Photovoltaic energy storage magnetic suspension power balancing method and system

Abstract

The invention belongs to the technical field of new energy and intelligent transportation, and discloses a photovoltaic energy storage magnetic suspension power balancing method and system, and the method comprises the following steps: collecting photovoltaic real-time power generation data and historical power generation data, and carrying out the preprocessing of the data to obtain a data set; extracting charging and discharging states and power output information of the energy storage system; a support vector regression model SVR is adopted to train and fit the relationship between the photovoltaic power generation power and the influence factors for the data set; dynamically optimizing a support vector regression model SVR through an adaptive multi-dimensional search artificial bee colony algorithm ABCDNDS; photovoltaic power generation power is predicted based on the optimized support vector regression model SVR, and the energy management system dynamically switches energy storage system module charging and discharging and power grid power supply control strategies according to the predicted photovoltaic power generation power. Through tight integration of software and hardware, efficient utilization of energy and maximization of transportation efficiency are achieved, and the system is suitable for various application scenes including but not limited to the fields of urban traffic, logistics transportation, industrial automation and the like.

Description

Technical Field

[0001] The present invention belongs to the field of new energy and intelligent transportation technology, and specifically relates to a photovoltaic energy storage magnetic levitation power balancing method and system. Background Art

[0002] Photovoltaic power generation, as a clean energy source, is environmentally friendly and renewable, but its efficiency is significantly affected by weather conditions and suffers from unstable power supply. If photovoltaic power stations deliver power to the grid in an unordered manner, this can cause frequent power on and off cycles, potentially impacting equipment in the photovoltaic power generation system. Maglev transportation, as an efficient mode of transportation, requires a large and stable power supply. However, existing power supply systems struggle to match the transient power demands of maglev equipment. Traditional energy storage systems also suffer from high charging and discharging losses and lack dynamic balancing strategies.

[0003] Hybrid energy storage systems with large capacity, long life, low cost, high stability and strong flexibility have gradually become an urgent pursuit of the market. How to optimize the control of hybrid energy storage systems remains an important research direction. Summary of the Invention

[0004] Purpose of the Invention: To address the aforementioned technical issues, the present invention provides a photovoltaic energy storage and magnetic levitation power balancing method and system. This optimizes energy management and control strategies, reduces power losses from direct grid power supply, and improves the efficiency of photovoltaic power generation and magnetic levitation transportation.

[0005] Technical solution: The present invention provides a photovoltaic energy storage magnetic levitation power balancing method, comprising the following steps:

[0006] (1) Collect real-time photovoltaic power generation data and historical power generation data and preprocess them to obtain a data set;

[0007] (2) Extract the charge and discharge status and power output information of the energy storage system to fully understand the current operating status of the system and provide a basis for optimizing the control strategy;

[0008] (3) The support vector regression model (SVR) is used to train the dataset to fit the relationship between photovoltaic power generation and influencing factors. The SVR model is expressed as follows:

[0009] y=f(x|C,ε,σ)

[0010] Among them, C represents the penalty coefficient, ε represents the insensitivity coefficient, and σ represents the kernel parameter;

[0011] (4) The artificial bee colony algorithm ABCDNDS with adaptive multidimensional search is used to dynamically optimize the penalty coefficient C, insensitivity coefficient ε, and kernel parameter σ in the support vector regression model SVR. By introducing adaptive mechanisms and multidimensional search strategies, the global search capability of the algorithm is enhanced, effectively preventing it from falling into local optimal solutions, thereby improving the accuracy and reliability of the optimization results.

[0012] (5) Predicting photovoltaic power generation based on the optimized support vector regression model SVR,

[0013] (6) The energy management system dynamically switches the energy storage system module charging and discharging and grid power supply control strategies based on the predicted photovoltaic power generation power. The control strategies include the following four control modes:

[0014] Mode 1: When the photovoltaic power generation power is less than the magnetic levitation demand, the energy storage is discharged and the grid power supply is enabled;

[0015] Mode 2: When the photovoltaic power generation power is greater than the magnetic levitation demand and the energy storage is sufficient, the energy storage is discharged and the grid is disconnected;

[0016] Mode 3: When the photovoltaic power generation power is less than the grid power supply capacity, only the grid power supply is enabled;

[0017] Mode 4: When the photovoltaic power generation power is greater than the magnetic levitation demand but the energy storage is insufficient, the grid charges the energy storage.

