A method for energy management of a charging station

Abstract

The present invention belongs to the technical field of power station energy management, and provides a charging station energy management method, including: electric pile data collection, data collection of allocation requests to be processed, generation of electric pile allocation schemes, determination of energy supply and demand status, user guidance, mobile charging pile recycling, combined charging by photovoltaic panels and the power grid, separate charging by the power grid, and battery replacement during peak periods. Through the energy management model, the present invention realizes the allocation of the best mobile charging piles for users, improves the utilization rate of electric piles, and reduces the waiting time of users; through the energy supply and demand status and environmental light intensity, different charging methods under different working conditions are realized, and the operating cost is reduced; by replacing the battery during the peak electricity consumption period, the energy replenishment efficiency of the mobile charging pile is further improved, and the pressure of power shortage at the charging station during peak periods is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power station energy management, and particularly to a method for energy management of a charging station. Background Art

[0002] The energy management technology of charging stations has developed rapidly with the large-scale application of electric vehicles and the low-carbon transformation of the energy system. It aims to solve the contradiction between the surging charging demand and the grid stability, economy, and sustainability through multi-dimensional collaborative optimization. Its core background technologies cover three aspects: First, the collaborative challenges between infrastructure and the grid, including local grid overload caused by large-scale centralized charging, the peak-valley electricity price-guided off-peak charging mechanism, and the charging station participating in the grid demand response as a flexible load; Second, the integration of distributed energy, by configuring renewable energy such as photovoltaic and wind power and energy storage systems, to build an integrated "photovoltaic-storage-charging" microgrid to improve the green electricity consumption capacity and power supply reliability; Third, the intelligent control system, relying on the Internet of Things, 5G communication, and AI algorithms to achieve dynamic scheduling, such as orderly charging based on user behavior prediction, V2G two-way charging and discharging to regulate the grid frequency, and using optimization models such as reinforcement learning to balance costs, carbon emissions, and user experience. In addition, power electronics technology has improved the energy conversion efficiency, and the policy side promotes the implementation of the technology through grid connection specifications and carbon trading mechanisms.

[0003] However, the existing energy management technologies of charging stations generally have problems such as lagging dynamic response, fragmentation of multi-objective optimization, and low automation level. Specifically, they include the inability to adapt to real-time changing queuing situations and user dynamic demands; real-time data not effectively driving dynamic scheduling, still relying on manual intervention, lacking temporal analysis and pattern mining of historical data, and being unable to predict future load trends; slow battery replacement efficiency and other problems. Summary of the Invention

[0004] In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide a method for energy management of a charging station, which shortens the waiting time of users and improves the utilization efficiency of mobile charging piles.

[0005] To achieve the above object, the present invention provides the following solution:

[0006] A method for energy management of a charging station, comprising:

[0007] Collecting the power information of each mobile charging pile in all charging stations to obtain charging pile data;

[0008] Collecting the electricity consumption applications and electricity consumption reservations of users to obtain data of requests to be allocated;

[0009] Inputting the charging pile data and the data of requests to be allocated into a pre-trained energy management model for scheme generation to obtain a charging pile allocation scheme;

[0010] According to the pile data, calculate the proportion of the mobile charging piles in use in each charging pile to obtain the energy supply and demand status;

[0011] Determine the mobile charging piles in the pile allocation plan as the to-be-operated charging piles, and guide the user to the to-be-operated charging piles for charging or provide the target pile location information of the to-be-operated charging piles to the user through a mobile terminal;

[0012] Use a mobile charging vehicle to transport the mobile charging piles with a power level lower than the normal power threshold in all the charging stations back to the charging base;

[0013] When the ambient light intensity exceeds the preset intensity benchmark and the energy supply and demand status is a low electricity consumption peak period, jointly charge the mobile charging piles using the power grid and the photovoltaic panels in the charging base;

[0014] When the ambient light intensity is lower than the preset intensity benchmark or the energy supply and demand status is a high electricity consumption peak period, charge the mobile charging piles using the power grid;

[0015] When the energy supply and demand status is a high electricity consumption peak period, if the proportion of the mobile charging piles with a power level lower than the normal power threshold in the charging station exceeds the warning ratio, use the mobile charging vehicle to transport the spare battery with a power proportion of 100% in the charging base to the charging station, and use the battery replacement device on the mobile charging vehicle to replace the spare battery into the mobile charging piles with a power level lower than the normal power threshold.

