Distributed source load system for accessing electric vehicle to rural micro-energy network and prediction method

Abstract

The invention belongs to the technical field of micro-energy network source load systems and prediction, and particularly relates to a distributed source load system for accessing an electric vehicle to a rural micro-energy network and a prediction method. The system comprises an electric vehicle intelligent monitoring integrated device, a direct current charging bus, an energy storage device, a charging booth roof photovoltaic device and a distributed roof photovoltaic device. The method aims at serving three to five households, and efficient utilization and balanced distribution of electric energy are achieved by establishing an intelligent grid in a small area. According to the invention, the rural energy structure can be optimized, the energy utilization efficiency is improved, the wide application of renewable energy sources is promoted, and the modernization process of rural area energy management is promoted. And after the electric vehicle is randomly accessed to the rural micro-energy network, the source load is accurately managed through an intelligent prediction method, so that the reliability and economy of rural power supply are ensured, the dependence on a traditional power grid can be effectively reduced, and the application and popularization of sustainable energy are promoted.

Description

Technical Field

[0001] The present invention belongs to the field of micro-energy grid source-load system and prediction technology, and in particular relates to a distributed source-load system and prediction method for connecting electric vehicles to a rural micro-energy grid. Background Art

[0002] In the context of global energy transformation and sustainable development, rural areas, as an important area of ​​energy utilization and management, urgently need innovation in their energy supply model. Traditional rural power grid systems usually rely on long-distance power transmission, and their energy supply is unstable and costly. With the popularization of electric vehicles and the advancement of photovoltaic technology, micro-energy grids in rural areas have become a key solution to solve energy shortages and improve energy self-sufficiency.

[0003] Micro-grid systems achieve local production, storage and consumption of energy by integrating distributed energy resources such as photovoltaic power generation, energy storage equipment and electric vehicle charging facilities. This system can significantly reduce dependence on external power grids, improve energy efficiency, and support intelligent charging and discharging management of electric vehicles. Photovoltaic power generation, as a green energy source, can effectively alleviate the impact of traditional energy on the environment, but its output is greatly affected by weather changes. Therefore, in micro-grid systems, it is necessary to accurately manage and optimize the relationship between photovoltaic power generation, energy storage and loads to achieve efficient energy utilization.

[0004] Intelligent prediction technology is crucial in the application of micro-energy grids. By analyzing historical data and real-time data, intelligent prediction methods can help the system accurately predict the charging needs of electric vehicles, photovoltaic power generation, and household load changes. This prediction capability can guide the system to automatically adjust the charging strategy and optimize the configuration of energy storage and power generation resources, thereby improving the stability and economy of the system. The long short-term memory network extended model xLSTM performs well in processing time series data and can effectively cope with the complex energy demand patterns and dynamic changes in micro-energy grids.

[0005] In view of this, those skilled in the art have been continuously conducting research and development of new topics. Summary of the invention

[0006] In view of the shortcomings of the above-mentioned prior art, the present invention provides a distributed source-load system and prediction method for electric vehicles connected to rural micro-energy networks. The purpose is to optimize the rural energy structure and improve energy utilization efficiency. After electric vehicles are randomly connected to the rural micro-energy network, the source-load is accurately managed through an intelligent prediction method to ensure the reliability and economy of rural power supply.

[0007] The technical solution adopted by the present invention to achieve the above-mentioned purpose is:

[0008] A distributed source-load system for connecting electric vehicles to a rural micro-energy network, comprising: an electric vehicle intelligent monitoring integrated device, a DC charging bus, an energy storage device, a charging kiosk photovoltaic power plant and a distributed roof photovoltaic power plant; wherein, after the electric vehicle is connected to the charging pile, it is connected to the DC charging bus, and the DC charging bus, the energy storage device, the charging kiosk photovoltaic power plant, the distributed roof photovoltaic power plant and the household load are respectively connected to the electric vehicle intelligent monitoring integrated device; the electric vehicle intelligent monitoring integrated device is connected to the system, and according to the rural micro-energy network, the status and energy requirements of each component are monitored and managed, and through real-time multi-source data collection, the charging, discharging and energy optimization strategies are controlled according to energy requirements and photovoltaic power generation conditions; the DC charging bus is connected to the charging pile of the electric vehicle, and the DC power is processed by normalization; the energy storage device is connected to the system, and the charging and discharging strategies are adjusted through the real-time data collection of the intelligent monitoring system and the prediction of the deep learning model, and the optimization control is performed according to the energy requirements and photovoltaic power generation conditions; the charging kiosk photovoltaic power plant and the distributed roof photovoltaic power plant convert the DC power into AC power through an inverter, and the output of the inverter is supplied to the household load.

[0009] A method for predicting distributed source load of electric vehicles connected to a rural micro-energy network is implemented by using the distributed source load system of electric vehicles connected to a rural micro-energy network, comprising:

[0010] Carry out real-time multi-source data collection and data preprocessing for the distributed source-load system of rural micro-energy grid;

[0011] Use deep learning algorithms to select and train data to obtain trained deep learning data;

[0012] Use the trained deep learning data to predict source load and optimize power configuration.