[0018] Among them, preprocessing includes data cleaning, denoising and normalization to improve data quality and accuracy, laying a solid foundation for subsequent model training and prediction

[0019] Furthermore, the mean square error (MSE) between the actual observed value and the predicted value is used as the evaluation function of the adaptive multidimensional artificial bee colony algorithm ABCDNDS to guide the algorithm to search for the optimal solution. The MSE is shown below:

[0020]

[0021] Where n is the number of samples, yi and yi* are the measured value and predicted value of the i-th sample respectively.

[0022] Furthermore, the objective function of the optimized SVR algorithm is as follows:

[0023]

[0024] Among them, w1, w2, ..., w5 are weight coefficients used to balance the importance of different objectives, T is the time range considered, L(t) is the function of grid load balance at time t, and its square term L 2 (t) is used to emphasize the importance of load balancing, η SE is the energy storage system efficiency, 1-ηSE represents energy loss, UPV is the utilization rate of photovoltaic power generation, 1-UPV is the proportion of photovoltaic power generation that is not utilized, CML is the energy consumption of the magnetic levitation equipment, and Ctotal is the total system cost.

[0025] Furthermore, the search strategy of the ABCDNDS algorithm is to adopt a multi-dimensional neighborhood search direction in the hired bee stage and a global-local hybrid search direction in the follower bee stage.

[0026] Furthermore, the SVR model uses a radial basis kernel function RBF, which is expressed as follows:

[0027]

[0028] Here, xi and xj are two points in the input space, and σ is the kernel function parameter, which controls the width of the function, that is, the range of its influence in the feature space. A larger σ parameter results in a smoother function and a stronger generalization ability of the model. A smaller σ parameter results in a more accurate fit of the model to the training data, but may lead to overfitting.

[0029] During the optimization process, the kernel function parameter σ in the SVR model refers to this kernel function parameter. By optimizing these parameters using the artificial bee colony algorithm and the adaptive artificial bee colony algorithm, we can find the optimal σ value, thereby achieving optimal performance in predicting photovoltaic power generation and energy storage status. This optimal σ value, the kernel function parameter in the SVR model mentioned above, is one of the key parameters solved during the optimization process.

[0030] Furthermore, when the following anomalies are detected: the photovoltaic power prediction error is greater than 25% for three consecutive hours; the vibration amplitude of the magnetic levitation equipment exceeds the safety threshold; emergency operations are performed: non-critical loads are cut off, and the energy storage enters the island power supply mode; the speed of the magnetic levitation equipment is limited to 50% of the rated value; the SVR model parameters are recalibrated every 30 minutes until the error is less than 15%.

[0031] The present invention also discloses a photovoltaic energy storage magnetic levitation power balancing system, comprising:

[0032] The grid interface module includes a grid interface and a grid interface controller; it is used to realize the transmission of electric energy and the reception of control commands;

[0033] The energy storage system module includes an energy storage system controller, energy storage status sensors, and an energy storage battery system. It is used to collect battery status information and control the battery charging and discharging process;

[0034] Data analysis and decision support module, including data analysis software and decision support system; used to analyze data from other modules and provide decision support, and realize information exchange through data bus or network connection;

[0035] Photovoltaic power generation module, including photovoltaic power generation module and photovoltaic relay; used for adjusting photovoltaic power generation parameters according to the instructions of decision support module;

[0036] An executive control module, including feedback loops and control signal generation, is used to adjust and optimize the operation of photovoltaic modules or other equipment;

[0037] The magnetic levitation device control module includes the magnetic levitation device; it is used to control the operating status of the magnetic levitation device, involving power drivers and sensor feedback;

[0038] Monitoring and alarm circuit module, including monitoring and alarm system; used to collect environment and equipment status information;

[0039] The sensor network module, including the weather forecast system and light intensity sensor, is used to collect environmental data and transmit it to the data analysis and decision support module via a wired network to optimize system operation.

