New energy charging station equipment state monitoring and maintenance optimization method

Abstract

The invention relates to the technical field of equipment management and maintenance, and discloses a new energy charging station equipment state monitoring and maintenance optimization method, which comprises the following steps: S1, equipment data acquisition: installing various sensors and counters on equipment for acquiring operation data of the equipment in real time, the data comprises voltage, current, temperature, humidity, equipment use duration and charging times; s2, data preprocessing and noise removal: preprocessing the data acquired in the S1 by using a generative adversarial network to remove noise in the data; and S3, equipment health state prediction: predicting the data processed in the S2 based on deep reinforcement learning and a graph neural network, and generating a health prediction model. Through combination of deep reinforcement learning, a generative adversarial network, a block chain and an intelligent contract technology, accurate equipment health prediction, dynamic market pricing, automatic transaction execution and data security guarantee are realized, and the efficiency and security of equipment management and transaction are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment management and maintenance, and specifically to a method for monitoring and optimizing the equipment status of new energy charging stations. Background Art

[0002] In modern society, the use of equipment has penetrated into all aspects of our lives. From daily household appliances to industrial equipment, equipment failures can not only lead to a decline in production efficiency but also cause economic losses. Therefore, how to accurately predict the health status of equipment, avoid sudden failures, and reduce maintenance costs has become an important issue in the field of equipment management.

[0003] Most existing equipment status monitoring systems are based on traditional data collection and analysis methods. These methods usually use sensors to collect basic equipment operation data and predict equipment failures by setting fixed thresholds or regression models based on historical data. This method is effective for some simple equipment and relatively stable environments. In addition, existing technologies also have a certain market demand prediction ability, using traditional regression analysis or linear models to predict market trends based on historical data, and these technologies can provide certain support for equipment pricing.

[0004] However, there are some deficiencies in the existing technologies. First, the accuracy and robustness of equipment health prediction are poor. Especially in the face of diversified equipment and complex operating environments, traditional methods often cannot effectively process complex data, resulting in insufficient early warnings of failures. Second, most existing market demand prediction and pricing models are static and cannot make rapid adjustments in real time according to market changes, lacking flexibility in dealing with competition and demand fluctuations. In addition, existing transaction and data management methods still rely on centralized storage and manual operations, leading to risks in data security and privacy protection and also reducing the efficiency of transaction execution. Summary of the Invention

[0005] In view of the deficiencies of the existing technologies, the present invention provides a method for monitoring and optimizing the equipment status of new energy charging stations, which solves the problems of inaccurate equipment health prediction, slow market demand response, low transaction execution efficiency, and insufficient data security in the existing technologies.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A method for monitoring and optimizing the equipment status of new energy charging stations, including the following steps:

[0007] S1. Equipment data collection: Install a variety of sensors and counters on the equipment to collect the operation data of the equipment in real time, and the data includes voltage, current, temperature, humidity, equipment usage duration, and charging times;

[0008] S2. Data preprocessing and noise removal: Use a generative adversarial network to preprocess the data collected in S1 and remove the noise in the data;

[0009] S3. Equipment health status prediction: Based on deep reinforcement learning and graph neural network, predict the data processed in S2 and generate a health prediction model;

[0010] S4. Fault mode identification: Use multi-task learning to train the health prediction model generated in S3 to achieve the identification of equipment fault modes;

[0011] S5. Maintenance decision optimization: Analyze the equipment fault modes identified in S4 based on the optimal control theory and optimize the maintenance decisions of the equipment;

[0012] S6. Market demand prediction: Dynamically predict the market demand of the equipment based on a generative adversarial network;

[0013] S7. Intelligent pricing: Optimally adjust the transaction price of the equipment according to the market demand data generated in S6 and the equipment health status information generated in S3;

[0014] S8. Blockchain recording: Store all the data and transaction records from S1 to S7 through blockchain technology to ensure the immutability of the data;

[0015] S9. Smart contract execution: Automatically execute the relevant terms of the transaction based on blockchain technology, including payment, delivery, and quality assurance.

[0016] Preferably, the equipment data collection includes:

[0017] Install voltage sensors, current sensors, temperature sensors, humidity sensors, and counters;

[0018] Real-time monitor and obtain the key parameters during the operation of the equipment, and the data includes voltage, current, temperature, humidity, equipment usage duration, and charging times.

[0019] Preferably, the data preprocessing and noise removal include:

[0020] Use the generator of the generative adversarial network to generate clean training data, the discriminator discriminates the authenticity of the data, and optimizes the generator and discriminator according to the training feedback;

[0021] Input the collected equipment operation data into the generative adversarial network for training, remove the noise and abnormal data therein, and generate a high-quality data set.

[0022] Preferably, the equipment health status prediction includes:

[0023] Based on the deep reinforcement learning algorithm, train a health prediction model on the basis of the device historical data to predict the future health status change of the device;

[0024] Use a graph neural network to perform multi-dimensional modeling on the operating parameters of the device to improve the accuracy and robustness of health status prediction.

[0025] Preferably, the graph neural network is carried out through the following steps:

[0026] Represent the device health data generated by S3 as a graph structure, where each node represents a device status parameter and each edge represents the relationship between device statuses;

[0027] Use graph convolution operations to propagate device status information to quickly extract deep features of the device health status.

[0028] Preferably, the fault mode recognition includes:

[0029] Use a multi-task learning model to simultaneously perform device health status prediction, fault mode recognition, and device fault repair requirement analysis;

[0030] Share the network layers of multiple tasks during training to optimize the classification accuracy of fault modes and improve the accuracy of device health status recognition.

[0031] Preferably, the maintenance decision optimization includes:

[0032] Based on the optimal control theory, define the objective function of device maintenance and optimize it. The objective function includes maintenance cost, fault risk, and device downtime;

[0033] By solving the optimal control problem, calculate the optimal maintenance strategy to reduce the unplanned downtime of the device and lower the maintenance cost.

[0034] Preferably, the market demand prediction includes:

[0035] Use a generative adversarial network to generate data related to market demand, and use a discriminator to judge the authenticity of the generated data;

[0036] Through adversarial training, generate a prediction model that better conforms to the actual market demand fluctuations and dynamically adjust the market demand prediction of the device.