[0013] Furthermore, the data collection includes historical charging mode data of electric vehicles, weather data, energy production data, and overall data of distributed sources and loads of rural micro-energy networks to which electric vehicles are connected; the data preprocessing includes using interpolation and filling methods to process missing data values ​​and eliminate abnormal data points caused by equipment failure and human interference.

[0014] Furthermore, the real-time multi-source data collection and preprocessing for the rural micro-energy grid distributed source-load system refers to normalizing the overall data collected from the rural micro-energy grid distributed source-load system to which the electric vehicle is connected;

[0015] After normalization, the expression is as follows:

[0016]

[0017] In the above formula, X represents a certain characteristic value in the distributed micro-energy grid source and load data, X min Indicates the minimum value of the feature value in the data set, Xmax It represents the maximum value of the feature value in the data set, and Y represents the normalized data, whose value range is [0,1].

[0018] Furthermore, the use of deep learning algorithms to select and train data to obtain trained deep learning data to achieve optimal prediction results includes:

[0019] Use LSTM model to select and train data;

[0020] Based on the LSTM model, exponential gating is introduced in the sLSTM model;

[0021] The LSTM model adjusted by exponential gating is used to determine the quality of the prediction based on performance evaluation indicators.

[0022] Furthermore, the use of trained deep learning data to perform source-load prediction and optimize power configuration to achieve stable operation of the system and efficient energy utilization is achieved by using a DC charging bus load prediction accuracy algorithm;

[0023] The DC charging bus load prediction accuracy algorithm includes the following steps:

[0024] Step (41) predicts data based on the xLSTM model and obtains the predicted value S of the i-th source load sampling point i , the actual value A of the i-th source-load sampling point i ;

[0025] Step (42) calculates the single bus error e i , single bus error e i It is expressed as:

[0026]

[0027] Step (43) calculates the root mean square σ of all bus errors in time period t i , the expression is as follows:

[0028]

[0029] In the above formula, i is the i-th source-charge sampling point, and N is the number of samples;

[0030] Step (44) is based on the single bus error ei calculated in step (42) and step (43) and the root mean square error σ of all buses in time period t i , evaluate the prediction result accuracy P e , as shown below:

[0031]

[0032] In the above formula, T is the total number of time steps, and k is the index variable used to traverse the time steps from 1 to T.

[0033] A distributed source-load prediction device for electric vehicles connected to a rural micro-energy network, used to implement the steps of any one of the methods for predicting distributed source-loads of electric vehicles connected to a rural micro-energy network, comprising:

[0034] Data collection and data preprocessing module, used for real-time multi-source data collection and data preprocessing for rural micro-energy grid distributed source-load system;

[0035] The selection and training module is used to select and train data using a deep learning algorithm to obtain trained deep learning data;

[0036] The prediction and optimization module is used to use the trained deep learning data to predict source loads and optimize power configuration.

[0037] Furthermore, the real-time data collection and preprocessing of the historical charging and discharging and energy of electric vehicles for the distributed source-load system of the rural micro-energy grid refers to normalizing the overall data collected from the distributed source-load system of the rural micro-energy grid to which the electric vehicles are connected;

[0038] After normalization, the expression is as follows:

[0039]

[0040] In the above formula, X represents a certain characteristic value in the distributed micro-energy grid source and load data, X min Indicates the minimum value of the feature value in the data set, X max Indicates the maximum value of the feature value in the data set, and Y represents the normalized data, whose value range is [0,1];

[0041] The use of trained deep learning data to perform source load prediction and optimize power configuration is achieved by using a DC charging bus load prediction accuracy algorithm;

[0042] The DC charging bus load prediction accuracy algorithm includes the following steps:

[0043] Step (41) predicts data based on the xLSTM model and obtains the predicted value S of the i-th source load sampling point i , the actual value A of the i-th source-load sampling point i ;

[0044] Step (42) calculates the single bus error e i , single bus error e i It is expressed as:

[0045]

[0046] Step (43) calculates the root mean square σ of all bus errors in time period t i , the expression is as follows:

[0047]

[0048] In the above formula, i is the i-th source-charge sampling point, and N is the number of samples;

[0049] Step (44) is based on the single bus error e calculated in step (42) and step (43) i and the root mean square error of all buses in period t σ i , evaluate the prediction result accuracy P e , as shown below:

[0050]

[0051] In the above formula, T is the total number of time steps, and k is the index variable used to traverse the time steps from 1 to T.

[0052] A computer device comprises a storage medium, a processor and a computer program stored on the storage medium and executable on the processor, wherein when the processor executes the computer program, the steps of any one of the methods for predicting distributed source and load of an electric vehicle connected to a rural micro-energy network are implemented.

[0053] A computer storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of any one of the methods for predicting distributed source and load of an electric vehicle connected to a rural micro-energy grid are implemented.

