[0155] Based on the same inventive concept, the present invention also provides a computer device, which includes: one or more processors, and a memory for storing one or more computer programs; the program includes program instructions, and the processor is used to execute the program instructions stored in the memory. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is used to implement one or more instructions, specifically for loading and executing one or more instructions in a computer storage medium to implement the above method.

[0156] It needs to be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium, on which a computer program is stored, and the computer program is executed by a processor to execute the above method. The storage medium can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electrical, magnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0157] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0158] The above shows and describes the basic principles, main features and advantages of the present disclosure. Those skilled in the art should understand that the present disclosure is not limited by the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present disclosure. Without departing from the spirit and scope of the present disclosure, the present disclosure may have various changes and improvements, and these changes and improvements fall within the scope of the present disclosure to be protected.

Claims

1. A method for evaluating the capacity of charging stations for intelligent operation of electric vehicle charging and swapping, characterized in that: The method comprises the following steps: Receiving charging operation data of an electric vehicle charging station, processing the charging operation data of the electric vehicle charging station, and obtaining processed charging operation data of the electric vehicle charging station, wherein the charging operation data of the electric vehicle charging station includes active power, power factor, and operation timestamp; Inputting the processed charging operation data of the electric vehicle charging station into a pre-established load decomposition model based on a deep neural network algorithm optimized by unsupervised learning, and outputting load data, wherein the pre-established load decomposition model based on a deep neural network algorithm optimized by unsupervised learning is trained based on a K-means clustering algorithm; Based on load data, electricity usage habits are predicted and available capacity is evaluated, thereby realizing charging station capacity assessment for intelligent operation of electric vehicle charging and swapping.

2. The method for evaluating the capability of charging stations for intelligent operation of electric vehicle charging and swapping according to claim 1 is characterized in that: When evaluating the controllable capability of the electric vehicle, the electric vehicle needs to enter a controllable idle state, wherein the determination process of whether to enter the idle state is as follows: The remaining power of electric vehicles entering the power grid is judged by the average daily travel distance, and the travel demand of the owner on the next day is used to determine how long it takes for the electric vehicle to enter the idle state after entering the power grid. The user's average daily travel distance obeys the log-normal distribution, and the probability density function is: Where d represents the average daily driving distance, μ d and σ d It represents the expectation and variance of the log-normal distribution, and sets the predicted remaining power of the vehicle when it enters the power grid and the conditions when the vehicle enters the idle state in the current electric vehicle charging station system according to probability.

3. The method for evaluating the capability of charging stations for intelligent operation of electric vehicle charging and swapping according to claim 2 is characterized in that: Calculate the estimated charging time required for the electric vehicle to enter the idle adjustable state: Among them, β refers to the amount of electricity required to travel one kilometer, d is the probability density of the average daily travel distance of electric vehicle users, and P is the minimum daily charging amount of this type of electric vehicle.

4. The method for evaluating the capability of charging stations for intelligent operation of electric vehicle charging and swapping according to claim 1 is characterized in that: The construction of the unsupervised learning optimized deep neural network algorithm includes the construction of a basic neural network algorithm and the construction of a neural network based on non-invasive equipment monitoring NILM.

5. The method for evaluating the capability of charging stations for intelligent operation of electric vehicle charging and swapping according to claim 4 is characterized in that: The construction process of the basic neural network algorithm is as follows: Artificial Neurons: x1~x i Represents all the inputs of this artificial neuron from other neurons or systems. Each input x i There is a corresponding weight w i , the direction of neural network learning is the direction of x transmission, x is the content transmitted from the previous neuron or system to the next neuron, then the input matrix X is described as: X=[x0,x1,x2,…,x n ] The expression of the weight vector W is: In=[in i0 ,In i1 ,In i2 ,…,In in ] The connection between the input and output of an artificial neuron is shown below: And net i As the net activation of the neural network neurons, it can be simplified using vectors as follows: net i =XW The final output y i The activation function expression is simplified as follows: y i =f(net i )=f(XW) For a three-layer feedforward neural network, X is used as the input vector of the neural network, W is used as the connection weight vector of each layer, and F is used as the activation function. Then the expression of the output Y1, Y2, and Y3 of each layer is: Y1=F1(XW1) Y2=F2(F1(XW1)W2) Y3=F3(F2(F1(XW1)W2)W3.