[0040] The energy management system EMS is used to coordinate the work among various modules and ensure the energy balance among photovoltaic energy storage, grid power supply and magnetic levitation equipment.

[0041] Furthermore, the energy management system EMS includes:

[0042] Multi-source data fusion unit, integrating light sensor, temperature and humidity sensor and magnetic levitation track vibration detector;

[0043] Emergency power controller, which switches to energy storage island power supply mode within 5 seconds when the grid loses power;

[0044] The heat recovery power generation unit recovers stored waste heat through a Stirling generator to generate electricity, with a conversion efficiency of ≥12%.

[0045] Furthermore, the charging and discharging strategy of the energy storage system module includes:

[0046] When the battery temperature is greater than 40°C, the supercapacitor will take on ≥90% of the discharge task;

[0047] A deep cycle calibration is performed every morning between 1:00 and 3:00 am to charge the battery SOC from 5% to 95%.

[0048] Furthermore, the magnetic levitation equipment control module includes:

[0049] Power driver, response time ≤ 100ms;

[0050] Energy consumption feedback unit generates energy efficiency optimization curves in real time and transmits them to EMS;

[0051] The emergency brake controller activates the mechanical brake device within 3 seconds when the power supply is abnormal.

[0052] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following significant advantages: by optimizing energy management and control strategies through the intelligent energy management system (EMS), the power loss of direct power supply from the power grid is reduced, and energy utilization efficiency is improved. The system combines photovoltaic power generation and energy storage modules, which can supplement energy when photovoltaic power generation is insufficient, ensure the stable operation of the magnetic levitation equipment, and improve transportation efficiency. In addition, the system has good stability and reliability, and can monitor and adjust the operating status of each module in real time to adapt to different changes in demand. The system also has high scalability and can be applied to multiple fields such as urban transportation, logistics transportation, and industrial automation, with good economic and environmental benefits. By adopting an improved artificial bee colony algorithm and support vector regression (SVR) model, the system can more accurately predict photovoltaic power generation power, optimize the charging and discharging control strategy, and further improve the overall performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a schematic diagram of the photovoltaic energy storage magnetic levitation power balancing system of the present invention. DETAILED DESCRIPTION

[0054] The technical solution of the present invention is further described in detail below through specific embodiments.

[0055] The photovoltaic energy storage magnetic levitation power balancing method of the present invention comprises the following steps:

[0056] (1) Collect real-time photovoltaic power generation data and historical power generation data and preprocess them to obtain a data set. Preprocessing includes data cleaning, denoising, and normalization to improve data quality and accuracy, laying a solid foundation for subsequent model training and prediction;

[0057] (2) Extract the charge and discharge status and power output information of the energy storage system to fully understand the current operating status of the system and provide a basis for optimizing the control strategy;

[0058] (3) The support vector regression model (SVR) is used to train the dataset to fit the relationship between photovoltaic power generation and influencing factors; the minimum output error is obtained, that is, during the training process, the parameters of the SVR model are continuously optimized to minimize the error between the photovoltaic power generation power predicted by the model and the actual value, thereby improving the prediction performance of the model;

[0059] The SVR model is expressed as follows:

[0060] y=f(x|C,ε,σ) (1)

[0061] Among them, C represents the penalty coefficient, ε represents the insensitivity coefficient, and σ represents the kernel parameter;

[0062] (4) The artificial bee colony algorithm ABCDNDS with adaptive multidimensional search is used to dynamically optimize the penalty coefficient C, insensitivity coefficient ε, and kernel parameter σ in the support vector regression model SVR. By introducing adaptive mechanisms and multidimensional search strategies, the global search capability of the algorithm is enhanced, effectively preventing it from falling into local optimal solutions, thereby improving the accuracy and reliability of the optimization results.

[0063] (5) Predicting photovoltaic power generation based on the optimized support vector regression model SVR,

[0064] (6) The energy management system dynamically switches the energy storage system module charging and discharging and grid power supply control strategies based on the predicted photovoltaic power generation power. The control strategies include the following four control modes:

[0065] Mode 1: When the photovoltaic power generation power predicted by the optimized SVR model is less than the real-time power demand of the maglev system, the hybrid energy storage module will release electricity to supplement the power shortage. At the same time, the grid will also provide the necessary power to ensure the normal operation of the maglev system.