[0054] The present invention has the following beneficial effects and advantages:

[0055] The present invention provides a distributed source-load system and prediction method for connecting electric vehicles to rural micro-energy networks, which realizes energy management and prediction optimization through intelligent means, and has important practical application value and market prospects. The system of the present invention can not only optimize the rural energy structure, but also improve the efficiency of energy utilization, promote the widespread application of renewable energy, and promote the modernization of energy management in rural areas. After randomly connecting electric vehicles to the rural micro-energy network, the source and load are accurately managed through intelligent prediction methods, ensuring the reliability and economy of rural power supply.

[0056] The present invention provides a distributed source-load system and prediction method for electric vehicles to access a rural micro-energy network, wherein the system mainly includes an integrated intelligent monitoring device for electric vehicles, a DC charging bus, an energy storage device, photovoltaic power generation on the top of a charging kiosk, distributed rooftop photovoltaic power generation and household loads, thereby establishing an efficient distributed source-load system for a rural micro-energy network. By using an intelligent prediction method, especially a deep learning algorithm based on an extended LSTM model, the access and energy management strategies of electric vehicles are optimized, thereby improving the operating efficiency and reliability of the entire system. Through prediction and optimization, the system maximizes the use of photovoltaic power generation, reduces dependence on traditional power grids, and enhances the stability and self-sufficiency of power supply in rural areas. In addition, through real-time data collection and dynamic adjustment strategies, the system of the present invention can also effectively balance energy supply and demand, reduce energy waste, and significantly improve economic benefits.

[0057] The intelligent monitoring system in the present invention optimizes charging strategies and energy configuration through real-time data collection and deep learning model prediction to meet the ever-changing energy needs. It can not only improve the stability and reliability of rural micro-energy grids, but also effectively reduce dependence on traditional power grids and promote the application and popularization of sustainable energy.

[0058] The present invention significantly improves the intelligent management and prediction optimization capabilities of the distributed source-load system of the rural micro-energy network by adding key components and steps such as an integrated intelligent monitoring device for electric vehicles, a deep learning algorithm based on an extended LSTM model, and an intelligent prediction method. Specifically, the integrated intelligent monitoring device for electric vehicles realizes real-time monitoring of the charging status and demand of electric vehicles, optimizes the access timing and method of electric vehicles, and ensures the coordination between electric vehicles and the micro-energy network system. The deep learning algorithm based on the extended LSTM model can accurately predict the charging demand of electric vehicles, photovoltaic power generation, and household load changes, and the intelligent prediction method dynamically adjusts the charging strategy, the charging and discharging operation of energy storage equipment, and the load distribution, effectively balancing energy supply and demand and improving energy utilization efficiency. In addition, through intelligent prediction and dynamic adjustment strategies, the system maximizes the use of photovoltaic power generation, reduces dependence on traditional power grids, enhances the stability and self-sufficiency of power supply in rural areas, significantly improves economic benefits, and promotes the widespread application of renewable energy. Compared with the prior art, the present invention optimizes the energy management and prediction process, solves the coordination problem between electric vehicle access, photovoltaic power generation, and energy storage equipment, improves the intelligence, stability, and economy of the system, and provides strong technical support for the modernization process of rural energy management. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0060] Figure 1This is a structural diagram of a distributed source-load system of a rural micro-energy network to which an electric vehicle is connected according to the present invention;

[0061] Figure 2 It is a system block diagram of the electric vehicle intelligent monitoring integrated device of the present invention;

[0062] Figure 3 It is a flow chart of the prediction method of the electric vehicle access to the distributed source-load system of the rural micro-energy network of the present invention;

[0063] Figure 4 This is a flow chart of the DC charging bus load prediction accuracy of the present invention;

[0064] Figure 5 It is a flow chart of training LSTM model of the present invention;

[0065] Figure 6 It is a performance evaluation index structure diagram of the present invention. DETAILED DESCRIPTION

[0066] In order to more clearly understand the above-mentioned objectives, features and advantages of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0067] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0068] Refer to the following Figure 1-Figure 6 The technical solutions of some embodiments of the present invention are described.

[0069] Example 1

[0070] The present invention provides an embodiment, which is a distributed source-load system for connecting electric vehicles to a rural micro-energy grid. Figure 1 As shown, Figure 1 This is a structural diagram of a distributed source-load system of a rural micro-energy network to which an electric vehicle is connected according to the present invention.

[0071] Figure 1 This is a structural diagram of a rural micro-energy grid distributed source-load system in which an electric vehicle is connected to the present invention. Figure 1As shown, the present invention discloses a rural micro-energy grid distributed source-load system connected to an electric vehicle, and the rural micro-energy grid distributed source-load system connected to the electric vehicle mainly includes: an electric vehicle intelligent monitoring integrated device, a DC charging bus, an energy storage device, a charging kiosk roof photovoltaic, a distributed roof photovoltaic and a household load. After the electric vehicle is randomly connected to the charging pile, it is connected to the DC charging bus. The DC charging bus, the energy storage device, the charging kiosk roof photovoltaic, the distributed roof photovoltaic and the household load (such as the use of household electricity such as electric vehicles, lighting, refrigerators, televisions, air conditioners, stoves, etc.) are respectively connected to the electric vehicle intelligent monitoring integrated device. The electric vehicle intelligent monitoring integrated device is connected to each component through multiple interfaces. The electric vehicle interface is connected to the electric vehicle, the DC charging bus interface is connected to the DC charging bus, the energy storage device interface is connected to the energy storage device, the charging kiosk roof photovoltaic and the distributed roof photovoltaic interface are connected to the photovoltaic power generation system, and the household load interface is connected to household appliances. The household appliances include: electric vehicles, lighting, refrigerators, televisions, air conditioners, stoves, etc.