[0016] Preferably, the pile data includes: power data, estimated power consumption for current vehicle charging, number of queuing vehicles, reserved power consumption of queuing vehicles, pile location data, charging type data, real-time status, rate, and historical utilization rate.

[0017] Preferably, the battery replacement device includes: a mechanical claw, a mechanical screwdriver, a camera, an upper guide rail, a lower guide rail, and a magnetic attraction plate;

[0018] The upper guide rail is arranged on the roof of the mobile charging vehicle; the mechanical claw and the mechanical screwdriver are respectively arranged on the upper guide rail; the lower guide rail is arranged on the inner bottom side of the cargo box of the mobile charging vehicle; the camera is fixed at the inner top end of the cargo box.

[0019] Preferably, when the mobile charging pile uses a ternary lithium battery, the normal power threshold is 20% to 80%; when the mobile charging pile uses a lithium iron phosphate battery, the normal power threshold is 20% to 90%.

[0020] Preferably, the calculation formula for the preset intensity benchmark is: ;

[0021] Among them, is the preset intensity benchmark; is the rated charging power of a single mobile charging pile; is the energy conversion efficiency of the photovoltaic panel; is the effective daylighting area of the photovoltaic panel; is the loss coefficient.

[0022] Preferably, the request data to be allocated includes: reservation time, reserved electricity consumption, user location data, charging type requirements, battery capacity, priority data, and waiting tolerance data.

[0023] Preferably, when the energy supply and demand state is the peak electricity consumption period, if the proportion of the mobile charging piles with electricity lower than the normal electricity threshold in the charging station exceeds the warning ratio, the mobile charging vehicle transports the spare battery with a power ratio of 100% in the charging base to the charging station, and uses the battery replacement device on the mobile charging vehicle to replace the spare battery into the mobile charging pile with electricity lower than the normal electricity threshold, including:

[0024] Collect image data using the camera, and use an object recognition algorithm to recognize the image data to obtain the spatial position information of the mobile charging pile where the battery needs to be replaced;

[0025] Move the robotic arm to the position on the upper rail closest to the mobile charging pile where the battery needs to be replaced according to the spatial position information;

[0026] Use a laser sensor to locate the grooves on both sides of the mobile charging pile, and transfer the mobile charging pile to the replacement station on the lower rail through the grooves using the robotic arm;

[0027] Use an object recognition algorithm to detect the fixing screws on the rear cover of the mobile charging pile to obtain the screw placement information;

[0028] Remove the fixing screws on the rear cover using a mechanical screwdriver according to the screw placement information;

[0029] Transfer the rear cover to the temporary storage area using the magnetic plate;

[0030] Detect the battery to be replaced in the mobile charging pile using the camera and an object recognition algorithm to obtain the battery placement information;

[0031] Transfer the battery to be replaced to the recycling area on the lower rail using the robotic arm according to the battery placement information, and install the prepared spare battery into the mobile charging pile;

[0032] After the spare battery is installed, use the mechanical screwdriver to install the back cover onto the mobile charging pile, and use the mechanical claw to transfer the mobile charging pile to its original position.

[0033] Preferably, the training process of the energy management model includes:

[0034] Collect the pile data and the allocation request data of the target research power station, and reversely formulate a pile allocation plan that meets each stage according to the pile data and the allocation request data;

[0035] Standardize the pile data using the Z-Score formula;

[0036] Perform one-hot encoding on the charging type data and the real-time status to obtain categorical features;

[0037] Calculate the Euclidean distance based on the pile location data and the user location data to obtain location features;

[0038] Standardize the allocation request data using Min-Max;

[0039] Construct an original model with a dual tower encoder and a cross-attention mechanism;

[0040] Construct a loss function;

[0041] Use the pile encoding tower of the original model to encode the standardized pile data, the categorical features, and the location features to obtain a pile embedding vector;

[0042] Use the request encoding tower of the original model to encode the standardized allocation request data to obtain a request embedding vector;

[0043] Use the request embedding vector as the Query, the pile embedding vector as the Key and Value, and calculate the attention weights through the cross-attention mechanism to obtain a context vector;

[0044] Use the mask layer of the original model to perform hard constraints and dynamic masking on the context vector to obtain a masked output;

[0045] Use the fully connected layer and Softmax of the original model to map and calculate the probability distribution of the masked output to obtain an allocation matrix;

[0046] Calculate the loss function based on the allocation matrix and the formulated pile allocation plan to obtain a predicted loss value;

[0047] Optimize the original model according to the predicted loss value by using an optimizer, and return to the step of "encoding the standardized pile data, the categorical features, and the location features by using the pile encoding tower of the original model to obtain a pile embedding vector", so as to obtain the trained energy management model.