[0037] Preferably, the intelligent pricing includes:

[0038] Based on the health status and market demand information of the device, use a game theory model to adjust the device price;

[0039] Determine the optimal transaction prices for both the supply and demand sides by calculating the Nash equilibrium point, ensuring market supply and demand balance, and maximizing the platform's revenue.

[0040] Preferably, the blockchain record includes:

[0041] All device operation data, health prediction data, maintenance decisions, and transaction records involved in S1 to S9 are distributedly stored through blockchain technology;

[0042] Use smart contracts to automatically execute device transaction terms, ensuring the transparency and immutability of the transaction process, and providing full traceability of device transactions.

[0043] The present invention provides a method for monitoring and optimizing the status of new energy charging station equipment. It has the following beneficial effects:

[0044] 1. The present invention combines deep reinforcement learning and graph neural networks to optimize the prediction of device health status, significantly improving the accuracy and stability of the prediction. Compared with traditional methods, it can detect potential device failures earlier, avoiding many device damages caused by inaccurate predictions.

[0045] 2. The present invention predicts market demand through a generative adversarial network and uses a game theory model for intelligent pricing, thus successfully achieving an accurate response to market dynamics, ensuring that the pricing strategy can be adjusted in real time, thereby improving the platform's market competitiveness and making up for the deficiency of the traditional pricing model's slow response to market changes.

[0046] 3. By adopting blockchain technology to record all transaction information, the present invention makes device transaction data completely transparent and immutable. This measure ensures the security and trust of transactions, far exceeding the existing centralized storage methods, and avoiding potential risks of data leakage and tampering.

[0047] 4. By using smart contracts to automatically execute device transaction terms, the present invention eliminates the need for manual intervention and improves transaction efficiency. Compared with traditional manual operations, smart contracts make transactions smoother and more transparent, reducing human errors and delays. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the specification of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0050] Please refer to the attached Figure 1 , the embodiment of the present invention provides a method for monitoring and optimizing the maintenance of the new energy charging station equipment, including the following steps:

[0051] S1. Equipment data acquisition: Install a variety of sensors and counters on the equipment to collect the operation data of the equipment in real time. The data includes voltage, current, temperature, humidity, equipment usage duration, and charging times;

[0052] S1 is to obtain various operation data of the equipment in real time by installing sensors and counters. These data include voltage, current, temperature, humidity, equipment usage duration, charging times, etc. The selection and layout of these sensors determine the accuracy and effect of subsequent health prediction, fault mode identification, and maintenance decision-making. To ensure the comprehensiveness and accuracy of data acquisition, a variety of sensors are used to monitor different operation parameters of the equipment.

[0053] First of all, voltage sensors, current sensors, temperature sensors, humidity sensors, and counters are installed on the equipment. These sensors will measure the voltage (V), current (I), temperature (T), humidity (H), equipment usage duration (t), and charging times (C) of the equipment in real time. All these sensors transmit the data to the cloud platform through the Internet of Things technology to ensure that the equipment operation data can be collected, stored, and used for subsequent analysis in real time.

[0054] In this embodiment, to ensure the accuracy of the equipment monitoring system, the design and layout of each sensor are optimized to ensure that they can efficiently reflect the real-time changes of the equipment state. The following are the specific functions of each sensor and among them:

[0055] Voltage sensor (V): Used to measure the voltage of the equipment to monitor the power supply status of the equipment. Voltage data has a direct impact on the working state of the equipment and is a basic parameter for judging whether the equipment is operating normally.

[0056] Voltage acquisition formula:

[0057] V = f voltage (sensor);

[0058] Among them, f voltage (sensor) represents the voltage of the equipment measured by the voltage sensor.

[0059] Current sensor (I): Measures the current of the equipment and reflects the degree of power consumption of the equipment. Through real-time current data, it can be inferred whether the equipment is overloaded or in an abnormal working state.

[0060] Current acquisition formula:

[0061] I = fcurrent (sensor);

[0062] Among them, f current (sensor) represents the device current measured by the current sensor.

[0063] Temperature sensor (T): Measures the temperature change during the operation of the device. An excessively high device temperature may cause overheating failures or performance degradation. Therefore, real-time temperature monitoring is crucial for device health.

[0064] Temperature acquisition formula:

[0065] T = f temperature (sensor);

[0066] Among them, f temperature (sensor) represents the device temperature measured by the temperature sensor.

[0067] Humidity sensor (H): Measures the humidity change in the device's surrounding environment. Especially in extreme weather conditions, excessive humidity may cause performance degradation or corrosion of the device.

[0068] Humidity acquisition formula:

[0069] H = f humidity (sensor);

[0070] Among them, f humidity (sensor) represents the humidity of the device's environment measured by the humidity sensor.

[0071] Counters (t and C): Record the usage duration and charging times of the device. The usage duration of the device reflects its usage cycle, and the charging times are one of the important indicators for evaluating the device's health status.

[0072] Device usage duration acquisition formula:

[0073] t = f time (counter);

[0074] Among them, f time (counter) represents the usage duration of the device recorded by the counter.

[0075] Charging times acquisition formula:

[0076] C = f charge (counter);

[0077] Among them, f charge (counter) represents the charging times of the device recorded by the counter.

[0078] Data from all these sensors is uploaded to the platform in real time through a high-speed communication channel, and the platform will process, store, and analyze it. By regularly or real-time updating the data, it is ensured that the platform can issue an alarm immediately when potential problems occur in the device.

[0079] As an option, in some embodiments, the acquisition frequency of the sensors may be dynamically adjusted according to the actual operating conditions of the device. For example, when the charging station is operating at a high load, the data acquisition frequency can be increased to enable more precise monitoring of the device status. Different types of devices may adopt acquisition strategies with different frequencies, avoiding data redundancy caused by over-acquisition while ensuring that important data can reflect the operating conditions of the device in real time.

[0080] In a possible implementation, the sensors of the device are calibrated regularly to ensure the accuracy of the measurement results. Voltage sensors, current sensors, temperature sensors, etc. are compared with standard devices to ensure that the working accuracy of the sensors meets the requirements. In addition, humidity sensors and counters can also be set with a regular calibration function to reduce measurement errors caused by hardware aging.