[0066] Mode 2: When the PV power generation predicted by the optimized SVR model exceeds the real-time power demand of the maglev system and the hybrid energy storage module has sufficient power, the hybrid energy storage module releases excess power and the grid stops supplying power, fully utilizing PV power generation and energy storage resources and reducing the burden on the grid.

[0067] Mode 3: When the PV power generation power predicted by the optimized SVR model is less than the grid’s supply capacity, the grid will only supply power to the magnetic levitation equipment module to meet its needs, while the hybrid energy storage module is on standby, ready to provide assistance when necessary;

[0068] Mode 4: When the optimized SVR model predicts that the photovoltaic power generation is greater than the real-time power demand of the magnetic levitation system, but the hybrid energy storage module is insufficient, the grid will charge the hybrid energy storage module to replenish its power, ensuring that the energy storage module can provide sufficient power support when needed.

[0069] The present invention establishes the following SVR prediction model based on the adaptive multi-dimensional artificial bee colony algorithm.

[0070] The present invention uses the radial basis kernel function as the kernel function for the SVR prediction model. Dynamic optimization is required for the penalty coefficient C, the insensitive loss coefficient ε, and the kernel parameter σ. These three parameters change dynamically and have a significant impact on the performance of the SVR prediction model. By correlating these three parameters with the location of nectar sources, solving for these three parameters translates to finding the optimal nectar source location within a population.

[0071] The evaluation function assesses the quality of the nectar source location and guides the entire population towards the optimal location. For evaluating the performance of SVR prediction models, prediction accuracy or error is often used as a criterion. This system uses the mean squared error (MSE) between the actual observed and predicted values ​​as the evaluation function for the adaptive multidimensional artificial bee colony algorithm to guide the algorithm in its search for the optimal solution. The definition of MSE is as follows:

[0072]

[0073] Where n is the number of samples, yi and yi* are the measured value and predicted value of the i-th sample respectively.

[0074] Through the above analysis, the location of the nectar source and the evaluation function are determined.

[0075] Define the objective function

[0076] The ABCDNDS algorithm is used to optimize the kernel function and penalty coefficient in the SVR model. The objective function in the SVR algorithm is:

[0077]

[0078] Where w1, w2, ..., w5 are weight coefficients used to balance the importance of different objectives. T is the time range considered. L(t) is the function of grid load balance at time t, and its square term L 2 (t) is used to emphasize the importance of load balancing. SE is the efficiency of the energy storage system, here we use 1-η SE To represent the energy loss. PV is the photovoltaic power generation utilization rate, using 1-U PV To express the proportion of photovoltaic power generation that is not utilized. ML is the energy consumption of the magnetic levitation equipment. C total is the total system cost.

[0079] When describing the adaptive multidimensional artificial bee colony algorithm for optimizing the parameters of the SVR prediction model, it is crucial to clarify the role of the objective function and how it is used within the algorithm. Objective function J is a quantitative metric used to evaluate the performance of the SVR model. It integrates multiple factors, including load balance, energy storage system efficiency, photovoltaic power generation utilization, magnetic levitation equipment energy consumption, and total system cost. The following are modifications and additions to the original steps to illustrate the application of the objective function within the algorithm.

[0080] Through the above analysis, the location of the nectar source and the evaluation function are determined. The next step is to use the adaptive multi-dimensional artificial bee colony algorithm to optimize the SVR prediction model parameters:

[0081] 1) Set the search range of the parameters {C, ε, σ} and select the training and test samples of the model;

[0082] 2) Use the objective function J as the evaluation function of the adaptive multidimensional artificial bee colony algorithm, set the population size to SN, the dimension to D, the maximum number of iterations to MaxFes, and the counter to limit;

[0083] 3). The SVR prediction model parameters C, ε, σ correspond to the position of the feasible solution xi=(x i1 ,x i2 ,x i3 )=(C i1 ,ε i2 ,σ i3 ), and initialize the position;