[0048] Preferably, determine the mobile charging piles in the pile allocation plan as the to-be-operated charging piles, and guide the user to go to the to-be-operated charging piles for charging or provide the user with the target pile location information of the to-be-operated charging piles through a mobile terminal, including:

[0049] Input the user location data and the target pile location information into a navigation software to obtain a travel route, and update the travel route to the user's mobile terminal.

[0050] Preferably, the expression of the loss function is: ;

[0051] where is the calculated value of the loss function; is the number of requests; is the number of charging piles; is the priority weight; is a dynamically generated mask; is the true label; is the predicted success probability; is the balance coefficient; is the variance of historical utilization rate.

[0052] The present invention discloses the following technical effects:

[0053] The present invention provides a charging station energy management method. Through the energy management model, the problem that the existing technology has an unreasonable solution formulation when dealing with the dynamic demands of users is solved, and the best mobile charging piles are allocated for users; through the energy supply and demand state and the environmental light intensity, the problems that the conventional charging stations have a single energy supplement source and a low charging efficiency caused by simply intervening in photovoltaic panels are solved, and different charging methods under different working conditions are realized; by replacing the battery during the peak electricity consumption period, the problem that users wait for a long time during the peak electricity consumption period is solved, and the rapid energy supplement of the mobile charging piles is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0055] Figure 1 Schematic diagram of the charging station energy management process provided by the embodiment of the present invention;

[0056] Figure 2 Schematic diagram of the battery replacement process provided by the embodiment of the present invention;

[0057] Figure 3 Schematic diagram of the energy management model training process provided by the embodiment of the present invention. Detailed implementation manners

[0058] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0059] The purpose of the present invention is to provide a charging station energy management method, which shortens the waiting time of users and improves the utilization efficiency of mobile charging piles.

[0060] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific implementation manners.

[0061] Figure 1 Schematic diagram of the charging station energy management process provided by the embodiment of the present invention, as Figure 1 shown, the present invention provides a charging station energy management method, including:

[0062] Step 100: Collect the power information of each mobile charging pile in all charging stations to obtain charging pile data;

[0063] Step 200: Collect the electricity consumption applications and electricity consumption reservations of users to obtain the to-be-allocated request data;

[0064] Step 300: Input the charging pile data and the to-be-allocated request data into a pre-trained energy management model for generating a solution to obtain a charging pile allocation solution;

[0065] Step 400: According to the charging pile data, count the occupancy ratio of the in-use mobile charging piles in each charging pile to obtain the energy supply and demand status;

[0066] Step 500: Determine the mobile charging piles in the charging pile allocation solution as the to-be-operated charging piles, and guide the users to the to-be-operated charging piles for charging or provide the target charging pile location information of the to-be-operated charging piles to the users through a mobile terminal;

[0067] Step 600: Use a mobile charging vehicle to transport the mobile charging piles with a power level lower than the normal power threshold in all charging stations back to the charging base;

[0068] Step 700: When the ambient light intensity exceeds the preset intensity benchmark and the energy supply and demand status is at a low electricity consumption peak, use the power grid and the photovoltaic panels in the charging base to jointly charge the mobile charging piles;

[0069] Step 800: When the ambient light intensity is lower than the preset intensity benchmark or the energy supply and demand status is at a high electricity consumption peak, use the power grid to charge the mobile charging piles;

[0070] Step 900: When the energy supply and demand status is at a high electricity consumption peak, if the proportion of mobile charging piles with a power level lower than the normal power threshold in the charging station exceeds the warning ratio, use a mobile charging vehicle to transport the backup battery with a power ratio of 100% in the charging base to the charging station, and use the battery replacement equipment on the mobile charging vehicle to replace the backup battery into the mobile charging piles with a power level lower than the normal power threshold.

