Charging station design method, charging station design system and charging station

Abstract

The present application discloses a charging station design method, a charging station design system and a charging station, which belong to the technical field of new energy vehicle charging facilities and comprehensive utilization of renewable energy, including obtaining real-time meteorological data and load data. The meteorological data and load data are preprocessed to obtain preprocessed data; and the preprocessed data is feature extracted to obtain prediction features. A prediction model for energy generation power and load demand is established, and based on the prediction features, the established prediction model is used to obtain the predicted wind power generation power, solar power generation power and load demand power at future times. An energy scheduling strategy is selected based on the predicted power results of the prediction model, meteorological sensor data and preset control rules. The selected energy scheduling strategy is executed to adjust the working mode of the charging station. The present application improves the energy utilization efficiency, economy and stability of the charging station, while reducing operating costs and environmental impact.

Description

Technical Field

[0001] The present application relates to the field of new energy vehicle charging facilities and comprehensive utilization of renewable energy, and in particular to a charging station design method, a charging station design system, and a charging station. Background Art

[0002] The rapid development of new energy vehicles is driving an increasing demand for charging infrastructure, particularly for fast and efficient charging technology. However, traditional charging stations present a range of challenges: most rely on mains electricity and fail to effectively integrate renewable energy, resulting in energy waste and high operating costs. Existing charging equipment suffers from low charging efficiency, particularly when high-power charging is required, leading to significant energy loss during the charging process. Furthermore, existing charging stations lack flexibility, making it difficult to adjust in real time to grid load fluctuations.

[0003] A small number of new energy charging stations combine wind power, solar power, and energy storage devices to effectively utilize renewable energy. However, these new energy charging stations still face some challenges in practical application. Although the combination of wind power, solar power, and energy storage devices can effectively utilize renewable energy, due to the instability and intermittent nature of wind and solar power, charging stations relying on these energy sources may experience unstable power supply, resulting in power outages or instability during the charging process, affecting the user's charging experience and charging efficiency. Some existing new energy charging stations still rely on traditional power networks as a backup power source. In the event of a grid failure or fluctuations in power demand, the stability and availability of the charging stations are restricted. Summary of the Invention

[0004] The present application aims to provide a charging station design method, a charging station design system, and a charging station that solve the problem of restricted stability and availability of charging stations.

[0005] To achieve the above objectives, the technical solution of this application is:

[0006] A charging station design method, comprising:

[0007] Step S1: Obtain real-time meteorological data and load data;

[0008] Step S2: preprocessing the meteorological data and the load data to obtain preprocessed data; performing feature extraction on the preprocessed data to obtain prediction features;

[0009] Step S3: Establishing a prediction model for energy generation power and load demand, and using the established prediction model to obtain the initial wind power generation prediction power, initial solar power generation prediction power, and initial load demand prediction power at a future time based on the prediction characteristics;

[0010] Step S4: selecting an energy scheduling strategy based on the predicted power results of the prediction model, meteorological sensor data, and preset control rules;

[0011] Step S5: Execute the selected energy scheduling strategy and adjust the working mode of the charging station.

[0012] Optionally, in step S1, the meteorological data includes: meteorological sensor data and meteorological forecast data, wherein the meteorological sensor data includes: wind speed, solar irradiance, ambient temperature, relative humidity; the load data includes: load demand, DC bus voltage, and energy storage battery state of charge.

[0013] Optionally, in step S2, the preprocessing includes: maximum normalization, sliding mean filtering, and data standardization;

[0014] The feature extraction includes: extracting valuable features from the preprocessed data, including time series features: diurnal variation trends of wind speed and solar irradiance; extracting weather variation features based on meteorological data in the preprocessed data; generating load prediction features based on historical load trends of load data in the preprocessed data; and constructing periodic features;

[0015] After the feature extraction, the method further includes: calculating the weights of the extracted features to obtain the prediction features, where the prediction features include: normalized wind speed, normalized solar irradiance, standardized temperature, hourly cycle, first-order difference of load power, and load power.

[0016] Optionally, in step S3, a long short-term memory network model is used to establish a prediction model for energy generation capacity and load demand, the input of the prediction model is the prediction feature, and the output of the prediction model is the initial wind power generation prediction power, initial solar power generation prediction power, and initial load demand prediction power at a future time.

[0017] Optionally, after the prediction model outputs the predicted power result, it also includes: correcting the initial wind power generation predicted power and the initial solar power generation predicted power to obtain corrected wind power generation predicted power and corrected solar power generation predicted power.

[0018] Optionally, in step S4, the preset control rules include: steps S41 to S413;

[0019] Step S41: Energy dispatch control starts;

[0020] Step S42: Determine whether the charging station status is normal. If so, proceed to step S43; if not, proceed to step S410;

[0021] Step S43: Determine whether the revised wind power generation forecast power and the revised solar power generation forecast power are greater than the load demand forecast power. If so, proceed to step S44; if not, proceed to step S46.

[0022] Step S44: selecting strategy 1: wind and solar priority mode;

[0023] Step S45: adjusting the output according to the revised wind power generation prediction power and the revised solar power generation prediction power; proceeding to step S411;

[0024] Step S46: Determine the state of charge of the energy storage battery. If the state of charge is greater than 70%, proceed to step S47; if the state of charge is greater than 30% and less than 70%, proceed to step S48; if the state of charge is less than 30%, proceed to step S49;

[0025] Step S47: Select strategy 2: wind-solar-storage collaborative mode, enable constant voltage-frequency ratio control, and proceed to step S411;

[0026] Step S48: Select strategy 3: storage-grid collaboration mode 1, enable droop control, and jointly output power from the storage and grid, and proceed to step S411;

[0027] Step S49: Select strategy 4: storage-grid collaboration mode 2, energy storage is charged from the grid, and proceed to step S411;

[0028] Step S410: Select strategy 5: island operation mode, enable virtual synchronous dual-winding generator control, and proceed to step S411;

[0029] Step S411: executing the policy;

[0030] Step S412: Evaluate the energy dispatch control status, including: the state of charge of the energy storage battery, the error of the predicted power result, and the quality of the grid voltage. If the actual requirements of the charging station are met, proceed to step S413; if not, return to step S41;

[0031] Step S413: Energy scheduling control ends.