[0081] Through this device data acquisition step, the present invention can comprehensively obtain the operating data of the device in real time, and this process can effectively improve the accuracy of device health management.

[0082] S2. Data preprocessing and noise removal: Use a generative adversarial network to preprocess the collected data and remove the noise in the data;

[0083] S2 accomplishes this task by using a generative adversarial network (GAN). A generative adversarial network is a deep learning algorithm that can effectively remove the noise in the data and generate high-quality training data during the adversarial training process of the generator and the discriminator.

[0084] In this embodiment, the generative adversarial network (GAN) includes two main components: a generator and a discriminator. The goal of the generator is to generate clean samples that conform to the real data distribution, while the goal of the discriminator is to distinguish between the generated fake data and the real data. The generator and the discriminator are continuously optimized through adversarial training until the samples generated by the generator are real enough that the discriminator cannot distinguish them.

[0085] The generator and the discriminator continuously play a game during the training process. Specifically, the generator generates new data based on random noise or the input sensor data, and the discriminator evaluates the generated data and the real data to determine whether it is real data. Through continuous iteration, the generator will learn how to generate cleaner data that is increasingly close to the real data.

[0086] To achieve this process, the training objectives of the generator and discriminator are quantified through loss functions. The loss function of the generator represents the difference between the generated data and the real data, while the loss function of the discriminator measures its error in distinguishing between real and fake data.

[0087] The loss function of the generator can be expressed as:

[0088]

[0089] The loss function of the discriminator is:

[0090]

[0091] Where: L G represents the loss of the generator, and the goal is to minimize the difference between the generated data and the real data; L D represents the loss of the discriminator, and the goal is to maximize the correct judgment of the discriminator on real data and generated data; D(G(z)) represents the evaluation value of the discriminator on the data generated by the generator, and the generator hopes to maximize this value to make the generated data as "real" as possible; D(x) represents the evaluation value of the discriminator on real data, and the discriminator hopes to maximize this value to accurately judge real data; p z represents the noise distribution, from which the generator samples to generate new data samples; p data represents the distribution of real data; z is a noise variable sampled from the noise distribution p z ; G(z) is the output of the generator, representing the fake data generated given the noise variable z; x is a real data sample, representing the data sampled from the real data distribution p data .

[0092] As an option, the training processes of the generator and discriminator usually adopt the gradient descent method for optimization. The goal of the generator is to improve the data generation ability by minimizing the difference between the generated data and the real data, while the goal of the discriminator is to improve the judgment ability by maximizing its correct judgment on real data. During the training process, the generator and discriminator play against each other in the parameter space, continuously improving their respective abilities.

[0093] Specifically, during the training process, by continuously adjusting the parameters of the generator and discriminator, the generator will learn how to generate clean data that conforms to the real data distribution, while the discriminator will learn how to more accurately distinguish between generated data and real data. After multiple training epochs, the generator will be able to generate increasingly real data.

[0094] In the data preprocessing stage, the original sensor data is input into the generative adversarial network. Through training, the generator can extract noise and outliers from the original data and generate a denoised and clean version. This clean version serves as the input data for subsequent health prediction and fault mode recognition, ensuring the efficiency and accuracy of the subsequent models.

[0095] For example, during the operation of a device, the current sensor may be subject to electromagnetic interference and produce inaccurate readings. Through the generative adversarial network, the generator will learn to identify such inaccurate readings and repair or replace them with data that more conforms to the normal operating state of the device. The finally output data will remove noise and outliers without losing key information, thereby improving the accuracy of subsequent analysis and decision-making.

[0096] The goal during the training process is to continuously optimize the loss function so that the generator and discriminator achieve the best performance when generating and identifying data.

[0097] After the data is cleaned and denoised, it can provide more reliable data support for subsequent health prediction, fault mode recognition, and maintenance decision-making. Through adversarial training, the generator and discriminator not only remove the noise in the data but also retain the key features of the device state. This process improves the accuracy, robustness, and adaptability of the system, effectively ensuring the smooth progress of subsequent technical steps.

[0098] S3. Prediction of the device health status: Based on deep reinforcement learning and graph neural network, the processed data is predicted to generate a health prediction model;

[0099] Through accurate prediction of the health status, S3 can identify potential fault risks of the device in advance, thereby providing a scientific basis for subsequent fault mode recognition and maintenance decision-making.

[0100] In this embodiment, the prediction of the device health status adopts a technical method that combines deep reinforcement learning (DRL) and graph neural network (GNN). Deep reinforcement learning (DRL) trains a model that can evaluate the current health status of the device and predict future changes through interaction with the device's historical data. The graph neural network (GNN) is used to process the complex relationships between the device's various operating parameters to further improve the accuracy and robustness of the prediction results.

[0101] Deep Reinforcement Learning (DRL) plays a crucial role in equipment health prediction. Specifically, the DRL model is trained using historical equipment health data and evaluates the current health status of the equipment. As the training progresses, the model can predict the possible future health status of the equipment by learning from the historical behavior of the equipment. The key in this process is to evaluate the health values in different states through the Q-learning algorithm and make corresponding decisions based on the changes in the states.

[0102] The update formula of deep reinforcement learning is as follows:

[0103]

[0104] Where: Q(s t ,a t ) represents the value of taking action a t in state s t , representing the estimated value of the health status in the current state of the equipment; r t represents the immediate reward obtained by performing action a t at the current time t, measuring the improvement or deterioration of the equipment health status; γ is the discount factor, used to adjust the model's emphasis on future rewards; α is the learning rate, indicating the degree of learning of the model for new data in each update; represents the maximum value action that may be taken in the next state s t+1 , and the model will optimize the current decision based on this value; s t represents the current equipment state; a t represents the action taken in state s t .

[0105] Through the above Q-learning formula, the deep reinforcement learning model can evaluate the health status of the equipment based on the historical data and current state of the equipment, and make predictions about the future health status. During the equipment health prediction process, the model can evaluate the possible future health changes according to different health states and provide a basis for subsequent maintenance decisions.