[0071] Further, the charging pile data includes: power data, estimated power consumption for current vehicle charging, number of queuing vehicles, reserved power consumption of queuing vehicles, charging pile location data, charging type data, real-time status, rate, and historical utilization rate.

[0072] Specifically, the battery replacement equipment includes: mechanical claws, mechanical screwdrivers, cameras, upper guide rails, lower guide rails, and magnetic attraction plates;

[0073] The upper guide rails are arranged on the roof of the mobile charging vehicle; the mechanical claws and mechanical screwdrivers are respectively arranged on the upper guide rails; the lower guide rails are arranged on the inner bottom side of the cargo box of the mobile charging vehicle; the cameras are fixed at the inner top of the cargo box.

[0074] Preferably, when the mobile charging pile uses a ternary lithium battery, the normal power threshold is 20% to 80%; when the mobile charging pile uses a lithium iron phosphate battery, the normal power threshold is 20% to 90%.

[0075] Further, the calculation formula for the preset intensity benchmark is: ;

[0076] Wherein, is the preset intensity benchmark; is the rated charging power of a single mobile charging pile; is the energy conversion efficiency of the photovoltaic panel; is the effective daylighting area of the photovoltaic panel; is the loss coefficient.

[0077] Specifically, the request data to be allocated includes: reservation time, reserved power consumption, user location data, charging type requirements, battery capacity, priority data, and waiting tolerance data.

[0078] Furthermore, when the energy supply and demand state is at the peak electricity consumption period, if the proportion of mobile charging piles with power below the normal power threshold in the charging station exceeds the warning ratio, a spare battery with a power ratio of 100% in the charging base is transported to the charging station by a mobile charging vehicle, and the spare battery is replaced into the mobile charging pile with power below the normal power threshold by using the battery replacement device on the mobile charging vehicle, including:

[0079] Step 901: Collect image data using a camera, and identify the image data using an object recognition algorithm to obtain the spatial position information of the mobile charging pile with the battery to be replaced.

[0080] Step 902: Move the robotic arm to the position on the upper guide rail closest to the mobile charging pile with the battery to be replaced according to the spatial position information.

[0081] Step 903: Use a laser sensor to locate the grooves on both sides of the mobile charging pile, and transfer the mobile charging pile to the replacement station on the lower guide rail through the grooves using the robotic arm.

[0082] Step 904: Use an object recognition algorithm to detect the fixing screws on the rear cover of the mobile charging pile to obtain the screw placement information.

[0083] Step 905: Remove the fixing screws on the rear cover using a mechanical screwdriver according to the screw placement information.

[0084] Step 906: Transfer the rear cover to the temporary storage area using a magnetic plate.

[0085] Step 907: Detect the battery to be replaced in the mobile charging pile using an object recognition algorithm through the camera to obtain the battery placement information.

[0086] Step 908: Transfer the battery to be replaced to the recycling area on the lower guide rail using the robotic arm according to the battery placement information, and install the prepared spare battery into the mobile charging pile.

[0087] Step 909: After the spare battery is installed, install the rear cover onto the mobile charging pile using a mechanical screwdriver, and transfer the mobile charging pile to its original position using the robotic arm.

[0088] Specifically, the training process of the energy management model includes:

[0089] Step 301: Collect the charging pile data of the target research station, the request data to be allocated, and reverse-formulate the charging pile allocation plan for each stage according to the charging pile data and the request data to be allocated.

[0090] Step 302: Standardize the charging pile data using the Z-Score formula;

[0091] Step 303: Perform one-hot encoding on the charging type data and real-time status to obtain categorical features;

[0092] Step 304: Calculate the Euclidean distance based on the charging pile location data and user location data to obtain location features;

[0093] Step 305: Standardize the to-be-allocated request data using Min-Max;

[0094] Step 306: Construct an original model with a dual tower encoder and a cross-attention mechanism;

[0095] Step 307: Construct a loss function;

[0096] Step 308: Encode the standardized charging pile data, categorical features, and location features using the charging pile encoding tower of the original model to obtain charging pile embedding vectors;

[0097] Step 309: Encode the standardized to-be-allocated request data using the request encoding tower of the original model to obtain request embedding vectors;

[0098] Step 310: Use the request embedding vector as the Query, the charging pile embedding vectors as the Key and Value, and calculate the attention weights through the cross-attention mechanism to obtain context vectors;