[0084] 4). During the bee search phase, the location of the nectar source is updated, and the objective function J is used to calculate the fitness value of the nectar source. In the above-mentioned MSE, in the adaptive multidimensional artificial bee colony algorithm, each time the nectar source location (i.e., the parameters of the SVR model) is updated, the MSE is used to calculate the fitness value of the new location. This fitness value reflects the prediction performance of the model under the current parameter settings, and the algorithm tends to select parameter combinations that can reduce the MSE. The smaller the fitness value, the better the model performance. If a better nectar source is found, the current nectar source is replaced, and the information of the optimal nectar source is recorded. This process is achieved by comparing the objective function values ​​under different parameter combinations to find the parameter combination that minimizes J;

[0085] 5) If the algorithm's stopping criteria are met (such as convergence or reaching the maximum number of iterations), proceed to step 6; otherwise, continue to step 4;

[0086] 6). Output the optimal nectar source location information, that is, the optimal parameter values ​​C, ε, σ of the SVR model, and use the optimized parameter values ​​to perform predictions on the SVR prediction model.

[0087] The penalty coefficient C, insensitive loss coefficient ε, and kernel parameter σ in the prediction model are set to [0.01, 500], [0, 1], and [0.1, 100], respectively. This setting aims to find the optimal parameter combination to improve the performance and prediction accuracy of the SVR prediction model.

[0088] During the hired bee phase, bees acquire information from multiple perspectives and are no longer restricted to a single search direction. Instead, they explore and optimize in multiple directions, targeting surrounding nectar sources with higher quality. When selecting a direction, the influence of the parameter success rate (SR) is considered. If a direction has greater potential for finding a superior nectar source, its success rate will also be higher. Therefore, this dimension is considered to have great potential for finding superior nectar sources. This strategy utilizes a multi-dimensional search direction to obtain more information and possibilities. The new search strategy is as follows:

[0089]

[0090] Among them, J represents the updated dimension, ∈[-1, 1], represents the best solution in the neighbor set of , is a randomly selected solution from the neighbor set of , and, this formula utilizes the effective information of the neighborhood, uses the best neighbor solution to guide the search, and selects more promising dimensions for update according to the success rate, further enhancing the convergence speed of the algorithm.

[0091] In the follower bee stage, the follower bee is also influenced by group learning information and social learning information in the bee colony. The information of the local neighborhood best solution and the global neighborhood best solution is used to guide the follower bee's learning in multiple dimensions. The new search formula is as follows:

[0092]

[0093] Among them, x gbest,J is: the global optimal nectar source location in the dynamic neighborhood, which represents the optimal solution searched by all bees in the current iteration, corresponding to the optimal energy allocation strategy of the photovoltaic energy storage system; best,J is the optimal nectar source location in the dynamic neighborhood, which represents the optimal solution in the local search space where the bee is currently located and is used to balance the local development capacity; φ is the local perturbation intensity coefficient, with a value range of [-1, 1], which controls the random perturbation step size of the neighborhood search; is: global guidance weight factor, ranging from [0, 1.5], which adjusts the probability of the follower bee moving to the global optimal solution; J is the energy storage module dimension index, which identifies the multidimensional coupling parameter.

[0094] In the above formula, x gbest is randomly selected from the dynamic neighborhood of (the local search space generated by the adaptive neighborhood partitioning mechanism) and represents the best solution found so far, φ, are a set of J random numbers between the range [-1, 1] and [0, 1.5]. If a nectar source is mined multiple times without being updated, that is, the trial value exceeds the predetermined threshold, then the algorithm may have fallen into a local optimal solution. In this case, the nectar source needs to be discarded and reinitialized. Although random initialization can cause the solution to randomly jump to another location, potentially finding the optimal value, this random search will disrupt the current convergence state, thereby reducing the search speed. To solve this problem, a new method based on reverse learning is proposed. Its formula is as follows:

[0095] X i,j =Ub+Lb-(r1·X i,j +r2·X best,j )(10)

[0096] Among them, Ub and Lb represent the upper and lower bounds, which are the best solutions found so far, and are two random numbers between [0,1].