[0106] The role of Graph Neural Network (GNN) in equipment health prediction is to model the various operating parameters of the equipment and the relationships between them through a graph structure. Different operating parameters of the equipment (such as voltage, current, temperature, etc.) are represented as nodes in the graph, and the relationships between these nodes are connected by edges in the graph. In the graph neural network, information propagates between each node through graph convolution operations, so as to extract the deep features of the equipment health status.

[0107] The calculation formula of the graph neural network is as follows:

[0108]

[0109] in: represents the feature representation of node v at the kth layer; W( k ) is the weight matrix of the kth layer, which determines the way node information is propagated in the graph convolution; N(v) represents the set of neighbor nodes of node v, reflecting the relationship between various state parameters of the device; for example, there may be a direct dependency between voltage and current, and between temperature and humidity; b( k 0 is the bias term of the kth layer, which is used to adjust the output value of the node; σ is the activation function, which can increase the nonlinear ability of the model; Represents the feature representation of node u at the kth layer.

[0110] Through graph convolution operations, GNN can efficiently propagate information and extract deep features between various parameters of the equipment. Through multi-layer graph convolution, GNN can effectively capture the complex dependencies between various operating parameters of the equipment, thereby improving the accuracy and robustness of health status prediction.

[0111] As an option, in order to further improve the accuracy of equipment health status prediction, deep reinforcement learning and graph neural network training can be carried out in a joint training mode. In this joint training mode, deep reinforcement learning evaluates the current health status of the equipment through the Q-learning algorithm, while the graph neural network captures the deep relationship between the various status parameters of the equipment through graph convolution operations. The two can be jointly trained by sharing certain network layers, thereby sharing information during the training process and improving the overall prediction accuracy of the model.

[0112] Specifically, joint training can be carried out in the following ways:

[0113] In each training cycle, the deep reinforcement learning model evaluates the current device state through the Q-learning algorithm and predicts the future state based on the change in health state.

[0114] At the same time, the graph neural network extracts the relationship between the various state parameters of the equipment through multi-layer graph convolution operations, and feeds this relationship information back to the deep reinforcement learning model to help it better predict the health status.

[0115] Through this joint training method, the system can take into account both the time series changes of equipment status (through deep reinforcement learning) and the dependencies between various equipment parameters (through graph neural networks) in equipment health prediction.

[0116] Through the joint training of deep reinforcement learning and graph neural networks, the model can not only make health predictions based on the historical data of the equipment, but also explore the deep correlations between the various operating parameters of the equipment, thereby improving the accuracy and robustness of the prediction.

[0117] Through the interactive learning of historical device data and the current device state, deep reinforcement learning continuously optimizes the prediction model, enabling effective evaluation of the current health state of the device and prediction of future health changes. Graph neural networks further improve the accuracy and robustness of prediction by modeling the complex relationships between various operating parameters of the device.

[0118] S4. Fault mode recognition: Use multi-task learning to train the generated health prediction model to achieve the recognition of device fault modes;

[0119] S4 identifies possible fault modes of the device through further analysis of the device health state. The core goal of fault mode recognition is to identify potential fault types of the device through data analysis and provide a basis for subsequent maintenance decisions. This process uses the multi-task learning (MTL) method to simultaneously perform device health state prediction, fault mode recognition, and fault repair requirement analysis.

[0120] In this embodiment, when performing fault mode recognition, a multi-task learning (MTL) model is used to simultaneously train the three tasks of health state prediction, fault mode recognition, and fault repair requirement analysis. Multi-task learning can effectively improve the generalization ability of the model by sharing a part of the network layer of the model and can improve the accuracy of fault recognition by sharing information.

[0121] Specifically, the processing process of step S4 includes the following aspects:

[0122] Health state prediction: In step S3, based on the joint training of deep reinforcement learning and graph neural networks, we have obtained the prediction of the device health state. Step S4 uses these prediction results as inputs to further identify fault modes.

[0123] Fault mode recognition: Based on the health state prediction and device historical data, use a multi-task learning model to perform fault mode recognition.

[0124] Fault repair requirement analysis: By analyzing the current fault mode and health state of the device, judge whether fault repair is required and the urgency of the repair.

[0125] In the multi-task learning framework, health state prediction and fault mode recognition share the same network layer, and the accuracy and efficiency of the model are improved through shared feature learning.

[0126] To train the multi-task learning model, it is first necessary to define the total loss function of multi-task learning. For each task i, there is a corresponding loss function L i (θ). The total multi-task loss function L MTL is represented by the following formula:

[0127]

[0128] Where: L MTL represents the total loss function of multi-task learning; L i (θ) represents the loss function of the i-th task; θ represents the parameters of the model; λ is the regularization parameter used to control the contribution of the loss functions of different tasks to the total loss function; w i and w j respectively represent the shared weights of the i-th and j-th tasks; ∥w i -w j ∥ 2 represents the difference in weights between shared tasks, and the regularization term is used to constrain the weight differences between different tasks; m represents the number of shared task weights, that is, how many tasks in the model share certain network layers.

[0129] Fault mode recognition is the core task in this step. This task identifies potential fault modes by learning the historical health status data of the device and the current health status prediction. In practical applications, fault modes may include problems such as device overheating, overload, sensor failure, electrical short circuit, etc.

[0130] Specifically, the fault mode recognition process first processes the time series data of the device health status through a multi-task learning model. The model not only predicts the current health status of the device but also identifies which state changes may indicate the occurrence of a fault through learning the historical data of the device. When the health status of the device exceeds the normal range, the model will output the corresponding fault mode.

[0131] For example, in power equipment, if the current and temperature of the device increase simultaneously, it may cause device overload or electrical short circuit. In fault mode recognition, the model will automatically identify the fault mode based on this characteristic information and combined with the health status prediction.

[0132] After fault mode recognition, the next step is to analyze whether fault repair is needed and judge the urgency of the repair. The analysis of fault repair requirements needs to be determined based on the current fault mode, health status of the device, and the historical maintenance records of the device.

[0133] As an option, the analysis of fault repair requirements can be combined with the environmental conditions of the device (such as temperature, humidity, etc.) to make a decision. Suppose the device has an overload fault in a high-temperature environment, the model may recommend immediate repair to avoid further damage.