[0072] The electric vehicle intelligent monitoring integrated device is connected to the control center of the entire system, responsible for monitoring and managing the status and energy requirements of each component, and controlling the charging, discharging and energy optimization strategies through real-time data acquisition. The DC charging bus is connected to the charging piles that are randomly connected to the electric vehicle, processing high-power DC power to ensure a fast and efficient charging process. The energy storage device is connected to the system as a key component of energy storage and balance. The charging and discharging strategy is adjusted through the electric vehicle intelligent monitoring system, and optimized control is performed according to energy demand and photovoltaic power generation. The photovoltaic power on the top of the charging kiosk and the distributed roof photovoltaic power convert DC power into AC power through an inverter. The output of the inverter can supply household loads to ensure that the power requirements of various parts of the system are met. The household load is connected to the AC side of the system and receives the AC power converted by the inverter. The household power consumption is predicted through the intelligent monitoring system management, and the power in the photovoltaic power generation and energy storage system is used preferentially to reduce dependence on the power grid.

[0073] Example 2

[0074] The present invention further provides an embodiment, which is a distributed source-load system for connecting electric vehicles to a rural micro-energy network, wherein the electric vehicle intelligent monitoring integrated device is as follows: Figure 2 As shown, Figure 2 It is a system block diagram of the electric vehicle intelligent monitoring integrated device of the present invention.

[0075] The electric vehicle intelligent monitoring integrated device includes a hardware part and a software part. The hardware part is composed of a sensor module, a processing unit, a communication module, a battery management system and a display interface, and the software part is responsible for data collection, processing, analysis and interaction with external systems.

[0076] Rural micro-grid distributed source-load systems usually use distributed rooftop photovoltaics as a key source of electricity. The system is mainly composed of three to five households, forming a sub-unit, and realizes the distribution and balance of electricity by establishing a smart grid in a small area. As the core component of the system, the rooftop photovoltaic array captures and converts the energy directly from the sun, converts direct current into alternating current through an inverter to supply various electricity needs of residents' households, and achieves electricity self-sufficiency and supply and demand balance on a small scale.

[0077] Example 3

[0078] The present invention further provides an embodiment, which is a method for predicting the distributed source and load of electric vehicles connected to a rural micro-energy network. Figure 3 As shown, Figure 3 It is a flow chart of the prediction method of the electric vehicle access to the distributed source-load system of the rural micro-energy network of the present invention.

[0079] The present invention provides a distributed source-load prediction method for a rural micro-energy network to which an electric vehicle is connected, comprising the following steps:

[0080] Step 1. Use multi-source data collection and processing technology to build a distributed source-load system for connecting electric vehicles to the rural micro-energy grid: clarify the needs, complete planning and design, and build an efficient and intelligent rural micro-energy grid infrastructure.

[0081] Step 2. For the rural micro-energy grid distributed source-load system, real-time multi-source data collection and data preprocessing are performed to determine the data source. The data source includes historical charging mode data of electric vehicles, weather data, energy production data, and the overall data of the distributed source-load of the rural micro-energy grid to which the electric vehicle is connected. The data preprocessing includes using interpolation and filling methods to process missing data values ​​and eliminate abnormal data points caused by equipment failure and human interference. The overall data collected from the distributed source-load of the rural micro-energy grid to which the electric vehicle is connected is normalized for later analysis and modeling.

[0082] The overall data collected from the distributed source and load of the rural micro-energy network connected to the electric vehicle is normalized, and its expression is as follows:

[0083]

[0084] In the above formula, X represents a certain characteristic value in the distributed micro-energy grid source and load data, X min Indicates the minimum value of the feature value in the data set, X max It represents the maximum value of the feature value in the data set, and Y represents the normalized data, whose value range is [0,1].

[0085] Step 3. Use the deep learning algorithm to select and train the overall data of the distributed source and load of the rural micro-energy network connected to the electric vehicle, and finally obtain the trained deep learning data to achieve the best prediction effect. Specifically, the overall data of the distributed source and load of the rural micro-energy network connected to the electric vehicle in step 2 is analyzed by selecting a deep learning algorithm. Based on the randomness and complexity of electric vehicle access, the deep learning algorithm training data is selected.

[0086] First, based on the randomness and complexity of electric vehicle access, a deep learning algorithm suitable for time series data prediction is selected, and a long short-term memory network (LSTM) is used. In order to improve the prediction accuracy of distributed source and load in rural micro-energy grids connected to electric vehicles, an extended LSTM model (xLSTM) is further used to enhance its adaptability and accuracy in dealing with complex source and load change patterns. Then, the xLSTM model is trained using the preprocessed data set in step 2, the model parameters are adjusted, and the model performance is evaluated through the validation set, and finally an optimized deep learning model is obtained. The model can accurately predict the charging demand of electric vehicles, photovoltaic power generation, energy storage equipment status, and household load changes, provide support for smart charging and energy management, and achieve the optimal prediction effect of the system.