[0099] Step 311: Use the mask layer of the original model to perform hard constraints and dynamic masking on the context vectors to obtain masked outputs;

[0100] Step 312: Use the fully connected layer and Softmax of the original model to map the masked outputs and calculate the probability distribution to obtain an allocation matrix;

[0101] Step 313: Calculate the loss function based on the allocation matrix and the formulated charging pile allocation plan to obtain a predicted loss value;

[0102] Step 314: Optimize the original model according to the predicted loss value using an optimizer, and return to the step "Encode the standardized charging pile data, categorical features, and location features using the charging pile encoding tower of the original model to obtain charging pile embedding vectors" to obtain a trained energy management model.

[0103] Preferably, determine the mobile charging piles in the charging pile allocation plan as the to-be-operated charging piles, and guide the user to the to-be-operated charging piles for charging or provide the user with the target charging pile location information of the to-be-operated charging piles through a mobile terminal, including:

[0104] Input the user location data and the target charging pile location information into the navigation software to obtain the travel route, and update the travel route to the user's mobile terminal.

[0105] Specifically, the expression of the loss function is: ;

[0106] where is the calculated value of the loss function; is the number of requests; is the number of charging piles; is the priority weight; is the dynamically generated mask; is the true label; is the predicted success probability; is the balance coefficient; is the variance of historical utilization rate.

[0107] Preferably, preprocess the charging pile data. Numerical feature standardization: For continuous numerical values such as power, queuing number, rate, historical utilization rate, etc., use Z-Score standardization (divide by the standard deviation after subtracting the mean) to eliminate the dimension difference. Supplement the time dimension for the historical utilization rate: Calculate the moving window mean of the past 1 hour to capture dynamic changes.

[0108] Furthermore, encode the categorical features. Use One-Hot encoding for the charging type (such as fast charging / slow charging) and real-time status (idle / occupied / faulty) to generate sparse binary vectors. Supplement the "expected idle time" for the status field: Dynamically calculate according to the current vehicle charging time.

[0109] Optionally, process the location data: Combine the longitude and latitude of the charging pile with the longitude and latitude requested by the user, and calculate the Euclidean distance or spherical distance (Haversine formula) as a feature. Cluster the locations into regions (such as 500-meter grids), and convert them into Embedding encodings of region IDs.

[0110] Specifically, preprocess the request data. Time feature processing: Split the reservation time into time series features such as hour, day of the week, and whether it is a holiday, and encode the periodicity with sine / cosine. Calculate the "difference between the current time and the reservation time" as an urgency indicator.

[0111] Furthermore, normalize the numerical features: Use Min-Max normalization for the reserved power consumption and battery capacity, and scale them to the interval [0, 1]. Convert the waiting tolerance into a numerical type, such as "the maximum waiting time (minutes) acceptable to the user".

[0112] Optionally, priority processing: Map priorities (such as VIP / ordinary users) to weight values (such as 1.5 and 1.0) for subsequent weighting of the loss function.

[0113] Specifically, implementation of the neural network architecture. Dual-tower encoder design:

[0114] 1) Charging pile encoding tower:

[0115] Input layer: Receive the preprocessed charging pile feature vector (including normalized values, One-Hot categories, distance, etc.);

[0116] Hidden layer: Use 3 fully connected layers (e.g., 256, 128, 64 nodes), followed by ReLU activation and BatchNormalization for each layer;

[0117] Output layer: Generate a 64-dimensional charging pile embedding vector to represent the static attributes and dynamic states of the charging pile.

[0118] 2) Request encoding tower:

[0119] Input layer: Receive the request feature vector (time, location, electricity demand, etc.);

[0120] Hidden layer: Symmetric structure with the charging pile tower (256, 128, 64 nodes), output a 64-dimensional request embedding vector.

[0121] Specifically, cross-attention mechanism: Use the request embedding vector as the Query, and the charging pile embedding vector as the Key and Value; Calculate the attention weights: Measure the matching degree between the request and each charging pile through dot product or additive attention; Output the context vector: Weightedly aggregate the charging pile features to enhance the perception of the global charging pile state.