[0134] In a possible implementation, the decision-making process for fault repair requirement analysis can also consider the life cycle of the device. If the device has been running for a long time and has few maintenance records, the model may recommend a comprehensive inspection and repair, rather than just a simple fault repair.

[0135] In some embodiments, fault mode recognition not only relies on the operating data of the device, but can also incorporate the working environment information of the device. Environmental factors (such as external temperature, humidity, vibration, etc.) may affect the operating state of the device, leading to the occurrence of certain fault modes. Therefore, when identifying fault modes, these environmental factors can be used as additional inputs to enhance the recognition ability of the model.

[0136] For example, in a high-humidity environment, the electrical part of the device may be corroded, resulting in a short circuit or a fault. If the environmental humidity data is taken into account, the model will be able to more accurately identify this type of fault mode.

[0137] Through fault mode recognition, the system can timely identify potential fault modes of the device, and combine health state prediction and device maintenance history for fault repair requirement analysis, so as to provide an accurate basis for subsequent maintenance decisions.

[0138] S5. Maintenance decision optimization: Analyze the identified device fault modes based on the optimal control theory to optimize the device maintenance decisions;

[0139] S5 makes optimized maintenance decisions based on the current health state of the device, predicted fault modes, and relevant economic indicators, in order to minimize the repair cost of the device, reduce the downtime of the device, and ensure the long-term stability of the device.

[0140] In this embodiment, the optimal control theory is adopted for maintenance decision optimization. The optimal control theory formulates an objective function for device maintenance, considering multiple influencing factors (such as maintenance cost, fault risk, and downtime, etc.), so as to formulate an optimal maintenance strategy for the device. The maintenance decision not only considers the current health state of the device, but also takes into account future risks, repair costs, and fault repair timeliness.

[0141] The objective function of the optimal control problem is:

[0142]

[0143] Where: J is the objective function of the optimal control problem, representing the comprehensive objective we want to optimize; T represents the time range of the control process; α1, α2, and α3 are the weight coefficients of various factors in the objective function, determining the contribution ratio of each factor to the total objective in the maintenance decision-making process; Cost(u(t)) represents the cost of performing the maintenance action u(t) at time t; u(t) represents the maintenance action performed at time t; Risk(x(t)) represents the failure risk of the equipment at time t, which is usually related to the health state x(t) of the equipment. The higher the risk, the more preventive maintenance the model usually recommends. x(t) represents the health state of the equipment at time t, obtained through the aforementioned health state prediction model; Downtime(t) represents the downtime of the equipment at time t; dt represents the infinitesimal time interval of integration, usually a discretized time step, used to calculate the total objective function within time T.

[0144] The optimal control problem is not only to minimize the objective function J, but also to satisfy certain constraint conditions. These constraint conditions usually include:

[0145] Constraints on the change of the equipment health state;

[0146] Constraints on the feasibility of maintenance operations;

[0147] Physical limitations of equipment operation and maintenance cycles.

[0148] These constraint conditions ensure the practical feasibility of the optimization decision. For example, the constraint on the equipment health state means that the health state of the equipment cannot be lower than a certain threshold, and the feasibility constraint of maintenance operations means that some maintenance operations (such as replacing components) may not be executable in a short time. All these constraint conditions will be considered in the optimization process to ensure that the maintenance decision is executable in actual operation.

[0149] To solve the optimal control problem, we can adopt numerical optimization methods. Common methods include dynamic programming, gradient descent method, and model predictive control (MPC). These methods solve the optimal maintenance strategy by updating the maintenance decision at each moment.

[0150] Specifically, when using numerical optimization methods, we first calculate the objective function through the current health state x(t) and failure risk Risk(x(t)) of the equipment. Then, we minimize J through the optimization algorithm to obtain the optimal maintenance strategy u(t) of the equipment, including when to perform maintenance, what maintenance operations to take, and how to allocate maintenance resources.

[0151] As an option, the model predictive control (MPC) method can be combined to continuously optimize maintenance decisions during real-time operation. Model predictive control makes the current optimal maintenance decision based on the historical health data and future predictions of the equipment. By continuously updating the prediction model and control strategy, MPC can provide real-time maintenance decisions for the equipment, thereby effectively improving the operation efficiency and reliability of the equipment.

[0152] In practical applications, the optimization of maintenance decisions can be carried out through the following steps:

[0153] Real-time health data collection: Through steps S1 and S2, the real-time data of the equipment is collected and processed to obtain the health status of the equipment.

[0154] Health status prediction and failure risk assessment: Using the health status prediction model in step S3 and the failure mode identification in step S4, the future failure risk of the equipment is evaluated.

[0155] Optimal control decision: Based on the health status and failure risk of the equipment, the optimal control theory is used to optimize the maintenance decision.

[0156] Execute the maintenance decision: Through the calculated optimal maintenance strategy, the equipment maintenance is carried out in a timely manner to reduce the downtime and maintenance costs.

[0157] By defining the objective function and comprehensively considering multiple factors such as maintenance costs, failure risks, and downtime, the optimal control theory can provide the most suitable maintenance decision for the equipment. This process can not only minimize the maintenance costs of the equipment, but also effectively reduce the probability of failures and downtime, thereby improving the utilization efficiency and reliability of the equipment.

[0158] S6. Market demand prediction: Dynamically predict the market demand of the equipment based on the generative adversarial network;

[0159] The goal of S6 is to predict the future market demand, which takes into account not only the health status and failure repair requirements of the equipment, but also multiple factors such as the dynamic changes in the market, fluctuations in user demands, and external environmental impacts. To achieve this goal, we adopt the generative adversarial network (GAN) technology to dynamically predict the market demand based on the existing data.

[0160] In this embodiment,

[0161] We conduct market demand forecasting by using Generative Adversarial Networks (GANs). A Generative Adversarial Network is a deep learning framework consisting of a generator and a discriminator. The generator generates prediction data that conforms to the actual demand fluctuation pattern through adversarial training, while the discriminator evaluates the difference between the generated data and the real market data. The adversarial training between the generator and the discriminator enables the finally generated data to better reflect the fluctuation trend of the real market demand.