[0032] Optionally, in step S5, after selecting the energy scheduling strategy, the step further includes: continuously optimizing the scheduling strategy through reinforcement learning.

[0033] A charging station design system, which executes a charging station design method as described above, comprising: an environment perception module, a data preprocessing and feature extraction module, a deep learning and prediction module, an adaptive control module, and an execution and scheduling module;

[0034] The environmental perception module, the data preprocessing and feature extraction module, the deep learning and prediction module, the adaptive control module and the execution and scheduling module are connected in sequence.

[0035] A charging station, performing a charging station design method as described in any one of the above, comprising:

[0036] Wind, solar and storage systems;

[0037] A dual-winding generator, wherein a first end of the dual-winding generator is connected to a first end of the wind-solar-storage system;

[0038] a first ACDC converter, wherein a first end of the first ACDC converter is connected to a first end of the dual-winding generator and a first end of the wind-solar-storage system respectively;

[0039] a second ACDC converter, wherein a first end of the second ACDC converter is connected to a second end of the dual-winding generator;

[0040] A 400V charging pile, wherein a first end of the 400V charging pile is connected to a second end of the first ACDC converter;

[0041] An 800V charging pile, wherein a first end of the 800V charging pile is connected to a second end of the second ACDC converter;

[0042] A mains interface is reserved, and a first end of the reserved mains interface is respectively connected to a first end of the wind-solar-storage system, a first end of the dual-winding generator, and a first end of the ACDC converter.

[0043] Optionally, the wind-solar-storage system includes:

[0044] Energy storage batteries;

[0045] a wind turbine generator, wherein a first end of the wind turbine generator is connected to a first end of the energy storage battery;

[0046] a solar panel, wherein a first end of the solar panel is connected to a first end of the energy storage battery and a first end of the wind turbine respectively;

[0047] A bidirectional DCAC converter, wherein the first end of the bidirectional DCAC converter is respectively connected to the first end of the energy storage battery, the first end of the wind turbine, and the first end of the solar panel; the second end of the bidirectional DCAC converter is the first end of the wind-solar-storage system.

[0048] This application provides a charging station design method, charging station design system, and charging station. Based on a long short-term memory (LSTM) network model, a prediction model for energy generation power and load demand is established. This model can more accurately predict renewable energy generation and charging demand, thereby achieving more efficient energy distribution and management and improving charging station operational efficiency. A dual-winding generator design can output different voltage levels to meet the charging needs of various vehicles. This eliminates the need for a DC-DC converter, simplifies the circuit structure, reduces losses during energy conversion, and improves charging efficiency. The dual-winding generator's multi-voltage output capability enables the charging station to accommodate the charging needs of various types of electric vehicles, enhancing its applicability and flexibility. By integrating a wind, solar, and storage system, the system fully utilizes renewable energy sources such as solar and wind power, reducing reliance on traditional utility power, lowering energy costs, and improving the charging station's energy sustainability. The charging station design system and energy allocation strategy enable real-time monitoring and control of various components within the charging station, ensuring efficient and flexible energy distribution and utilization, and improving the stability and cost-effectiveness of the charging station's operations.

[0049] In order to make the above features and advantages of the application more obvious and easy to understand, the following embodiments are given and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Flowchart of the charging station design method of the present application.

[0051] Figure 2 This is a flow chart of the preset control rules of this application.

[0052] Figure 3 This is a module diagram of the charging station design system for this application.

[0053] Figure 4 This is a circuit diagram of the charging station of this application. DETAILED DESCRIPTION

[0054] To make the purpose and technical solutions of the embodiments of the present application clearer, the technical solutions of the embodiments of the present application will be clearly and completely described below in conjunction with the drawings of the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the described embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0055] In one embodiment of this application, please refer to Figure 1 , Figure 1 This is a flow chart of the charging station design method of the present application. The present application provides a charging station design method, including: steps S1 to S5.

[0056] Step S1: Acquire real-time meteorological data and load data.

[0057] Step S2: Preprocessing the meteorological data and the load data to obtain preprocessed data; performing feature extraction on the preprocessed data to obtain prediction features.

[0058] Step S3: Establish a prediction model for energy generation power and load demand. Based on the prediction characteristics, use the established prediction model to obtain the initial wind power generation prediction power, initial solar power generation prediction power and initial load demand prediction power at the future moment.

[0059] Step S4: Select an energy scheduling strategy based on the predicted power results of the prediction model, meteorological sensor data, and preset control rules.

[0060] Step S5: Execute the selected energy scheduling strategy and adjust the working mode of the charging station.

[0061] In step S1, see Figure 1 In step S1, real-time weather data and load data are obtained.

[0062] As an example, the meteorological data includes meteorological sensor data and meteorological forecast data, wherein the meteorological sensor data includes wind speed, solar irradiance, ambient temperature, and relative humidity; the load data includes load demand, DC bus voltage, and energy storage battery state of charge. The meteorological sensor data, meteorological forecast data, and load data are numerically represented as follows:

[0063] Formula (1)

[0064] in, X meteo Represents weather sensor data; v w Indicates wind speed ( ); G solar represents the solar irradiance ( ); T amb Indicates temperature ( ); RH Relative humidity ( ); X load Represents load data; P demand Indicates the charging power requirement ( ); V bus Indicates the DC bus voltage ( ); SOC batIndicates the state of charge of the energy storage battery ( ); X pred Represents weather forecast data; NWP represents the output of the numerical weather forecast model (spatial resolution 1km×1km), lat Indicates latitude, lon Indicates longitude, t represents time; T represents the transposed symbol.