[0122] Optionally, implementation of the mask layer. Hard constraint filtering: Charging type matching: If the request requires fast charging, mask the charging piles that do not support fast charging; Status filtering: Exclude the charging piles with "fault" or "current utilization rate exceeding 90%"; Waiting time limit: If the total waiting time of the vehicles in line for a charging pile is greater than the user's tolerance time, mask the charging pile. Dynamic mask generation: Generate a boolean mask matrix according to the real-time charging pile status during each batch of inferences to mask invalid options.

[0123] Furthermore, output layer and decision-making: Map the attention output to a charging pile matching score through a fully connected layer; Apply Softmax to generate a probability distribution, and select the charging pile with the highest probability as the allocation result.

[0124] Specifically, the model training implementation includes: input data organization; batch construction: each training sample contains a set of charging pile data and an allocation request to be assigned; positive samples: the charging piles actually assigned for this request are marked as 1, and the others are marked as 0; negative samples: randomly sample unassigned charging piles (invalid items need to be filtered using a mask); loss function design. The loss term for high-priority requests is multiplied by a weight (such as 1.5 times) to enhance the allocation accuracy of the model.

[0125] Furthermore, the optimization strategy includes: optimizer: use AdamW (Adam with weight decay), and set the initial learning rate to 0.0003; learning rate scheduling: adopt the cosine annealing strategy, and reset the learning rate every 50 epochs to avoid local optima.

[0126] Optionally, regularization: L2 weight decay (coefficient 0.0001) is used to prevent overfitting. Dropout (probability 0.3) is applied to the embedding layer and the fully connected layer.

[0127] Furthermore, the training process includes: load the preprocessed charging pile and request datasets, and divide them into a training set and a validation set in a ratio of 8:2; randomly sample 100 requests and their associated charging pile data in each batch; perform forward propagation to calculate the matching score, and calculate the loss after applying the dynamic mask; perform backpropagation to update the parameters, and verify the model performance every 100 batches.

[0128] Optionally, for model lightweighting: use TensorRT or ONNX to convert the model into a high-performance inference format. Pre-compute the embedding vectors (features that do not change in real time) for the charging pile encoding tower to reduce the online computation amount. Real-time data processing: receive the charging pile status stream data (such as updated every second) through Kafka or RabbitMQ. Store the request data in the Redis cache to ensure low-latency reading. Online inference service: deploy a Flask / FastAPI service to receive the charging pile and request data and return the allocation plan. Use GPU to accelerate batch inference and support processing thousands of requests per second.

[0129] Specifically, the collaborative operation of the robotic arm and the guide rail. Integrated design of the robotic arm and the guide rail: The upper guide rail of the mobile charging vehicle and the robotic arm adopt an integrated structure. The base of the robotic arm is driven by a servo motor and can slide horizontally along the upper guide rail. An encoder is built into the upper guide rail to provide real-time feedback on the position of the robotic arm (accuracy ±1mm) to ensure the synchronization of the movement of the robotic arm and the positioning of the charging pile.

[0130] Furthermore, the charging pile grasping process: The robotic arm slides along the guide rail to the position of the target charging pile, locates the grooves on both sides of the charging pile through a laser sensor (accuracy ±2mm), clamps it, and then moves the entire charging pile along the guide rail to the battery replacement station. The upper guide rail only bears the movement and positioning of the charging pile, and the robotic arm is responsible for operations such as clamping, disassembling, and replacing. The two cooperate through the central controller to avoid positioning deviations caused by separation.

[0131] Optionally, the adsorption distance between the magnetic adsorption plate and the rear cover is fixed at 1 cm (calibrated in real time by a laser rangefinder). After adsorption, it is translated along the guide rail to the temporary storage area to avoid collision with the charging pile body.

[0132] Specifically, the battery replacement process is as follows:

[0133] Fixing the charging pile and disassembling the rear cover: The robotic arm holds the charging pile to the guide rail station, and the bottom buckle automatically locks the position of the charging pile. The mechanical screwdriver disassembles the rear cover screws. After completion, the electromagnet is triggered to adsorb the rear cover and move it out.

[0134] Battery replacement operation: The robotic claw extends into the charging pile. Through 3D vision, it locates the battery slot, holds the old battery and pulls it out along the guide rail, and places it in the recycling bin. The spare battery compartment of the mobile charging vehicle ejects a fully charged battery. The robotic arm holds the new battery and aligns it with the slot, and inserts the fully charged battery into the charging pile.