[0162] During the market demand forecasting process, the Generative Adversarial Network can generate more realistic demand data by continuously optimizing the game between the generator and the discriminator. This process helps us accurately predict market demand and provides strong support for equipment operation and resource scheduling.

[0163] The generator and the discriminator in market demand forecasting are adversarially trained through loss functions. The goal of the generator is to generate data that approximates the real market demand, while the goal of the discriminator is to accurately distinguish between the generated data and the real data. Use the "loss function of the generator" and the "loss function of the discriminator" disclosed in S2:

[0164] The training process of the Generative Adversarial Network is an adversarial game process, in which the generator and the discriminator play against each other. The generator tries to generate more and more realistic data, while the discriminator continuously improves its ability to distinguish between real and fake data. Through this adversarial process, the generator gradually generates more realistic market demand prediction data.

[0165] Initialization phase: After the generator and the discriminator are initialized, the generator generates market demand data based on the initial noise zzz, while the discriminator is trained according to the comparison between the real market data and the generated data.

[0166] Generative adversarial training: In each training, the generator tries to make the generated data closer to the real market demand data by modifying the generation strategy, while the discriminator continuously learns how to identify the difference between the generated data and the real data.

[0167] Model optimization: As the training progresses, the generator gradually learns to generate data that conforms to the market demand fluctuation pattern, and the discriminator's recognition ability is also improved. Eventually, the generator can generate accurate prediction data that conforms to the actual market demand fluctuation pattern.

[0168] The specific implementation process of market demand forecasting can be divided into the following steps:

[0169] Data input: The system inputs the existing market demand data into the Generative Adversarial Network as the data basis for training. The market demand data includes historical sales data, customer demand data, industry trend data, etc.

[0170] The generator generates predicted data: The generator generates predicted data on future market demand based on the input data. The generated data should be able to reflect the fluctuation trend of future demand.

[0171] The discriminator evaluates the generated data: The discriminator evaluates the generated data to determine whether it conforms to the real market demand data. The discriminator learns how to distinguish real data from generated data through backpropagation.

[0172] Optimization process: The generator and the discriminator are optimized through adversarial training. The generator continuously adjusts its generation strategy under the feedback of the discriminator until the generated data cannot be distinguished as fake data by the discriminator.

[0173] Output the market demand prediction result: Finally, after multiple rounds of training, the generator can output real and market-law-compliant demand prediction data, providing an accurate basis for the operation decision-making, resource allocation, and market scheduling of the equipment.

[0174] As an option, the generative adversarial network has higher flexibility and prediction accuracy compared to traditional market demand prediction methods (such as linear regression or time series prediction methods). GAN can handle complex non-linear relationships and the influence of multiple factors. Especially when market demand is affected by multiple factors, traditional methods may be difficult to capture the deep features of market fluctuations. However, through generative adversarial training, GAN can simulate the real laws of market demand fluctuations and generate more accurate prediction results.

[0175] Specifically, the generative adversarial network also has advantages in dealing with irregular or scarce data. Even when the amount of market demand data is insufficient, the generator can still learn the laws of demand fluctuations from the existing data and generate reliable market demand predictions.

[0176] Through the adversarial training of the generator and the discriminator, not only data consistent with the actual market demand fluctuations is generated, but also the complex laws of market dynamic changes can be captured, thus providing accurate market demand prediction support for equipment operation and resource scheduling.

[0177] S7, Intelligent pricing: Optimize and adjust the transaction price of the equipment according to the generated market demand data and the generated equipment health status information;

[0178] Step S7 is to provide a dynamically optimized pricing strategy for the equipment and services based on these data. Intelligent pricing not only needs to consider market demand, equipment health status, and maintenance requirements, but also factors such as the competitiveness of the equipment and the price sensitivity of consumers, so as to formulate a reasonable price strategy.

[0179] In this embodiment, intelligent pricing dynamically adjusts prices through the combination of game theory models and market demand prediction data. The game theory model helps the platform simulate the game process with consumers, and optimizes the transaction price by calculating the Nash equilibrium to achieve the goal of maximizing the platform's revenue. The platform needs to adjust its pricing strategy in a changing market demand and competitive environment, ensuring that the equipment can meet market demand while maintaining profitability under the pressure of competitors.

[0180] In the game theory model, the interaction between the platform and consumers is expressed through price decisions. The price in the market is generated by the game between the platform and consumers, and the platform maximizes its revenue through pricing decisions. The game model can be expressed as:

[0181]

[0182] Where: is the maximum revenue of the platform, indicating the economic benefits achieved by the platform through selecting the optimal pricing strategy p platform ; p platform is the equipment pricing of the platform, representing the selling price set by the platform for the equipment or service in the market; p competitor is the pricing strategy of competitors, the equipment prices set by other competitors in the market. The platform needs to adjust according to the pricing strategy of competitors to maintain market competitiveness; D is the market demand prediction, based on the market demand prediction results in the aforementioned step S6, considering the demand quantities at different prices; C(p platform ) is the cost function of the platform, indicating the production, operation, maintenance, etc. costs paid by the platform according to the equipment pricing p platform . It is the cost related to pricing, reflecting the economic burden brought by each pricing level; The revenue function R(p platform , p competitor , D) and the cost function C(p platform ) are crucial for pricing decisions. The revenue of the platform depends on the pricing strategy, competitor pricing, and market demand. The specific expressions are as follows:

[0183] R(p platform , p competitor , D) = p platform ×D(p platform , p competitor , D);

[0184] Where: R(p platform , p competitor , D) is the revenue of the platform, and the revenue generated by the platform through pricing p platform and market demand D; D(p platform , p competitor , D) is the demand function, indicating the demand at price pplatform and the market demand volume at the price p of competitors competitor of the market. This function is dynamically adjusted according to the market demand forecast and the pricing strategy of competitors.

[0185] The cost function C(p platform ) describes the costs incurred by the platform to provide equipment or services, usually including maintenance costs, production costs, etc.:

[0186] C(p platform ) = α4· platform + β;

[0187] where: C(p platform ) is the cost function of the platform, representing the expenses incurred by the platform when providing equipment services; α4 is the proportional coefficient of equipment costs, representing the fixed costs of each unit of equipment that the platform needs to consider when pricing; β is an additional fixed expense item, such as operating costs, management expenses, etc.