[0065] In step S2, see Figure 1 In step S2, the meteorological data and the load data are preprocessed to obtain preprocessed data; and feature extraction is performed on the preprocessed data to obtain prediction features.

[0066] As an example, the preprocessing includes: maximum value normalization, sliding mean filtering, and data standardization. After the preprocessing, preprocessed data of the load data and the meteorological data are obtained.

[0067] Furthermore, the maximum normalization includes: normalizing the wind speed and solar irradiance to their maximum values. The maximum normalization of wind speed is expressed as:

[0068] Formula (2)

[0069] in, represents the normalized wind speed, v w,min Indicates the minimum value of the wind speed data obtained, v w,max Indicates the maximum value of the obtained wind speed data. By linearly transforming the obtained wind speed data, the resulting wind speed value is mapped to the range [0-1] to obtain the normalized wind speed.

[0070] The maximum normalization formula for solar irradiance is:

[0071] Formula (3)

[0072] in, represents the normalized solar irradiance, G STC Standard test conditions refer to the parameters used to test solar panels under standard laboratory conditions: irradiance: 1000W / m²; cell temperature: 25°C; spectral distribution: AM1.5. The actual output efficiency of a photovoltaic system is evaluated by comparing real-time solar irradiance with standard test conditions.

[0073] Furthermore, the sliding mean filter includes smoothing and differential processing of time series data such as temperature and load data. Set the window size N=60, and the input data includes time series data such as temperature and load data, which can be expressed as:

[0074] Formula (4)

[0075] in, y t represents the filtered time series data, x k Indicates the first k Input data, x t Indicates the current time t Input data, 3σ represents three times the standard deviation (the threshold for determining outliers), sgn represents the sign function, μ window Represents the mean of the input data within the window. High-frequency noise in the input data is suppressed by a sliding mean filter. When the input data at the current moment deviates from the mean by more than 3σ, the sign function triggers the Kalman filter to correct the input data.

[0076] Furthermore, data standardization includes: standardizing the load data and meteorological data to make their mean 0 and standard deviation 1, which can be expressed as:

[0077] Formula (5)

[0078] Among them, Δ x t represents the first-order difference of the time series at time t, x t-1 Indicates time t-1 The input data is used to eliminate trend items through first-order differences, eliminate trend items in load data and meteorological data (such as the inherent changes in temperature rising during the day and falling at night), highlight short-term fluctuation characteristics (such as sudden changes in wind speed), make time series data more stable, facilitate subsequent model learning, and improve model training effects.

[0079] Furthermore, feature extraction is performed on the preprocessed data to obtain prediction features. This feature extraction includes: extracting valuable features from the preprocessed data, including time series features such as diurnal trends in wind speed and solar irradiance; extracting weather change features based on meteorological data in the preprocessed data to provide more efficient input for subsequent deep learning models; and generating load prediction features based on historical load trends in the load data in the preprocessed data.

[0080] Furthermore, the periodic characteristics are constructed and expressed as follows:

[0081] Formula (6)

[0082] in, C day represents the hour period, C year represents the daily cycle, t Indicates the hours of the day (0≤t<24 hours), t =12 o'clock corresponds to noon, t =0 corresponds to midnight, d Indicates the cumulative days per year (1≤ d <365 days). The intra-day periodicity of solar irradiance and load demand is characterized by periodic characteristics.

[0083] Furthermore, the weights of the extracted features are calculated and the random forest is used to calculate the feature weights, which can be expressed as follows:

[0084] Formula (7)

[0085] in, W i represents the weight of the i-th feature, N tree Indicates the number of decision trees in the random forest, Imp i (t) represents the weight of the i-th feature in the t-th decision tree, M represents the total number of features in the dataset, Imp j (t) Represents the weight of the jth feature in the tth decision tree. By calculating and averaging the weights of each feature in each decision tree, the importance of each feature in the entire random forest model is comprehensively evaluated.

[0086] Furthermore, according to the weight calculation results of the extracted features, the top 6 features with the highest weights are selected as the prediction features of the model input, including normalized wind speed, normalized solar irradiance, standardized temperature, hourly period, first-order difference of load power, and load power.

[0087] Step S3: Establish a prediction model for energy generation capacity and load demand. Based on the prediction characteristics, use the established prediction model to obtain the initial wind power generation prediction power, initial solar power generation prediction power and initial load demand prediction power at the future moment.

[0088] As an example, a long short-term memory network model (LSTM) is used to establish a prediction model for energy generation capacity and load demand. The prediction model LSTM is trained using historical data of normalized wind speed, normalized solar irradiance, standardized temperature, hourly period, first-order difference of load power, load power, wind power generation power, solar power generation power, and load demand power to ensure that the prediction model LSTM can accurately predict the initial wind power generation forecast power, initial solar power generation forecast power, and initial load demand forecast power at future times. The prediction model LSTM input is normalized wind speed, normalized solar irradiance, standardized temperature, hourly period, first-order difference of load power, and load power, and the output is the initial wind power generation forecast power at future times. (t+1), initial solar power generation prediction power (t+1), initial load demand forecast power (t+1). It can be expressed as:

[0089] Formula (8)