[0087] The details are as follows:

[0088] 1. Use the LSTM model to select and train data;

[0089] Second, based on the LSTM model, exponential gating is introduced into the sLSTM model. After the exponential gating calculation, the weighted gating output is obtained, which enables the model to more effectively selectively retain or discard information according to the time step, thereby improving the accuracy and robustness of the prediction. The weighted gating output will be used as the core input of the LSTM model to further improve the prediction ability of the distributed source and load of electric vehicles connected to the rural micro-energy grid.

[0090] 3. Determine the quality of the prediction effect based on the performance evaluation index; the prediction performance of the LSTM model adjusted by exponential gating will be objectively verified by the performance evaluation index. If the evaluation index shows that the prediction effect is poor, it is necessary to further optimize the model or adjust the gating mechanism.

[0091] like Figure 5 As shown, Figure 5 The present invention is a flow chart of training an LSTM model. The present invention also discloses a training LSTM algorithm model. When training the LSTM model, the specific steps are as follows:

[0092] Step (31) initializes the parameters of the LSTM model, including the weight vector w z 、w i 、w f 、wo and the bias term b z , b i , b f , b o . Where w z The output weight matrix, w, represents the unit input i represents the output weight matrix of the input gate, w f represents the output weight matrix of the forget gate, w o represents the output weight matrix of the output gate, b z represents the bias vector of the unit input, b i represents the bias vector of the input gate, b f represents the bias vector of the forget gate, b o Represents the bias vector of the output gate.

[0093] Step (32) constructs an LSTM unit, which includes four key components: forget gate, input gate, unit state and output gate, which determines the flow of information through interaction and captures the long-term dependencies in sequence data.

[0094] Step (33) Input the source load data set x at the current time after the electric vehicle is dynamically connected t and the previous hidden state h t-1 .

[0095] Step (34) Calculate the output f of the forget gate t , used to determine the cell state from the previous t-1 How much information is discarded in the equation is expressed as follows:

[0096]

[0097] In the formula, represents the output weight matrix transposed matrix of the forget gate, r f represents the weight of the forget gate, b f represents the bias vector of the forget gate, and σ represents the activation function sigmoid, that is,

[0098] Step (35) calculates the output z of the unit input t , determine what needs to be added to the cell state c t The amount of new information is expressed as follows:

[0099]

[0100] In the formula, The output weight matrix transposed matrix represents the unit input, r z represents the weight of the unit input, b z represents the bias vector of the unit input, It represents the activation function tanh.

[0101] Step (36) calculates the output i of the input gate t , the input gate controls what new information should be added to the cell state c t In the equation, the updated amount is determined as follows:

[0102]

[0103] In the formula, represents the output weight matrix transposed matrix of the input gate, r i represents the weight of the input gate, b i Represents the bias vector of the input gate.

[0104] Step (37) updates the cell state based on (34), step (35) and step (36), using the output of the forget gate, the cell input and the input gate, and determines how much of the previous state c to retain. t-1 , and add new information, the expression is as follows:

[0105] c t =f t c t-1 +i t z t

[0106] In the above formula, c t represents the cell state, z t The output of the unit input, i t represents the output of the input gate, f t Represents the output of the forget gate.

[0107] Step (38) calculates the output o of the output gate t , determine how to use the updated cell state c t Generate the hidden state h of the current time step t , which is expressed as follows:

[0108]

[0109] h t =o t ψ(c t )

[0110] In the formula, represents the output weight matrix transposed matrix of the output gate, r o represents the weight of the output gate, b o represents the bias vector of the output gate, and ψ represents the activation function tanh, which is used to standardize or compress the unit state, otherwise it will be unbounded.

[0111] The present invention discloses a distributed source-load system and prediction method for electric vehicles connected to a rural micro-energy network, including introducing exponential gating and appropriate normalization and stabilization technology based on an LSTM model, modifying the storage structure of the LSTM, and obtaining an sLSTM with a mixture of scalar storage, scalar update and new storage and a fully parallelized mLSTM with matrix storage and covariance update rules. The LSTM model is extended and integrated into a residual block to generate an xLSTM block, and the xLSTM blocks are stacked into an xLSTM architecture.

[0112] In the present invention, the steps of "forward transfer of sLSTM" and "forward transfer of mLSTM" are closely related to the previous step of "steps of LSTM model during training". In the previous step, the LSTM model learned the rules in the time series data through training, and optimized the structure and parameters of the extended LSTM (xLSTM) model to improve the adaptability to the complex source-load change pattern in the distributed source-load system of the rural micro-energy grid connected to electric vehicles. At this time, the trained model provides the initial conditions for this step, and the trained model includes the storage rules and update mechanism of sLSTM and mLSTM. In this step, sLSTM and mLSTM perform forward transfer respectively, calculate and output the prediction results through input data and historical states. sLSTM plays a role in processing small or frequently changing data through scalar storage and update rules; while mLSTM is more suitable for processing large-scale data and complex time series through matrix storage and covariance update rules. These forward transfer operations rely on the model parameters and structure obtained by the previous step of training, and integrate them into the residual block to form an optimized xLSTM architecture, thereby improving prediction accuracy and computational efficiency. Therefore, the training results of the previous step directly affect the forward transfer process of this step, ensuring that the model can make predictions efficiently in practical applications.