[0188] Through the concept of Nash equilibrium in game theory, the platform can find the optimal pricing strategy p platform under the given market demand and competitive environment,

[0189]

[0190] such that the revenue of the platform is maximized. Specifically, the Nash equilibrium can be obtained by solving the following equation:

[0191] When pricing, the platform also needs to consider the price elasticity of the market, that is, how the change in price affects the demand volume. The demand elasticity function ∈(p) can be expressed as:

[0192]

[0193] where: ∈(p) represents the sensitivity of market demand to price changes. The greater the demand elasticity, the greater the impact of price changes on demand; is the derivative of the demand function with respect to price, representing the rate of change of demand volume when the price changes; p is the pricing level, the equipment transaction price set by the platform; D(p) is the demand function, representing the market demand volume at the price p.

[0194] Through demand elasticity, the platform can adjust the pricing strategy according to different price levels. For example, when the market demand is more sensitive, the platform may choose a lower price to attract consumers; while in the case of less sensitive demand, the price can be appropriately increased.

[0195] The specific implementation process of intelligent pricing is as follows:

[0196] Market demand forecasting: Based on the market demand forecasting in step S6, the platform obtains future demand fluctuation information.

[0197] Game theory model calculation: The platform uses the game theory model to play against the prices of competitors and calculates the optimal pricing strategy.

[0198] Adjusting demand elasticity: The platform further adjusts the pricing strategy according to the demand elasticity analysis to achieve the best balance between demand and revenue.

[0199] Implementing the pricing strategy: According to the final pricing strategy, the platform adjusts the price of the device and conducts market promotion to attract consumers to purchase.

[0200] By calculating the Nash equilibrium and price elasticity, the platform can adjust the pricing strategy in real time to adapt to the changing market demand, maximize revenue and ensure market competitiveness.

[0201] S8. Blockchain recording: All data and transaction records are stored through blockchain technology to ensure the immutability of the data;

[0202] S8 ensures the security, immutability and transparency of this data through blockchain technology. By using blockchain, all transaction records and device status information will be reliably stored and can be traced and verified. Blockchain technology can ensure that the records of device transactions are publicly transparent while protecting the privacy and security of the data.

[0203] In this embodiment, all transaction records and device data will be stored in the blockchain and automatically executed through smart contracts. Smart contracts can automatically execute the terms of device transactions according to predefined rules to ensure the transparency and compliance of device transactions. The distributed ledger technology of blockchain guarantees the immutability of transaction records, while the automated execution of smart contracts improves the efficiency of the system.

[0204] Specifically, the system will write these data into the chain through blockchain technology based on the health status prediction, market demand and intelligent pricing data in the foregoing steps to ensure the integrity and security of the data.

[0205] Data collection and encryption: All data generated in the foregoing steps S1 to S7 (such as device status, transaction data, pricing information, etc.) will first be encrypted to protect its privacy.

[0206] Smart contract execution: The smart contract automatically executes the relevant conditions of the device transaction according to the pricing information and transaction terms generated in step S7 and writes the transaction data into the blockchain.

[0207] Blockchain Record: All transaction records will be stored in the form of blocks. Each block contains detailed information about device transactions and is connected to the previous block through a hashing algorithm to form an immutable blockchain.

[0208] Consensus Mechanism and Data Verification: Each data update in the blockchain is verified through a consensus mechanism to ensure the consistency and reliability of transaction data.

[0209] Smart contract is a tool for automatically executing transaction terms. In step S8, when a device transaction occurs, the smart contract automatically verifies and executes relevant terms. The execution process can be expressed as:

[0210]

[0211] where: p platform is the transaction price set by the platform, which is generated by the pricing strategy in the aforementioned step S7; p consumer is the price paid by the consumer. The consumer pays the device fee according to the terms of the smart contract; T is the terms of the device transaction, including various conditions of the transaction, such as price, device information, buyer and seller information, etc.; BlockData is the transaction data stored in the blockchain, including all information of the transaction, such as device type, price, buyer and seller information, etc.; StoreData is the execution function of the smart contract, which is used to write the transaction data into the blockchain.

[0212] In the blockchain, device transaction data is stored in the form of blocks. The structure of each block is as follows:

[0213] Block = [Header, TransactionData, Timestamp, PreviousHash, Hash];

[0214] where: Header is the block header information, including block version, timestamp, hash value of the previous block, etc. TransactionData is the transaction data contained in the block, such as the price of the device transaction, buyer and seller information, etc.; Timestamp is the timestamp of the transaction, recording the time when the transaction occurs; PreviousHash is the hash value of the previous block, which is used to connect the current block with the previous block to form the blockchain; Hash is the hash value of the current block, which is used to uniquely identify the current block and ensure the immutability of the data.

[0215] By storing the transaction data in the blockchain, the platform can ensure the immutability and transparency of the data. All transactions are encrypted and recorded in a distributed ledger to prevent data loss or tampering. Blockchain technology can provide reliable data verification and transaction traceability for the platform and consumers.

[0216] As an option, blockchain technology also has the characteristic of decentralization, that is, it does not need to rely on a single intermediary. All participants (such as platforms, consumers, third-party verification agencies, etc.) can access and verify transaction data, thereby improving the trust and transparency of the system.

[0217] Specifically, blockchain technology can provide a complete historical record of all transactions of the device. Both users and platforms can query various data of device transactions at any time to ensure the authenticity and reliability of device transactions.

[0218] By automatically executing the terms of device transactions through smart contracts, blockchain technology provides a decentralized and reliable data storage and transaction verification system. Every transaction of the device is recorded on the blockchain, enabling full traceability, ensuring the authenticity of data, and enhancing the trust of the platform.

[0219] S9. Smart contract execution: Based on blockchain technology, automatically execute relevant terms of the transaction, including payment, delivery, and quality assurance;

[0220] The core task of S9 is to automatically execute relevant terms of device transactions through smart contract technology. Smart contracts can ensure that transactions are automatically completed according to predetermined conditions, reduce manual intervention, and ensure the transparency, fairness, and efficiency of transactions.