[0090] in, x t Represents the input vector at the current moment, used to carry input features; h t-1 Represents the hidden state of the previous moment, used as a short-term memory carrier: conveys the network state information of the previous moment; C t-1 Represents the cell state at the previous moment and is used to store important feature information across time steps; W f represents the forget gate of the weight matrix, W i represents the input gate of the weight matrix, W c represents the memory unit of the weight matrix, W o Represents the output gate of the weight matrix, which is used for the feature transformation matrix: mapping the input and hidden states to the gating signal; b f 、 b i 、 b c 、 b o Both represent bias terms, which are used to control the offset of the gate: adjusting the threshold of the gate activation function; f t Represents the output of the forget gate, which is used to decide how much historical information to retain: controlling the cell state at the previous moment C t-1 The proportion of forgetting; Represents a candidate memory unit, used for new information candidate value: contains the feature information extracted at the current moment; tanh represents the hyperbolic tangent function; i t Represents the output of the input gate, which is used to decide how much new information to update: controlling the candidate memory unit Write ratio; C t Represents the updated cell state, which is used for long-term memory update: fusing historical memory and the current input vector; h t Represents the current hidden state, used for short-term memory output: the hidden state passed to the next time step; o t Represents the output of the output gate, which is used to decide what information to output: controlling the updated cell state C t To the current hidden state h t conversion.

[0091] Specifically, the input vector x t Expressed as:

[0092] Formula (9)

[0093] in, P load (t) represents the load power, Δ P load (t) represents the first-order difference of load power. Normalized wind speed Impact on initial wind power generation forecast (t+1); normalized solar irradiance With standardized temperature Impact on initial solar power generation forecast power (t+1); hour cycle C day , first-order difference of load power Δ P load (t) and load power P load (t) Impact on initial load demand forecast power (t+1).

[0094] Furthermore, the output gate controls the cell state To hidden state Information flow, hidden state In order to pass the intermediate features to the next time step, the temporal dependency of the historical input information is encoded. The final prediction output of the prediction model LSTM requires an additional fully connected layer to process the hidden state, which can be expressed as:

[0095] Formula (10)

[0096] in, W wind is the weight matrix of the wind power generation prediction power output layer, W solar The weight matrix of the power output layer for solar power prediction, W load The weight matrix of the power output layer for load demand prediction, b wind is the bias term for the wind power generation prediction power output layer, b solar The bias term for the power output layer to predict solar power generation, b load Bias term for the power output layer to predict load demand.

[0097] As an example, the network parameters in the prediction model LSTM are 3 layers of LSTM layers. LSTM layers are used to capture temporal dependencies in time series. Increasing the number of network layers allows the model to learn more complex and abstract feature representations. The number of units in each LSTM layer is 128, 64, and 32, respectively. The number of units determines the dimension of the hidden state of each LSTM layer, indicating the number of features that each LSTM layer can learn. In the established prediction model LSTM, in order to prevent the prediction model LSTM from overfitting, a regularization method is used. Dropout Randomly ignore the outputs of some neurons with a certain probability.

[0098] In one embodiment, Dropout= 0.2.

[0099] As an example, based on the established prediction model LSTM, the prediction features are input: normalized wind speed, normalized solar irradiance, standardized temperature, hourly cycle, first-order difference of load power, and load power, to obtain the initial wind power generation prediction power, initial solar power generation prediction power, and initial load demand prediction power at the future moment.

[0100] Furthermore, the mean square error of the three prediction tasks of wind power generation, solar power generation and load demand is optimized, which can be expressed as follows:

[0101] Formula (11)

[0102] in, It represents the mean square error of the three prediction tasks of simultaneously optimizing wind power generation, solar power generation and load demand; P windrepresents the actual wind power generation power, P solar Indicates the actual solar power generation power, P load Indicates the actual load demand power, represents the initial wind power generation forecast power; represents the initial solar power generation forecast power; Represents the initial load demand forecast power. The initial wind power forecast, solar power forecast, and load demand forecast power after the LSTM prediction model's initial prediction must approximate the actual wind power, solar power, and load demand power. ɑ, β, and γ represent weight coefficients used to adjust the importance of each prediction task.

[0103] In one embodiment, ɑ=0.4, β=0.3, and γ=0.3.

[0104] Furthermore, due to the significant characteristics of wind speed probability distribution, the Bayesian method is used to correct the wind power prediction results for wind power generation, which can be expressed as follows:

[0105] Formula (12)

[0106] in, represents the revised wind power generation forecast power, represents the historical actual wind speed dataset, Indicates the predicted wind speed value at the current moment, Represents the marginal probability distribution of historical wind speeds.

[0107] Furthermore, for the solar power generation power prediction task, the Bayesian method is used to correct the solar power generation power prediction result, which can be expressed as follows:

[0108] Formula (13)

[0109] in, Indicates the revised solar power generation forecast power, represents the historical irradiance distribution, represents the predicted irradiance value, Represents the marginal probability distribution of historical solar irradiance.

[0110] As an example, the power prediction results of the prediction model LSTM include: corrected wind power generation prediction power, corrected solar power generation prediction power, and load demand prediction power.

[0111] Step S4: Select an energy scheduling strategy based on the predicted power results of the prediction model LSTM, meteorological sensor data, and preset control rules.

[0112] As an example, see Figure 2 , Figure 2 This is a flow chart of a preset control rule, which includes steps S41 to S413.

[0113] Step S41: Energy dispatch control starts.

[0114] Step S42: Determine whether the charging station status is normal. If so, proceed to step S43; if not, proceed to step S410.

[0115] Step S43: Determine whether the revised wind power generation forecast power and the revised solar power generation forecast power are greater than the load demand forecast power. If so, proceed to step S44; if not, proceed to step S46.

[0116] Step S44: Select strategy 1: wind and solar priority mode.

[0117] Step S45: Adjust the output according to the revised wind power generation forecast power and the revised solar power generation forecast power; proceed to step S411.