[0113] To achieve the above object, the present invention provides a distributed source-load system and prediction method for electric vehicles connected to a rural micro-energy grid. The method comprises: based on an LSTM model, exponential gating is introduced into the sLSTM model, and normalization and stabilization processing are combined to enhance the ability of LSTM to modify storage decisions. The forward transfer of the sLSTM is:

[0114] c t =f t c t-1 +i t z t

[0115] n t =f t n t-1 +i t

[0116]

[0117] In the above formula, exp(·) represents the exponential function. In the above formula, n t Represents the normalization factor at the current time t, ensuring that the model information transmission process remains smooth and stable, n t-1 It represents the normalization factor of the previous moment t-1, which is used to accumulate and transfer to the current moment t, reflecting the cumulative characteristics of the normalized weights between time steps. z represents the bias vector of the unit input, b i represents the bias vector of the input gate, b f represents the bias vector of the forget gate, b o Represents the bias vector of the output gate.

[0118] Adding a bias term to the LSTM gating technique, propagated to the new architecture, the exponential activation function may lead to large values, causing overflow. Therefore, adding state m t To stabilize the gate:

[0119] m t =max(log(f t )+m t-1 ,log(i t ))

[0120]

[0121] f t '=exp(log(f t )+m t-1 -m t )

[0122] In the above formula, i′ t represents the normalized input gate output, f ′ t Represents the normalized forget gate output.

[0123] To increase the storage capacity of the LSTM, its storage unit is increased from scalar c∈R to matrix C∈R d×d , retrieval is performed by matrix multiplication. At time t, to store a pair of vectors, key k t ∈R d Sum value v t ∈R d . Then at time t+τ, v t The value should be determined by the query vector q t+τ ∈R d Retrieve. The covariance update rule for storing key-value pairs is:

[0124]

[0125] In the above formula, Ct Indicates the unit state, C t-1 represents the unit state at the previous moment, k T t Represents the vector matrix at the current moment, v t A vector representing the current time.

[0126] The mLSTM forward pass is:

[0127]

[0128] n t =f t n t-1 +i t k t

[0129]

[0130] q t =W q x t +b q

[0131]

[0132] v t =W v x t +b v

[0133]

[0134]

[0135] o t =σ(W o x t +b o )

[0136] In the above formula, ⊙ represents the same or same operation, k t represents the key input at time step t, b k Indicates k t The bias vector, q t represents the gate variable at time step t, b q represents the query input bias vector, b v Represents the value input bias vector, b i represents the bias vector of the input gate, b f Represents the bias vector of the forget gate, w v Represents the value input weight matrix, w q represents the query input weight matrix, w k represents the key input weight matrix, wT i represents the input gate weight vector transpose, w T f represents the transpose of the forget gate weight vector, w o represents the transpose of the output gate weight vector, and σ represents the activation function sigmoid.

[0137] The prediction effect in step 3 is determined based on the performance evaluation index, such as Figure 6 As shown, Figure 6 The present invention discloses a method for determining the quality of prediction effect based on performance evaluation indicators, wherein the performance evaluation indicators include deterministic prediction results and probabilistic prediction results, and the specific steps are as follows:

[0138] First, the LSTM model is used for data selection and training to learn valuable patterns and features from the original data, including selecting appropriate data and adjusting the model weights through the back-propagation algorithm. Then, exponential gating is introduced in the sLSTM model. This gating mechanism can control the flow of information more flexibly, has better performance in information forgetting and updating, and helps to stabilize the calculation process. Then, the performance evaluation indicators such as MAE, RMSE, and MAPE are used to determine the quality of the prediction effect. These indicators provide a feedback mechanism for the first two steps. If the prediction effect is not good, the data selection strategy or model structure parameters can be adjusted accordingly. In this series of processes, stacking xLSTM blocks into an xLSTM architecture can enhance the model's feature learning ability, allowing it to learn more abstract and representative features at different levels, and at the same time help to deal with long-term dependencies in long sequence data. However, the stacking architecture will directly affect the prediction performance of the model, and its rationality needs to be verified by performance evaluation indicators. If there are problems such as overfitting, the stacking architecture can be adjusted and optimized based on the performance evaluation results.

[0139] The performance evaluation indicators of the deterministic prediction in step (1) include mean absolute error, root mean square error and mean absolute percentage error, which are expressed as follows:

[0140]

[0141] In the above formula, MAE is the mean absolute error, RMSE is the root mean square error, MAPE is the mean absolute percentage error, y i is the actual value, is the predicted value, N is the number of samples, and i is the i-th sample in the data set.