[0097] The steps to obtain the target input variable groups involved in defining the objective function based on the SVR model are as follows:

[0098] Grid load balancing (LB): Smart metering devices monitor the power supply and demand of the grid in real time and calculate the difference between the two to ensure the balance of power supply and demand.

[0099] Energy Storage System Efficiency (ESE): Evaluates the performance of an energy storage system by measuring its energy conversion efficiency during charging and discharging.

[0100] Photovoltaic power utilization (PU): By monitoring the total power generation of the photovoltaic power generation modules and the actual power used by the system, the photovoltaic power generation utilization rate is calculated to optimize the coordinated operation of photovoltaic power generation and energy storage systems.

[0101] Magnetic levitation equipment energy consumption (MEC): The energy consumption monitoring system records the energy consumption data of the magnetic levitation equipment in real time and evaluates the operating efficiency of the equipment.

[0102] The calculation of the total system cost (TSC) may require a more detailed financial analysis, including energy purchase costs, equipment depreciation, operation and maintenance expenses, etc., to fully reflect the economic feasibility of the system.

[0103] The objective function is optimized using the SVR model to obtain optimized SVR model parameters obtained by the adaptive multidimensional artificial bee colony algorithm. The model parameters obtained by the adaptive multidimensional artificial bee colony algorithm are specifically:

[0104] Parameters include: regularization parameter C; kernel function parameter σ; kernel function type: radial basis kernel (RBF).

[0105] The adaptive multi-dimensional artificial bee colony algorithm simulates the foraging behavior of bees in nature to find the optimal solution. The algorithm uses three types of bees: employed bees, observer bees, and scout bees. Through an iterative process, these bees search the solution space to find the optimal combination of SVR parameters.

[0106] During the optimization process, each bee represents a candidate solution, or a set of SVR parameters. The algorithm determines which parameter combinations are best by evaluating the fitness of each candidate solution (typically based on cross-validation error). Through continuous iteration and parameter updates, the algorithm eventually converges to an optimal set of parameters.

[0107] Based on the objective function, kernel function parameters in the SVR model are obtained. Specifically, the kernel function parameters in the SVR model are: Based on the given SVR model and the overall objective function equation, in support vector regression (SVR), the kernel function is used to map the input space to a higher-dimensional feature space, in which the data may become more easily separated. The choice and parameters of the kernel function have a significant impact on the performance of the model.

[0108] The above SVR model uses the radial basis kernel function (RBF), and its kernel function can be expressed as:

[0109]

[0110] Here, xi and xj are two points in the input space, and σ is the kernel function parameter, which controls the width of the function, that is, the range of its influence in the feature space. A larger σ parameter results in a smoother function and a stronger generalization ability of the model. A smaller σ parameter results in a more accurate fit of the model to the training data, but may lead to overfitting.

[0111] During the optimization process, the kernel function parameter σ in the SVR model refers to this kernel function parameter. By optimizing these parameters using the artificial bee colony algorithm and the adaptive artificial bee colony algorithm, we can find the optimal σ value, thereby achieving optimal performance in predicting photovoltaic power generation and energy storage status. This optimal σ value, the kernel function parameter in the SVR model mentioned above, is one of the key parameters solved during the optimization process.

[0112] When the following anomalies are detected: the photovoltaic power prediction error is greater than 25% for three consecutive hours; the vibration amplitude of the magnetic levitation equipment exceeds the safety threshold; emergency operations are performed: non-critical loads are disconnected, and the energy storage enters the island power supply mode; the speed of the magnetic levitation equipment is limited to 50% of the rated value; the SVR model parameters are recalibrated every 30 minutes until the error is less than 15%.