[0135] Resetting the rear cover and returning the charging pile: The magnetic adsorption plate adsorbs the rear cover and moves it back above the charging pile. The robotic arm assists in aligning the screw holes, and the screwdriver tightens the screws.

[0136] The charging pile releases the guide rail lock, and the robotic arm pushes it back to its original position. The accuracy of the return position is confirmed through visual detection of the marker.

[0137] Optionally, the object recognition algorithm uses the YOLOv8 network, which balances real-time performance and accuracy and is applicable to object recognition of mobile charging piles, screws, batteries, etc. in a dynamic environment. The CSPDarknet53 is used to extract image features, and the PANet is combined for multi-scale feature fusion to improve the detection ability for small objects (such as screws). The YOLOv8 network is pre-trained with labeled data. To ensure the recognition accuracy of the screws, a large number of images of the screw head shapes are included in the labeled data to enhance the recognition accuracy of the driving slot orientation of the screws on the mobile charging pile, ensuring that the mechanical screwdriver can fit accurately with the driving slot.

[0138] The beneficial effects of the present invention are as follows:

[0139] Through the energy management model, the present invention realizes the allocation of the best mobile charging pile for users, improves the utilization rate of the charging pile, and reduces the waiting time of users; through the energy supply and demand status and environmental light intensity, it realizes different charging methods under different working conditions and reduces the operating cost; by replacing the battery during the peak electricity consumption period, it further improves the energy replenishment efficiency of the mobile charging pile and reduces the pressure of power shortage at the charging station during the peak period.

[0140] In this specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other.

[0141] In this article, specific examples are used to illustrate the principles and implementation modes of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation modes and application scopes. In summary, the content of this specification should not be construed as a limitation on the present invention.

Claims

1. A charging station energy management method, characterized in that: include: Collect the power information of each mobile charging pile in all charging stations to obtain the charging pile data; Collect users' electricity applications and reservations to obtain request data to be allocated; The charging pile data and the request data to be allocated are input into the pre-trained energy management model to generate a plan and obtain the charging pile allocation plan; Determine the mobile charging pile in the charging pile allocation plan as the waiting charging pile, and guide the user to the waiting charging pile for charging or provide the user with the target charging pile location information of the waiting charging pile through the mobile terminal; The training process of the energy management model includes: Collect the charging pile data and the request data to be allocated of the target research power station, and reversely formulate the charging pile allocation plan that meets each stage based on the charging pile data and the request data to be allocated; Use the Z-Score formula to standardize the charging pile data; Perform one-hot encoding on the charging type data and real-time status to obtain category features; Euclidean distance calculation is performed based on the electric pile location data and the user location data to obtain location features; Use Min-Max to normalize the request data to be allocated; Build an original model with a dual-tower encoder and criss-cross attention mechanism; Construct loss function; The electric pile encoding tower of the original model is used to encode the standardized electric pile data, category features, and location features to obtain the electric pile embedding vector; The request encoding tower of the original model is used to encode the standardized request data to be assigned to obtain the request embedding vector; The request embedding vector is used as the query, the electric pile embedding vector is used as the key and value, and the attention weight is calculated through the cross attention mechanism to obtain the context vector; Use the mask layer of the original model to hard-constrain and dynamically mask the context vector to obtain the mask output; Use the fully connected layer and Softmax of the original model to map the mask output and calculate the probability distribution to obtain the allocation matrix; Calculate the loss function based on the allocation matrix and the established charging pile allocation plan to obtain the predicted loss value; The original model is optimized using the optimizer according to the predicted loss value, and the process returns to step "using the electric pile encoding tower of the original model to encode the standardized electric pile data, category features, and position features to obtain the electric pile embedding vector" to obtain the trained energy management model.

2. A charging station energy management method according to claim 1, characterized in that: Also includes: According to the electric pile data, statistics are made on the proportion of the mobile charging piles in use in each of the charging piles to obtain the energy supply and demand status; Using a mobile charging vehicle to transport the mobile charging piles in all the charging stations whose power levels are lower than a normal power threshold back to the charging base; When the ambient light intensity exceeds a preset intensity benchmark and the energy supply and demand state is in a low-peak period, the mobile charging pile is charged by using the power grid and the photovoltaic panels in the charging base; When the ambient light intensity is lower than a preset intensity benchmark or the energy supply and demand state is at a peak period, the mobile charging pile is charged using the power grid; When the energy supply and demand status is during the peak period of electricity consumption, if the proportion of the mobile charging piles in the charging station with power below the normal power threshold exceeds the warning ratio, the spare batteries with 100% power in the charging base are transported to the charging station by the mobile charging vehicle, and the battery replacement equipment on the mobile charging vehicle is used to replace the spare batteries with the mobile charging piles with power below the normal power threshold.