[0142] The performance evaluation index of the probability prediction in step (2) mainly includes Pinball loss, Winkler score and continuous grade probability score, which are expressed as follows:

[0143]

[0144] In the above formula, q is the probability value, is the predicted value of the q quantile, y i is the true value of the i-th sample, α is the confidence level, β is the width of the confidence interval, and U i is the upper limit of the confidence interval, L i is the lower limit of the confidence interval, F i (z) is the predicted probability distribution, Y and Y' are derived from the cumulative distribution function F i (z) is an independent random variable sampled from the distribution of To find the expected function.

[0145] Step 4. Use the trained deep learning model data to perform source-load prediction and optimize power configuration to achieve stable system operation and efficient energy utilization.

[0146] Using the deep learning data trained in step 3, the real-time and accurate prediction of electric vehicles after they are dynamically connected to the distributed source-load system of the rural micro-energy grid is realized. The system adjusts the charging strategy according to the prediction results and real-time data, optimizes the distributed source-load configuration of the micro-energy grid, and completes the stable operation and efficient energy utilization of the system.

[0147] like Figure 4 As shown, Figure 4 The present invention discloses a DC charging bus load prediction accuracy rate flow chart. The present invention discloses a DC charging bus load prediction accuracy rate algorithm. The DC charging bus load prediction accuracy rate specific steps are as follows:

[0148] Step (1) According to the xLSTM model prediction data, obtain the predicted value S of the i-th source load sampling point i , the actual value A of the i-th source-load sampling point i .

[0149] Step (2) Calculate the single bus error e i , single bus error e i It can be expressed as:

[0150]

[0151] Step (3) Calculate the root mean square σ of all bus errors in time period t i , which is expressed as follows:

[0152]

[0153] In the above formula, i is the i-th source-charge sampling point, and N is the number of samples;

[0154] Step (4) The single bus error e calculated according to step 2 and step 3 i and the root mean square error of all buses in period t σ i , evaluate the accuracy of the prediction results, using P e It is expressed as follows:

[0155]

[0156] In the above formula, T is the total number of time steps, and k is the index variable used to traverse the time steps from 1 to T.

[0157] Example 4

[0158] The present invention further provides an embodiment, which is a distributed source-load prediction device for electric vehicles connected to a rural micro-energy network, comprising:

[0159] Building modules for building a distributed source-load system for electric vehicles to access rural micro-energy grids using multi-source data acquisition and processing technology;

[0160] The data collection and data preprocessing module is used to collect and preprocess real-time multi-source data for the distributed source-load system of the rural micro-energy grid and determine the data source;

[0161] The selection and training module is used to select and train data using deep learning algorithms to obtain trained deep learning data and achieve optimal prediction results;

[0162] The prediction and optimization module is used to use the trained deep learning data to predict the source and load and optimize the power configuration to achieve stable operation of the system and efficient energy utilization;

[0163] The device is used to implement the steps of a method for predicting distributed source and load of electric vehicles connected to a rural micro-energy network as described in Example 3.

[0164] Example 5

[0165] Based on the same inventive concept, an embodiment of the present invention further provides a computer device, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor. When the processor executes the computer program, the steps of the method for predicting the access of an electric vehicle to a distributed source-load system of a rural micro-energy grid as described in Example 3 are implemented.

[0166] Example 6

[0167] Based on the same inventive concept, an embodiment of the present invention further provides a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a method for predicting the access of an electric vehicle to a distributed source-load system of a rural micro-energy grid as described in Example 3 are implemented.

[0168] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0169] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0170] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0171] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0172] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A distributed source-load system for connecting electric vehicles to a rural micro-energy grid, characterized by: include: An integrated intelligent monitoring device for electric vehicles, a DC charging bus, an energy storage device, photovoltaic power generation on the top of a charging kiosk and distributed rooftop photovoltaic power generation; wherein, after the electric vehicle is connected to the charging pile, it is connected to the DC charging bus, and the DC charging bus, the energy storage device, photovoltaic power generation on the top of the charging kiosk, the distributed rooftop photovoltaic power generation and the household load are respectively connected to the integrated intelligent monitoring device for electric vehicles; the integrated intelligent monitoring device for electric vehicles is connected to the system, and the status and energy requirements of each component are monitored and managed according to the rural micro-energy network. Through real-time multi-source data acquisition, the intelligent monitoring system of the electric vehicle controls the charging, discharging and energy optimization strategies according to the energy demand and photovoltaic power generation situation; the DC charging bus is connected to the charging pile of the electric vehicle, and the DC power is processed by normalization; the energy storage device is connected to the system, and the charging and discharging strategies are adjusted through the real-time data acquisition and deep learning model prediction of the intelligent monitoring system, and the optimization control is performed according to the energy demand and photovoltaic power generation situation; the photovoltaic power generation on the top of the charging kiosk and the distributed rooftop photovoltaic power generation convert the DC power into AC power through the inverter, and the output of the inverter is supplied to the household load.