[0118] Step S46: Determine the state of charge of the energy storage battery. If it is greater than 70%, proceed to step S47; if it is greater than 30% and less than 70%, proceed to step S48; if it is less than 30%, proceed to step S49.

[0119] Step S47: Select strategy 2: wind-solar-storage collaborative mode, enable constant voltage-frequency ratio (V / F) control, and proceed to step S411.

[0120] Step S48: Select strategy 3: storage-grid collaboration mode 1, enable droop control, and jointly output power from the storage and grid, and proceed to step S411.

[0121] Step S49: Select strategy 4: storage-grid collaboration mode 2, energy storage is charged from the grid, and proceed to step S411.

[0122] Step S410: Select strategy 5: island operation mode, enable virtual synchronous dual-winding generator (vsg) control, and proceed to step S411.

[0123] Step S411: Execute the policy.

[0124] Step S412: Evaluate the energy dispatch control status, including: the state of charge of the energy storage battery, the error of the predicted power result, and the quality of the grid voltage. If it meets the actual needs of the charging station, proceed to step S413; if not, return to step S41.

[0125] Step S413: Energy scheduling control ends.

[0126] As an example, the strategy 1: wind-solar priority mode includes: when wind energy and solar energy are sufficient, the electricity generated by them is used to charge vehicles first.

[0127] Strategy 2, a wind-solar-storage synergy, involves using energy storage batteries to supplement power when wind and solar power fluctuate or are temporarily insufficient. If wind and solar power generation exceeds the required power for vehicle charging, the excess energy is stored in the energy storage system. If wind and solar power generation is less than the required power for vehicle charging, the shortfall is covered by the energy storage system.

[0128] The strategy 3: storage-grid synergy mode 1 includes: when the combined power supply of the wind-solar-storage synergy mode still cannot meet the vehicle charging needs, obtaining electricity from the power grid.

[0129] The strategy 4: storage-grid collaborative mode 2 includes: during the off-peak period of electricity consumption or when the energy storage battery charge is low, low-cost electricity can be obtained from the power grid and stored in the energy storage battery, and then released for use during the peak period of electricity consumption.

[0130] Strategy 5: Island operation mode, including: when the energy storage battery and the power grid are unable to supply power, starting the generator to output voltage levels of 400V and 800V respectively for vehicle use.

[0131] Furthermore, when the environment changes, the prediction model LSTM will continuously optimize the scheduling strategy through reinforcement learning. For example, when the wind speed fluctuates greatly, the weight of wind power generation will be increased; when it is cloudy, the weight of battery discharge power will be increased.

[0132] As an example, the reinforcement learning includes: Q-learning update rules and rolling horizon control (RHC).

[0133] As an example, define the state-action space, which can be expressed as:

[0134] Formula (14)

[0135] in, s t represents the state vector, a t represents the action vector, K w represents the wind power generation dispatch weight, K s represents the solar power generation scheduling weight, K grid represents the grid interaction weight, K i Indicates the energy scheduling priority.

[0136] Furthermore, the wind power generation dispatch weight is adjusted according to the membership function K w , which can be expressed as:

[0137] Formula (15)

[0138] in, represents the adjusted wind power generation dispatch weight, μ wind (v w ) Represents wind speed membership, which is used to characterize wind energy availability.

[0139] Wind speed membership μ wind (v w ) It can be expressed as:

[0140] Formula (16)

[0141] in, v cut-in Indicates the fan startup wind speed; v rated Indicates the rated wind speed, i.e. the maximum output wind speed; the real-time wind speed is used in this formula v w The calculation is in line with the physical definition of the wind turbine power curve and directly corresponds to the technical parameters of the wind turbine. However, the normalized wind speed may destroy the physical meaning of the membership function and cause the control strategy to lose the equipment protection function. Therefore, the real-time wind speed is used instead of the normalized wind speed.

[0142] Furthermore, the Q-learning update rule is used to optimize the long-term scheduling strategy, which is expressed as:

[0143] Formula (17)

[0144] Among them, Q(s,a) represents the current state-action pair; Q(s',a') represents the next state-optimal action pair; It means taking the maximum value of Q(s',a') for all a'; η Represents the learning rate, which is used to update the step size; γ represents the discount factor, i.e. the future reward attenuation coefficient; r represents the reward function.

[0145] Furthermore, the design of the reward function r is expressed as:

[0146] Formula (18)

[0147] in, P curt It represents the abandoned wind and solar power, that is, the renewable energy power that cannot be used. Penalizing abandoned energy can improve the utilization rate of wind and solar energy.

[0148] In one embodiment, the design λ 1=0.4 means energy abandonment penalty; λ 2=0.2 means the energy storage battery is charged SOC bat balanced; λ 3=0.1 means the voltage is stable.

[0149] As an example, rolling horizon control is used to accurately track short-term scheduling strategies. Rolling horizon control and Q-learning together constitute hierarchical control of scheduling strategies. Rolling horizon control is expressed as:

[0150] Formula (19)

[0151] Among them, u t:t+H represents the control input sequence from time t to t+H; P gen ( k ) represents the total generated power at the kth moment; P demand (k) represents the charging demand power at the kth moment, which is obtained by forward-looking optimization scheduling based on load demand forecast power; ρ Represents the control cost weight coefficient, which is used to balance tracking accuracy and motion amplitude; R Represents the control input weight matrix, usually a diagonal matrix; u ( k ) represents the control input sequence at time k. The corresponding variable value scheme that makes the rolling horizon control formula reach the minimum value is used to optimize the energy scheduling strategy.