3. A charging station energy management method according to claim 1, characterized in that: The charging pile data includes: power data, estimated power consumption of current vehicle charging, number of vehicles in queue, scheduled power consumption of vehicles in queue, charging pile location data, charging type data, real-time status, rates and historical utilization.

4. A charging station energy management method according to claim 2, characterized in that: The battery replacement equipment includes: a mechanical claw, a mechanical screwdriver, a camera, an upper guide rail, a lower guide rail and a magnetic suction plate; The upper guide rail is arranged on the roof of the mobile charging vehicle; the mechanical claw and the mechanical screwdriver are respectively arranged on the upper guide rail; the lower guide rail is arranged on the bottom side of the cargo box of the mobile charging vehicle; and the camera is fixed on the top inner side of the cargo box.

5. A charging station energy management method according to claim 2, characterized in that: When the mobile charging pile adopts a ternary lithium battery, the normal power threshold is 20% to 80%; when the mobile charging pile adopts a lithium iron phosphate battery, the normal power threshold is 20% to 90%.

6. A charging station energy management method according to claim 2, characterized in that: The calculation formula of the preset strength reference is: ; in, is the preset strength reference; is the rated charging power of a single mobile charging pile; is the energy conversion efficiency of the photovoltaic panel; is the effective lighting area of ​​the photovoltaic panel; is the loss coefficient.

7. A charging station energy management method according to claim 1, characterized in that: The request data to be allocated includes: reservation time, reservation power consumption, user location data, charging type requirement, battery capacity, priority data and waiting tolerance data.

8. A charging station energy management method according to claim 4, characterized in that: When the energy supply and demand state is at a peak period of electricity consumption, if the proportion of the mobile charging piles in the charging station whose power is lower than the normal power threshold exceeds the warning ratio, the spare batteries with a power ratio of 100% in the charging base are transported to the charging station by the mobile charging vehicle, and the spare batteries are replaced with the mobile charging piles whose power is lower than the normal power threshold by the battery replacement equipment on the mobile charging vehicle, including: Using the camera to collect image data, and using an object recognition algorithm to identify the image data, to obtain spatial position information of the mobile charging pile whose battery is to be replaced; Moving the mechanical claw to a position where the upper guide rail is closest to the mobile charging pile of the battery to be replaced according to the spatial position information; Using a laser sensor to locate the grooves on both sides of the mobile charging pile, and using the mechanical claw to transfer the mobile charging pile to the replacement station of the lower guide rail through the grooves; Using an object recognition algorithm to detect the fixing screws on the back cover of the mobile charging pile to obtain screw placement information; Remove the fixing screws on the back cover using a mechanical screwdriver according to the screw placement information; Using the magnetic plate to transfer the back cover to a temporary storage area; The camera uses an object recognition algorithm to detect the battery to be replaced in the mobile charging pile to obtain battery placement information; Using the mechanical claw to transfer the battery to be replaced to the recovery area of ​​the lower rail according to the battery placement information, and installing the prepared spare battery into the mobile charging pile; After the backup battery is installed, the rear cover is installed on the mobile charging pile using the mechanical screwdriver, and the mobile charging pile is transferred to its original position using the mechanical claw.

9. A charging station energy management method according to claim 1, characterized in that: Determine the mobile charging pile in the charging pile allocation plan as the waiting charging pile, and guide the user to the waiting charging pile for charging or provide the user with the target charging pile location information of the waiting charging pile through the mobile terminal, including: The user location data and the target electric pole location information are input into the navigation software to obtain a travel route, and the travel route is updated on the user's mobile terminal.

10. A charging station energy management method according to claim 1, characterized in that: The expression of the loss function is: ; in, is the calculated value of the loss function; is the number of requests; is the number of charging piles; is the priority weight; is a dynamically generated mask; is the true label; To predict the probability of success; is the balance coefficient; is the historical utilization variance.

Images