2. A method for predicting the distributed source and load of electric vehicles connected to a rural micro-energy grid, which is implemented by using the distributed source and load system of electric vehicles connected to a rural micro-energy grid as claimed in claim 1, and is characterized by: include: Carry out real-time multi-source data collection and data preprocessing for the distributed source-load system of rural micro-energy grid; Use deep learning algorithms to select and train data to obtain trained deep learning data; Use the trained deep learning data to predict source load and optimize power configuration.

3. A method for predicting distributed source and load of electric vehicles connected to a rural micro-energy grid according to claim 2, characterized in that: The data collection includes: historical charging mode data of electric vehicles, weather data, energy production data and overall data of distributed sources and loads of rural micro-energy networks to which electric vehicles are connected; the data preprocessing includes: using interpolation and filling methods to process missing data values ​​and eliminate abnormal data points caused by equipment failure and human interference.

4. A method for predicting distributed source and load of electric vehicles connected to a rural micro-energy grid according to claim 2, characterized in that: The real-time multi-source data collection and preprocessing for the rural micro-energy grid distributed source-load system refers to normalizing the overall data collected from the rural micro-energy grid distributed source-load system to which the electric vehicle is connected; After normalization, the expression is as follows: In the above formula, X represents a certain characteristic value in the distributed micro-energy grid source and load data, X min Indicates the minimum value of the feature value in the data set, X max It represents the maximum value of the feature value in the data set, and Y represents the normalized data, whose value range is [0,1].

5. The method for predicting distributed source and load of electric vehicles connected to a rural micro-energy grid according to claim 2 is characterized by: The method of selecting and training data using a deep learning algorithm to obtain trained deep learning data includes: Use LSTM model to select and train data; Based on the LSTM model, exponential gating is introduced in the sLSTM model; The LSTM model adjusted by exponential gating is used to determine the quality of the prediction based on performance evaluation indicators.

6. A method for predicting distributed source and load of electric vehicles connected to a rural micro-energy grid according to claim 2, characterized in that: The use of trained deep learning data to perform source load prediction and optimize power configuration is achieved by using a DC charging bus load prediction accuracy algorithm; The DC charging bus load prediction accuracy algorithm includes the following steps: Step (41) predicts data based on the xLSTM model and obtains the predicted value S of the i-th source load sampling point i , the actual value A of the i-th source-load sampling point i ; Step (42) calculates the single bus error e i , single bus error e i It is expressed as: Step (43) calculates the root mean square σ of all bus errors in time period t i , the expression is as follows: In the above formula, i is the i-th source-charge sampling point, and N is the number of samples; Step (44) is based on the single bus error e calculated in step (42) and step (43) i and the root mean square error of all buses in period t σ i , evaluate the prediction result accuracy P e , as shown below: In the above formula, T is the total number of time steps, and k is the index variable used to traverse the time steps from 1 to T.

7. A distributed source-load prediction device for electric vehicles connected to a rural micro-energy grid, used to implement the steps of a distributed source-load prediction method for electric vehicles connected to a rural micro-energy grid as described in any one of claims 2 to 6, characterized in that: include: Data collection and data preprocessing module, used for real-time multi-source data collection and data preprocessing for rural micro-energy grid distributed source-load system; The selection and training module is used to select and train data using a deep learning algorithm to obtain trained deep learning data; The prediction and optimization module is used to use the trained deep learning data to predict source loads and optimize power configuration.

8. The distributed source-load prediction device for electric vehicles connected to a rural micro-energy grid according to claim 7, characterized in that: The real-time data collection and preprocessing of the historical charging and discharging and energy of electric vehicles for the distributed source-load system of the rural micro-energy grid refers to normalizing the overall data collected from the distributed source-load system of the rural micro-energy grid to which the electric vehicles are connected; After normalization, the expression is as follows: In the above formula, X represents a certain characteristic value in the distributed micro-energy grid source and load data, X min Indicates the minimum value of the feature value in the data set, X max Indicates the maximum value of the feature value in the data set, and Y represents the normalized data, whose value range is [0,1]; The use of trained deep learning data to perform source load prediction and optimize power configuration is achieved by using a DC charging bus load prediction accuracy algorithm; The DC charging bus load prediction accuracy algorithm includes the following steps: Step (41) predicts data based on the xLSTM model and obtains the predicted value S of the i-th source load sampling point i , the actual value A of the i-th source-load sampling point i ; Step (42) calculates the single bus error e i , single bus error e i It is expressed as: Step (43) calculates the root mean square σ of all bus errors in time period t i , the expression is as follows: In the above formula, i is the i-th source-charge sampling point, and N is the number of samples; Step (44) is based on the single bus error ei calculated in step (42) and step (43) and the root mean square error σ of all buses in time period t i , evaluate the prediction result accuracy P e , as shown below: In the above formula, T is the total number of time steps, and k is the index variable used to traverse the time steps from 1 to T.

9. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the computer program, the steps of a method for predicting distributed source and load of electric vehicles connected to a rural micro-energy grid as described in any one of claims 2 to 6 are implemented.

10. A computer storage medium, characterized in that: The computer storage medium stores a computer program, and when the computer program is executed by the processor, the steps of a method for predicting distributed source and load of electric vehicles connected to a rural micro-energy grid as described in any one of claims 2 to 6 are implemented.

Images