[0113] The present invention also discloses a photovoltaic energy storage magnetic levitation power balancing system, comprising:

[0114] The grid interface module includes a grid interface and a grid interface controller; it is used to realize the transmission of electric energy and the reception of control commands;

[0115] The energy storage system module includes an energy storage system controller, energy storage status sensors, and an energy storage battery system. It is used to collect battery status information and control the battery charging and discharging process;

[0116] Data analysis and decision support module, including data analysis software and decision support system; used to analyze data from other modules and provide decision support, and realize information exchange through data bus or network connection;

[0117] Photovoltaic power generation module, including photovoltaic power generation module and photovoltaic relay; used for adjusting photovoltaic power generation parameters according to the instructions of decision support module;

[0118] An executive control module, including feedback loops and control signal generation, is used to adjust and optimize the operation of photovoltaic modules or other equipment;

[0119] The magnetic levitation device control module includes the magnetic levitation device; it is used to control the operating status of the magnetic levitation device, involving power drivers and sensor feedback;

[0120] Monitoring and alarm circuit module, including monitoring and alarm system; used to collect environment and equipment status information;

[0121] The sensor network module, including the weather forecast system and light intensity sensor, is used to collect environmental data and transmit it to the data analysis and decision support module via a wired network to optimize system operation.

[0122] The energy management system EMS is used to coordinate the work among various modules and ensure the energy balance among photovoltaic energy storage, grid power supply and magnetic levitation equipment.

[0123] The energy management system EMS includes:

[0124] Multi-source data fusion unit, integrating light sensor, temperature and humidity sensor and magnetic levitation track vibration detector;

[0125] Emergency power controller, which switches to energy storage island power supply mode within 5 seconds when the grid loses power;

[0126] The heat recovery power generation unit recovers stored waste heat through a Stirling generator to generate electricity, with a conversion efficiency of ≥12%.

[0127] The charging and discharging strategies of the hybrid energy storage module include:

[0128] When the battery temperature is greater than 40°C, the supercapacitor will take on ≥90% of the discharge task;

[0129] A deep cycle calibration is performed every morning between 1:00 and 3:00 am to charge the battery SOC from 5% to 95%.

[0130] The magnetic levitation equipment control module includes:

[0131] Power driver, response time ≤ 100ms;

[0132] Energy consumption feedback unit generates energy efficiency optimization curves in real time and transmits them to EMS;

[0133] The present invention reduces energy losses from direct grid power supply and improves the efficiency of photovoltaic power generation and magnetic levitation transportation by optimizing energy management and control strategies. An intelligent energy management system coordinates the operation of various modules. The photovoltaic power generation module converts solar energy into electricity, which is then converted into a form suitable for grid use via an inverter. The energy storage module stores excess photovoltaic power for future use. The grid power supply module connects to the grid via a substation, enabling bidirectional flow of electricity. The magnetic levitation equipment module utilizes the principle of magnetism to achieve contactless transportation and reduce energy losses. The intelligent energy management system (EMS) is the core of this system. Using an improved artificial bee colony algorithm and control strategy, it monitors and regulates photovoltaic power generation, energy storage, and grid power supply in real time to ensure optimal system operation. Furthermore, the EMS can dynamically adjust the power supply based on the operational requirements of the magnetic levitation equipment to achieve the highest transportation efficiency. Through the tight integration of hardware and software, the system achieves efficient energy utilization and maximizes transportation efficiency, making it suitable for a variety of application scenarios, including but not limited to urban transportation, logistics, and industrial automation.

Claims

1. A photovoltaic energy storage magnetic levitation power balancing method, characterized in that: The following steps are involved: (1) Collect real-time photovoltaic power generation data and historical power generation data and preprocess them to obtain a data set; (2) Extracting the charge and discharge status and power output information of the energy storage system; (3) The support vector regression model (SVR) is used to train the dataset to fit the relationship between photovoltaic power generation and influencing factors. The SVR model is expressed as follows: y=f(x|C,ε,σ) Among them, C represents the penalty coefficient, ε represents the insensitivity coefficient, and σ represents the kernel parameter; (4) Dynamically optimize the penalty coefficient C, insensitivity coefficient ε and kernel parameter σ in the support vector regression model SVR through the adaptive multidimensional search artificial bee colony algorithm ABCDNDS; (5) Predicting photovoltaic power generation based on the optimized support vector regression model SVR; (6) The energy management system dynamically switches the energy storage system module charging and discharging and grid power supply control strategies based on the predicted photovoltaic power generation power. The control strategies include the following four control modes: Mode 1: When the photovoltaic power generation power is less than the magnetic levitation demand, the energy storage is discharged and the grid power supply is enabled; Mode 2: When the photovoltaic power generation power is greater than the magnetic levitation demand and the energy storage is sufficient, the energy storage is discharged and the grid is disconnected; Mode 3: When the photovoltaic power generation power is less than the grid power supply capacity, only the grid power supply is enabled; Mode 4: When the photovoltaic power generation power is greater than the magnetic levitation demand but the energy storage is insufficient, the grid charges the energy storage.