6. The method for evaluating the capability of charging stations for intelligent operation of electric vehicle charging and swapping according to claim 5 is characterized in that: The neural network construction process based on non-invasive device monitoring NILM is as follows: Select the differentiable activation function sigmoid as the activation function, as shown in the following formula: For the three-layer BP neural network model, when the input and output are not equal, there is an output error E, which is expressed as: When the output error is transferred to the hidden layer, the error expression is: If it is further transferred to the input layer, it becomes the input error, which is expressed as: The input error is related to the weights w of each layer jk and v ij Therefore, the adjustment of the weights of each layer will affect the change of the function error result E. The principle of weight adjustment is to reduce the error so that the weight is proportional to the gradient decrease of the error. The expression is: It is determined by three factors: the learning rate η, the error signal δ of the output signal of this layer, and the input signal Y of this layer. The difference between the expected output and the actual output of the network is directly expressed as the output error, which is fed back by the output layer after forward propagation.

7. The method for evaluating the capability of charging stations for intelligent operation of electric vehicle charging and swapping according to claim 1 is characterized in that: The process of processing the charging operation data of the electric vehicle charging station is as follows: A part of the initial data of the charging operation data of the electric vehicle charging station is used to train the deep neural network. Part of the initial data is multiplexed, and the dimension of the neural network is increased by multi-layer training. The expression is: Where P is the matrix of all measured powers, P norm is the normalized power matrix of power, P min , P max are the minimum and maximum powers in the power matrix.

8. The method for evaluating the capability of charging stations for intelligent operation of electric vehicle charging and swapping according to claim 1 is characterized in that: When training is performed based on the K-means clustering algorithm, the process of the K-means clustering algorithm is as follows: The first step of the K-means algorithm is to determine K initial centers. The expectation of each determined cluster center is the cluster center of each point. The Euclidean distance is the distance from each collected sample point to the initial cluster center. Each PQ power point is divided into each cluster center point according to the calculated Euclidean distance, and then the average distance is calculated. Repeat this process until the square error standard function meets the requirements. The square error is defined as: K is the number of cluster centers, and below is the formula for dist(m,x) to calculate the Euclidean distance between points m and x: When N types of vehicles are designed for clustering, multiple different combinations of cluster centers are obtained from the switching actions on M charging piles. The total number of cluster centers is: The number of cluster centers calculated by the above formula is used as K for clustering, and the obtained cluster centers can decompose the charging characteristics of each electric vehicle.

9. The charging station capacity evaluation system for electric vehicle charging and swapping intelligent operation is characterized by: include: A data processing module, used for receiving charging operation data of the electric vehicle charging station, processing the charging operation data of the electric vehicle charging station, and obtaining the processed charging operation data of the electric vehicle charging station, wherein the charging operation data of the electric vehicle charging station includes active power, power factor and operation timestamp; A load decomposition module, used for inputting the processed charging operation data of the electric vehicle charging station into a pre-established load decomposition model based on a deep neural network algorithm optimized by unsupervised learning, and outputting load data, wherein the pre-established load decomposition model based on a deep neural network algorithm optimized by unsupervised learning is trained based on a K-means clustering algorithm; The capacity assessment module is used to predict electricity usage habits and evaluate available capacity based on load data, thereby realizing charging station capacity assessment for intelligent operation of electric vehicle charging and swapping.

10. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: The memory stores a computer program that can be run on the processor. When the processor loads and executes the computer program, the charging station capacity assessment method for intelligent operation of electric vehicle charging and swapping described in any one of claims 1 to 8 is adopted.

Images