[0152] As an example, after optimizing the energy scheduling strategy using reinforcement learning, the second step is to use the interior point method to solve the quadratic programming of the scheduling strategy, which can be expressed as:

[0153] Formula (20)

[0154] in, represents the minimum value of the revised wind power generation forecast power, Indicates the maximum value of the revised wind power generation forecast power, Indicates the minimum charge value of the energy storage battery. Indicates the maximum charge of the energy storage battery.V nom Indicates the nominal voltage of the system. For example, the allowable fluctuation range of an 800V DC bus is 760V~840V.

[0155] For example, when the wind speed fluctuates greatly: relax the maximum value of the wind power forecast power Limitation allows greater fluctuations in wind power output and reduces the penalty for wind power regulation by adjusting the control input weight matrix R; when photovoltaic power is insufficient on cloudy days: tighten the minimum charge value of the energy storage battery , such as increasing from 20% to 30%, increasing battery discharge power The LSTM prediction model is used to predict energy generation capacity and load demand, and the strategy is adjusted according to real-time fluctuations to ensure stable operation of the charging station.

[0156] In step S5, see Figure 1 In step S5, the selected energy scheduling strategy is executed to adjust the working mode of the charging station.

[0157] In another embodiment of the present application, please refer to Figure 3 , Figure 3 This is a module diagram of the charging station design system of the present application. The present application provides a charging station design system for executing the above-mentioned charging station design method, including: an environmental perception module 21, a data preprocessing and feature extraction module 22, a deep learning and prediction module 23, an adaptive control module 24 and an execution and scheduling module 25.

[0158] As an example, the environment perception module 21, the data preprocessing and feature extraction module 22, the deep learning and prediction module 23, the adaptive control module 24 and the execution and scheduling module 25 are connected in sequence.

[0159] As an example, the environmental perception module 21 obtains real-time meteorological data and load data, and outputs them to the data preprocessing and feature extraction module 22; the data preprocessing and feature extraction module 22 preprocesses the meteorological data and load data to obtain preprocessed data, performs feature extraction on the preprocessed data to obtain prediction features, and outputs them to the deep learning and prediction module 23; the deep learning and prediction module 23 establishes a prediction model for energy generation capacity and load demand, and according to the prediction features, uses the established prediction model to obtain the initial wind power generation prediction power, initial solar power generation prediction power and initial load demand prediction power at several moments in the future, and outputs them to the adaptive control module 24; the adaptive control module 24 selects an energy scheduling strategy based on the predicted power results of the prediction model, meteorological sensor data and preset control rules, and outputs it to the execution and scheduling module 25; the execution and scheduling module 25 executes the selected energy scheduling strategy and adjusts the working mode of the charging station.

[0160] In another embodiment of the present application, please refer to Figure 4 , Figure 4 This is a circuit diagram of a charging station of the present application. The present application provides a charging station for executing the above-mentioned charging station design method. The charging station includes:

[0161] Wind, solar and storage system 31;

[0162] A dual-winding generator 32 , wherein a first end of the dual-winding generator 32 is connected to a first end of the wind-solar-storage system 31 ;

[0163] ACDC converter 33, a first end of the ACDC converter 33 is connected to a first end of the dual-winding generator 32 and a first end of the wind-solar-storage system 31 respectively;

[0164] An ACDC converter 34 , wherein a first end of the ACDC converter 34 is connected to a second end of the dual-winding generator 32 ;

[0165] A 400V charging pile 35 , wherein a first end of the 400V charging pile 35 is connected to a second end of the ACDC converter 33 ;

[0166] An 800V charging pile 36 , wherein a first end of the 800V charging pile 36 is connected to a second end of the ACDC converter 34 ;

[0167] A mains interface 37 is reserved, and a first end of the mains interface 37 is connected to a first end of the wind-solar-storage system 31 , a first end of the dual-winding generator 32 , and a first end of the ACDC converter 33 , respectively.

[0168] As an example, the wind-solar-storage system 31 includes:

[0169] Energy storage battery 311;

[0170] A wind turbine 312, wherein a first end of the wind turbine 312 is connected to a first end of the energy storage battery 311;

[0171] A solar panel 313, wherein a first end of the solar panel 313 is connected to a first end of the energy storage battery 311 and a first end of the wind turbine 312 respectively;

[0172] The bidirectional DCAC converter 314 has a first end connected to the first end of the energy storage battery 311, the first end of the wind turbine 312, and the first end of the solar panel 313 respectively; the second end of the bidirectional DCAC converter 314 is the first end of the wind-solar-storage system 31.

[0173] As an example, the execution and scheduling module 25 of the charging station design system controls the charging station to execute and schedule the energy scheduling strategy according to the selected energy scheduling strategy.

[0174] As an example, in the wind-solar-storage system 31 , the wires connecting the energy storage battery 311 , the wind turbine 312 , the energy storage battery 311 , and the bidirectional DCAC converter 314 are 400V DC busbars.

[0175] As an example, the wires connecting the wind-solar-storage system 31, the dual-winding generator 32, the ACDC converter 33, and the reserved AC power interface 37 are 400V DC busbars; the wires connecting the ACDC converter 33 and the 400V charging pile 35 are 400V DC busbars.

[0176] As an example, the wires connecting the dual-winding generator 32 and the ACDC converter 34 are 800V DC bus bars; the wires connecting the ACDC converter 34 and the 800V charging station 36 are 800V DC bus bars.

[0177] As an example, energy storage battery 311 uses a lithium battery pack as its energy storage medium, connected to a 400V DC bus and connected to a 400V AC bus via a bidirectional DC / AC converter 314. Energy storage battery 311 can store excess energy. When the power generated by the solar panels 313 and wind turbine 312 exceeds the required power for vehicle charging, energy storage battery 311 is charged to store energy, releasing it during periods of insufficient wind and solar power generation or peak demand. When wind and solar power fluctuate or are temporarily insufficient to meet vehicle charging requirements, energy storage battery 311 discharges to replenish the shortfall and ensure continuous vehicle charging. Furthermore, energy storage battery 311 stabilizes the 400V DC bus voltage. The battery management system (BMS) monitors the battery's charge, voltage, temperature, and other parameters in real time to ensure the safe and efficient operation of the wind, solar, and storage system 31.