[0221] In this embodiment, the execution of the smart contract in device transactions is based on the pricing strategy and market demand data in step S7. The smart contract will automatically verify and execute according to the terms of device transactions. All transaction data and conditions will be entered into the blockchain to ensure the immutability and transparency of transactions.

[0222] When the smart contract is executed, the system will automatically perform operations such as fund settlement and device delivery according to the transaction terms. Use the "execution process" disclosed in S8:

[0223] Verify transaction conditions: The smart contract will first verify transaction conditions, such as the payment amount and device status. If all conditions are met, the contract will continue to execute.

[0224] Automatic payment and delivery: After the transaction conditions are verified, the smart contract will automatically execute the payment operation and confirm the device delivery.

[0225] Record transaction data: The smart contract will record the transaction data into the blockchain to ensure the transparency and immutability of the data.

[0226] Confirm transaction status: After the transaction is completed, the system will return the transaction status indicating whether the transaction has been successfully completed.

[0227] Generally, the use of smart contracts can reduce manual intervention, improve transaction efficiency, and ensure that transactions are automatically executed according to predefined conditions. Through smart contracts, the transaction process is more efficient and fair, and both parties to the transaction can trust the platform more.

[0228] As an option, smart contracts can also ensure the transparency of transactions. Each transaction is automatically recorded on the blockchain, and the relevant data of any transaction can be traced and verified, ensuring the reliability and transparency of the transaction.

[0229] Specifically, smart contracts provide an automated and decentralized execution method for device transactions, reducing the risk of manual operations and improving the security and efficiency of the system.

[0230] Through blockchain technology, the authenticity and immutability of all transaction data are guaranteed, ensuring the fairness and transparency of device transactions.

[0231] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring and optimizing the equipment status of a new energy charging station, characterized in that: The following steps are involved: S1. Equipment data collection: Various sensors and counters are installed on the equipment to collect the equipment's operating data in real time, including voltage, current, temperature, humidity, equipment usage time, and charging times; S2, data preprocessing and noise removal: Use the generative adversarial network to preprocess the data collected in S1 and remove the noise in the data; S3, equipment health status prediction: predict the data processed by S2 based on deep reinforcement learning and graph neural network, and generate a health prediction model; S4, Fault mode identification: Use multi-task learning to train the health prediction model generated in S3 to identify equipment failure modes; S5, maintenance decision optimization: analyze the equipment failure mode identified in S4 based on optimal control theory to optimize the equipment maintenance decision; S6. Market demand forecasting: Dynamically forecast the market demand for equipment based on generative adversarial networks; S7, intelligent pricing: optimize and adjust the transaction price of the equipment based on the market demand data generated by S6 and the equipment health status information generated by S3; S8, blockchain records: All data and transaction records from S1 to S7 are stored through blockchain technology to ensure that the data cannot be tampered with; S9. Smart contract execution: Based on blockchain technology, it automatically executes relevant terms of the transaction, including payment, delivery and warranty.

2. The method for monitoring and optimizing the equipment status of a new energy charging station according to claim 1, characterized in that: The equipment data collection includes: Install voltage sensors, current sensors, temperature sensors, humidity sensors and counters; Real-time monitoring and acquisition of key parameters during device operation, including voltage, current, temperature, humidity, device usage time and number of charging times.

3. The method for monitoring and optimizing the equipment status of a new energy charging station according to claim 1, characterized in that: The data preprocessing and noise removal include: Use the generator of the generative adversarial network to generate clean training data, the discriminator determines the authenticity of the data, and optimizes the generator and discriminator based on training feedback; The collected equipment operation data is input into the generative adversarial network for training to remove noise and abnormal data and generate a high-quality data set.

4. The method for monitoring and optimizing the equipment status of a new energy charging station according to claim 1, characterized in that: The equipment health status prediction includes: Based on the deep reinforcement learning algorithm, a health prediction model is trained based on the historical data of the device to predict the future health status changes of the device; Use graph neural networks to perform multi-dimensional modeling of various operating parameters of the equipment to increase the accuracy and robustness of health status prediction.

5. The method for monitoring and optimizing the equipment status of a new energy charging station according to claim 1, characterized in that: The graph neural network is performed through the following steps: The device health data generated by S3 is represented as a graph structure, where each node represents a device state parameter and each edge represents the relationship between device states; Graph convolution operations are used to propagate device status information to quickly extract deep features of the device health status.

6. The method for monitoring and optimizing the equipment status of a new energy charging station according to claim 1, characterized in that: The fault mode identification includes: Use multi-task learning models to simultaneously predict equipment health status, identify fault patterns, and analyze equipment fault repair needs; Network layers of multiple tasks are shared during training to optimize the classification accuracy of failure modes and improve the accuracy of equipment health status identification.

7. The method for monitoring and optimizing the equipment status of a new energy charging station according to claim 1, characterized in that: The maintenance decision optimization includes: Based on optimal control theory, define and optimize the objective function of equipment maintenance, which includes maintenance cost, failure risk and equipment downtime; By solving the optimal control problem, the optimal maintenance strategy is calculated to reduce the unplanned downtime of equipment and reduce maintenance costs.

8. The method for monitoring and optimizing the equipment status of a new energy charging station according to claim 1, characterized in that: The market demand forecast includes: Generate data related to market demand using a generative adversarial network, and use a discriminator to determine the authenticity of the generated data; Through adversarial training, a prediction model that is more in line with actual market demand fluctuations is generated, and the market demand forecast of the equipment is dynamically adjusted.

9. The method for monitoring and optimizing the equipment status of a new energy charging station according to claim 1, characterized in that: The smart pricing includes: Use game theory models to adjust equipment prices based on equipment health status and market demand information; By calculating the Nash equilibrium point, the optimal transaction price for both supply and demand sides is determined to ensure market supply and demand balance and maximize platform profits.

10. The method for monitoring and optimizing the equipment status of a new energy charging station according to claim 1, characterized in that: The blockchain records include: All equipment operation data, health prediction data, maintenance decisions and transaction records involved in S1 to S9 are distributed and stored through blockchain technology; Smart contracts are used to automatically execute equipment transaction terms, ensuring the transparency and immutability of the transaction process, and providing full traceability of equipment transactions.

Images