2. The method according to claim 1, characterized in that The mean square error (MSE) between the actual observed value and the predicted value is used as the evaluation function of the adaptive multidimensional artificial bee colony algorithm ABCDNDS to guide the algorithm to search for the optimal solution. The MSE is shown below: Among them, n is the number of samples, yi and yi * are the measured value and predicted value of the i-th sample respectively.

3. The method according to claim 1, characterized in that The objective function of the optimized support vector regression model SVR algorithm is as follows: Among them, w1, w2, ..., w5 are weight coefficients used to balance the importance of different objectives, T is the time range considered, L(t) is the function of grid load balance at time t, and its square term L 2 (t) is used to emphasize the importance of load balancing, η SE is the energy storage system efficiency, 1-η SE Indicates energy loss, U PV is the photovoltaic power generation utilization rate, 1-U PV Indicates the proportion of unused photovoltaic power generation, C ML is the energy consumption of the magnetic levitation equipment, C total is the total system cost.

4. The system according to claim 1, wherein: The search strategy of the artificial bee colony algorithm ABCDNDS is to adopt a multi-dimensional neighborhood search direction in the employed bee stage and a global-local hybrid search direction in the follower bee stage.

5. The method according to claim 1, characterized in that The SVR model uses the radial basis kernel function RBF, and its expression is: Among them, xi and xj are two points in the input space, σ is the parameter of the kernel function, which controls the width of the function, that is, the range of influence in the feature space.

6. The method according to claim 1, characterized in that When the following anomalies are detected: the photovoltaic power prediction error is greater than 25% for three consecutive hours; the vibration amplitude of the magnetic levitation equipment exceeds the safety threshold; emergency operations are performed: non-critical loads are disconnected, and the energy storage enters the island power supply mode; the speed of the magnetic levitation equipment is limited to 50% of the rated value; the SVR model parameters are recalibrated every 30 minutes until the error is less than 15%.

7. A photovoltaic energy storage magnetic levitation power balancing system, used to implement the method according to any one of claims 1 to 6, characterized in that: include: A grid interface module, comprising a grid interface and a grid interface controller; Energy storage system module, including energy storage system controller, energy storage status sensor, and energy storage battery system; Data analysis and decision support module, including data analysis software and decision support system; Photovoltaic power generation module, including photovoltaic power generation module and photovoltaic relay; Executive control module, including feedback loop and control signal generation; A magnetic levitation device control module, including a magnetic levitation device; Monitoring and alarm circuit module, including monitoring and alarm system; Sensor network modules, including weather prediction systems and light intensity sensors; Energy Management System EMS.

8. The photovoltaic energy storage magnetic levitation power balancing system according to claim 7 is characterized in that: The energy management system EMS includes: Multi-source data fusion unit, integrating light sensor, temperature and humidity sensor and magnetic levitation track vibration detector; Emergency power controller, which switches to energy storage island power supply mode within 5 seconds when the grid loses power; The heat recovery power generation unit recovers stored waste heat through a Stirling generator to generate electricity, with a conversion efficiency of ≥12%.

9. The photovoltaic energy storage magnetic levitation power balancing system according to claim 7, characterized in that: The charging and discharging strategies of the energy storage system module include: When the battery temperature is greater than 40°C, the supercapacitor will take on ≥90% of the discharge task; A deep cycle calibration is performed every morning between 1:00 and 3:00 am to charge the battery SOC from 5% to 95%.

10. The photovoltaic energy storage magnetic levitation power balancing system according to claim 7, characterized in that: The magnetic levitation equipment control module includes: Power driver, response time ≤ 100ms; Energy consumption feedback unit generates energy efficiency optimization curves in real time and transmits them to EMS; The emergency brake controller activates the mechanical brake device within 3 seconds when the power supply is abnormal.

Images