[0178] Wind turbine 312 is installed at an appropriate height and location to capture stable wind energy. The AC power generated by wind turbine 312 is converted to DC by a rectifier and then fed into a 400V DC bus via a maximum power point tracking (MPPT) controller. This controller adjusts the operating parameters of wind turbine 312 based on wind speed fluctuations, ensuring efficient power generation under varying wind speed conditions. When wind speeds are within the rated range, wind turbine 312 generates stable power and supplies it to the 400V DC bus, providing energy for the subsequent charging process and the energy storage battery 311. When wind speeds are excessively high or low, wind turbine 312 implements appropriate protective measures, such as speed limiting or shutdown, to ensure the safety of the charging station.

[0179] Solar panels 313 are arranged on the roof of the charging station or in an open area around it, converting solar energy into direct current (DC) electricity, which is then connected to a 400V DC bus via a maximum power point tracking (MPPT) controller. The MPPT controller monitors the output voltage and current of the solar panels 313 in real time, dynamically adjusting the operating point based on environmental factors such as solar irradiance, ensuring that the solar panels 313 always output electrical energy at maximum power. This electrical energy is then transmitted via the 400V DC bus to provide power for the subsequent charging process and the energy storage battery 311.

[0180] The bidirectional DCAC converter 314 connects the 400V DC bus and the 400V AC bus. When the energy storage battery 311 is fully charged, the bidirectional DCAC converter 314 converts DC power into AC power for charging the vehicle. When the energy storage battery 311 is undercharged and wind and solar power cannot be replenished in time, the energy storage battery 311 can be charged by converting the AC power generated by the dual-winding generator 32 into DC power.

[0181] When the charging station is operating normally (i.e., when strategies 1, 2, 3, or 4 are selected), the dual-winding generator 32 functions as a transformer. Its control winding is connected to the 400V DC busbar via a bidirectional DC / AC converter 314, converting the 400V AC output of the wind, solar, and storage system 31 into 800V AC for fast charging of the vehicle, while also outputting 400V AC for conventional charging. Furthermore, based on energy supply conditions and cost-benefit analysis, the decision to activate the dual-winding generator 32 is made to optimize energy utilization and reduce operating costs. In the event of an abnormal charging station emergency, when neither the wind, solar, and storage system 31 nor the grid is providing power, strategy 5 is selected. The charging station adopts strategy 5: island operation mode, in which the dual-winding generator 32 is activated as a generator. Its control winding outputs 400V AC, which is connected to the ACDC converter 33 to output 400V DC for conventional charging of the electric vehicle. The power winding side outputs 800V AC power, which is connected to the ACDC converter 34 to output 800V DC power to supply electric vehicles for fast charging, achieving simultaneous output of 400V and 800V.

[0182] A reserved mains power interface 37 is connected to the 400V AC busbar. When the combined power supply of the wind, solar, and energy storage system 31 still cannot meet the vehicle's charging needs, that is, when Strategy 3 or Strategy 4 is selected, the control device obtains power from the grid and converts it accordingly to charge the vehicle. During low-demand periods, based on the electricity price strategy and the charge status of the energy storage battery 311, low-cost power is obtained from the grid and stored in the energy storage battery 311. During peak periods, the energy stored in the energy storage battery 311 is preferentially used to charge the vehicle, and power is then drawn from the grid when necessary, achieving a two-way flow of power and optimized allocation between the grid and the charging station.

[0183] This application provides a charging station design method, charging station design system, and charging station. Based on a long short-term memory (LSTM) network model, a prediction model for energy generation power and load demand is established. This model can more accurately predict renewable energy generation and charging demand, thereby achieving more efficient energy distribution and management and improving charging station operational efficiency. A dual-winding generator design can output different voltage levels to meet the charging needs of various vehicles. This eliminates the need for a DC-DC converter, simplifies the circuit structure, reduces losses during energy conversion, and improves charging efficiency. The multi-voltage output characteristics of the dual-winding generator 32 enable the charging station to accommodate the charging needs of various types of electric vehicles, enhancing its applicability and flexibility. By integrating a wind, solar, and storage system 31, it fully utilizes renewable energy sources such as solar and wind power, reducing dependence on traditional utility power, lowering energy costs, and improving the charging station's energy sustainability. The charging station design system and energy allocation strategy enable real-time monitoring and control of various components within the charging station, ensuring efficient and flexible energy distribution and utilization, and improving the stability and cost-effectiveness of the charging station's operations.

[0184] In summary, the charging station design method, charging station design system and charging station provided in this application significantly improve the energy utilization efficiency, economy and stability of the charging station, while reducing operating costs and environmental impact.

[0185] Although the present application has been disclosed above with reference to the embodiments, they are not intended to limit the present application. Anyone with ordinary knowledge in the technical field may make slight changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be determined by the scope of the appended patent application.

Claims

1. A charging station design method, characterized in that: include, Step S1: Obtain real-time meteorological data and load data; Step S2: preprocessing the meteorological data and the load data to obtain preprocessed data; Performing feature extraction on the preprocessed data to obtain prediction features; Step S3: Establishing a prediction model for energy generation power and load demand, and using the established prediction model to obtain the initial wind power generation prediction power, initial solar power generation prediction power, and initial load demand prediction power at a future time based on the prediction characteristics; Step S4: selecting an energy scheduling strategy based on the predicted power results of the prediction model, meteorological sensor data, and preset control rules; Step S5: Execute the selected energy scheduling strategy and adjust the working mode of the charging station; In the step S2, after the feature extraction, the step further includes: calculating the weights of the extracted features, and selecting the features with higher weights as the prediction features; the prediction features include: normalized wind speed, normalized solar irradiance, standardized temperature, hourly cycle, first-order difference of load power, and load power; In step S3, a long short-term memory network model is used to establish a prediction model for energy generation capacity and load demand, wherein the input of the prediction model is the prediction feature, and the output of the prediction model is the initial wind power generation prediction power, the initial solar power generation prediction power, and the initial load demand prediction power at the future time; the network parameters in the prediction model are three layers of long short-term memory network layers, and the number of units in each long short-term memory network layer is 128, 64, and 32, respectively; After the prediction model outputs the predicted power result, the method further includes: correcting the initial wind power generation predicted power and the initial solar power generation predicted power to obtain a corrected wind power generation predicted power and a corrected solar power generation predicted power.

2. A charging station design method according to claim 1, characterized in that: In step S1, the meteorological data includes: meteorological sensor data and meteorological forecast data, wherein the meteorological sensor data includes: wind speed, solar irradiance, ambient temperature, and relative humidity; the load data includes: load demand, DC bus voltage, and energy storage battery state of charge.

3. A charging station design method according to claim 1, characterized in that: In step S2, the preprocessing includes: maximum normalization, sliding mean filtering, and data standardization; The feature extraction includes: extracting valuable features from the preprocessed data, including time series features: daily variation trends of wind speed and solar irradiance; extracting weather change features based on meteorological data in the preprocessed data; generating load prediction features based on historical load trends of load data in the preprocessed data; and constructing periodic features.

4. A charging station design method according to claim 1, characterized in that: In the step S4, the preset control rules include: steps S41 to S413; Step S41: Energy dispatch control starts; Step S42: Determine whether the charging station status is normal. If so, proceed to step S43; if not, proceed to step S410; Step S43: Determine whether the revised wind power generation forecast power and the revised solar power generation forecast power are greater than the load demand forecast power. If so, proceed to step S44; if not, proceed to step S46. Step S44: selecting strategy 1: wind and solar priority mode; Step S45: adjusting the output according to the revised wind power generation prediction power and the revised solar power generation prediction power; proceeding to step S411; Step S46: Determine the state of charge of the energy storage battery. If the state of charge is greater than 70%, proceed to step S47; if the state of charge is greater than 30% and less than 70%, proceed to step S48; if the state of charge is less than 30%, proceed to step S49; Step S47: Select strategy 2: wind-solar-storage collaborative mode, enable constant voltage-frequency ratio control, and proceed to step S411; Step S48: Select strategy 3: storage-grid collaboration mode 1, enable droop control, and jointly output power from the storage and grid, and proceed to step S411; Step S49: Select strategy 4: storage-grid collaboration mode 2, energy storage is charged from the grid, and proceed to step S411; Step S410: Select strategy 5: island operation mode, enable virtual synchronous dual-winding generator control, and proceed to step S411; Step S411: executing the policy; Step S412: Evaluate the energy dispatch control status, including: the state of charge of the energy storage battery, the error of the predicted power result, and the quality of the grid voltage. If the actual requirements of the charging station are met, proceed to step S413; if not, return to step S41; Step S413: Energy scheduling control ends.

5. A charging station design method according to claim 1, characterized in that: In the step S5, after selecting the energy scheduling strategy, the step further includes: continuously optimizing the scheduling strategy through reinforcement learning.

6. A charging station design system, characterized in that: Executing a charging station design method according to any one of claims 1 to 5, comprising: an environment perception module, a data preprocessing and feature extraction module, a deep learning and prediction module, an adaptive control module, and an execution and scheduling module; The environmental perception module, the data preprocessing and feature extraction module, the deep learning and prediction module, the adaptive control module and the execution and scheduling module are connected in sequence.

7. A charging station, characterized in that: Executing a charging station design method according to any one of claims 1 to 5, comprising: Wind, solar and storage systems; A dual-winding generator, wherein a first end of the dual-winding generator is connected to a first end of the wind-solar-storage system; a first ACDC converter, wherein a first end of the first ACDC converter is connected to a first end of the dual-winding generator and a first end of the wind-solar-storage system respectively; a second ACDC converter, wherein a first end of the second ACDC converter is connected to a second end of the dual-winding generator; A 400V charging pile, wherein a first end of the 400V charging pile is connected to a second end of the first ACDC converter; An 800V charging pile, wherein a first end of the 800V charging pile is connected to a second end of the second ACDC converter; A mains interface is reserved, wherein a first end of the reserved mains interface is connected to a first end of the wind-solar-storage system, a first end of the dual-winding generator, and a first end of the ACDC converter respectively; When the charging station is in normal condition, the dual-winding generator is used as a transformer when strategies 1, 2, 3, and 4 are selected. In an emergency situation where the charging station is in abnormal condition, the dual-winding generator 32 is started as a generator when strategy 5 is selected.

8. A charging station according to claim 7, characterized in that: The wind-solar-storage system comprises: Energy storage batteries; a wind turbine generator, wherein a first end of the wind turbine generator is connected to a first end of the energy storage battery; a solar panel, wherein a first end of the solar panel is connected to a first end of the energy storage battery and a first end of the wind turbine respectively; A bidirectional DCAC converter, wherein the first end of the bidirectional DCAC converter is respectively connected to the first end of the energy storage battery, the first end of the wind turbine, and the first end of the solar panel; the second end of the bidirectional DCAC converter is the first end of the wind-solar-storage system.

Images