Power distribution equipment state evaluation system and method

Abstract

The invention relates to the technical field of power distribution equipment maintenance, and discloses a power distribution equipment state evaluation system and method, and the system obtains the electrical parameters, physical state, operation environment, maintenance history and other multi-dimensional data of power distribution equipment in real time through a data collection module. And the control strategy output module constructs an intelligent control model, models an equipment operation process into a Markov decision process, and dynamically outputs a control strategy. The control strategy output module comprises a state space definition sub-module, an action space definition sub-module, a reward function design sub-module, a strategy making sub-module, a model training sub-module and the like, and optimization of the strategy is ensured. And the equipment state evaluation module comprehensively considers the electrical parameter stability, the physical state health degree, the operating environment adaptability, the maintenance historical record good degree and the intelligent control strategy effectiveness based on a preset evaluation index system, carries out quantitative evaluation on the equipment state, and outputs an evaluation grade. According to the invention, the comprehensiveness and accuracy of power distribution equipment state evaluation are improved.

Description

Technical Field

[0001] This invention relates to the field of power distribution equipment maintenance technology, specifically to a power distribution equipment condition evaluation system and method. Background Technology

[0002] In power systems, distribution equipment serves as a crucial link connecting the power grid and users, and its operational status directly affects the stability of the power system and the quality of electricity supply for users. Traditional condition assessment of distribution equipment mainly relies on periodic maintenance, which has drawbacks such as low efficiency, difficulty in reflecting real-time changes in equipment status, and difficulty in predicting potential faults. With the development of smart grid and Internet of Things (IoT) technologies, real-time and accurate condition assessment of distribution equipment has become an important means to improve the operational efficiency and reliability of power systems.

[0003] However, existing power distribution equipment condition assessment systems often rely solely on single monitoring data, such as electrical parameters or physical condition, while neglecting multi-dimensional information such as the operating environment, maintenance history, and inter-equipment interactions. This limits the comprehensiveness and accuracy of the assessment results, making it difficult to accurately reflect the true condition of the equipment. Furthermore, traditional assessment methods often lack intelligent decision support, failing to dynamically adjust control strategies based on equipment condition to achieve optimized operation and preventative maintenance. Summary of the Invention

[0004] The purpose of this invention is to provide a power distribution equipment condition evaluation system and method to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a power distribution equipment condition evaluation system, comprising:

[0006] The data acquisition module is used to collect real-time operating data of power distribution equipment, including electrical parameters, physical status, operating environment, and maintenance history data.

[0007] A control strategy output module is used to construct an intelligent control model, which models the operation process of the power distribution equipment as a Markov decision process and dynamically outputs the control strategy for the power distribution equipment by analyzing the operating data of the power distribution equipment in real time. The control strategy output module further includes:

[0008] The state space definition submodule is used to define the state space of the power distribution equipment, including the electrical parameter dimension, physical state dimension, operating environment dimension, and maintenance history dimension.

[0009] The Action Space Definition submodule is used to define the set of actions that may be taken when the power distribution equipment is in an abnormal state;

[0010] The reward function design submodule, connected to the action space definition submodule, is used to design reward functions based on equipment reliability, operating efficiency, maintenance costs, power quality, and safety performance, and to evaluate action results.

[0011] The strategy formulation submodule, connected to the reward function design submodule, is used to formulate a strategy for selecting actions in a given state, and continuously optimizes the strategy selection through an algorithm to maximize the cumulative reward.

[0012] The model training submodule, which employs a deep reinforcement learning algorithm, is connected to the policy formulation submodule and is used to train the intelligent control model.

[0013] The equipment status evaluation module, connected to the data acquisition module, is used to receive real-time collected operating data of the power distribution equipment and to quantitatively evaluate the overall status of the power distribution equipment based on a preset evaluation index system, and output the status evaluation level. The evaluation index system includes at least five aspects: electrical parameter stability, physical health, operating environment adaptability, maintenance history record quality, and the effectiveness of the control strategy output by the intelligent control model.

[0014] Preferably, the electrical parameter dimensions include voltage, current, power factor, active power, reactive power, power quality, and harmonic content;

[0015] The physical state dimensions include the device's temperature, vibration, noise, partial discharge status, and insulation status;

[0016] The operating environment dimensions include ambient temperature, ambient humidity, atmospheric pressure, and pollution level;

[0017] The maintenance history dimension includes the equipment's historical repair records, maintenance records, fault types, fault frequencies, and repair times.

[0018] Preferably, the action space definition submodule takes the following action set: adjusting electrical parameters, switching equipment operating modes, starting backup equipment, performing equipment maintenance, notifying operators, and scheduling resources.

[0019] Preferably, the model training submodule uses the Deep Q-Network (DQN) algorithm to train the intelligent control model.

[0020] Preferably, the steps of the model training submodule in training the intelligent control model include:

[0021] Step 1: Initialize model parameters: Set the neural network structure in the deep reinforcement learning algorithm, including the input layer, hidden layer and output layer, as well as the learning rate, discount factor and exploration rate;

[0022] Step 2: Construct an experience replay buffer: This buffer is used to store experience samples generated by the interaction between the agent and the power distribution equipment environment, including states, actions, rewards, and the next state.

[0023] Step 3: Collect training data: Prepare the training dataset as input for the intelligent control model;

[0024] Step 4: Execute the action: Select an action based on the current state and strategy, execute the action through actual control or simulation environment, and observe the next state, reward value, and whether to terminate.

[0025] Step 5: Store the experience samples: Store the experience samples obtained in Step 4 into the experience replay buffer;

[0026] Step 6: Sampling from the buffer: Randomly select a batch of experience samples from the experience replay buffer for model training;

[0027] Step 7: Calculate the target value: For each sampled empirical sample, calculate the target Q value using the Bellman equation. The formula is as follows:

[0028] Q target =r+γmax a′ Q(s ′ ,a ′ ;θ - )

[0029] Among them, Q target Here, γ is the target Q value, r is the reward value, γ is the discount factor, and s is the target Q value. ′ The next state is a ′ The action that can be selected for the next state, Q(s) ′ ,a ′ ;θ - ) is the target network parameter θ - Calculated Q value;

[0030] Step 8: Update model parameters: Calculate the difference between the current Q-value and the target Q-value using the mean squared error loss function, and update the neural network parameters θ. The loss function formula is:

[0031]

[0032] Where N is the number of samples;

[0033] Step 9, Iterative Training: Repeat steps 4 to 8 until the preset number of training rounds is reached;

[0034] Step 10: Save the model: Save the parameters of the trained intelligent control model for subsequent power distribution equipment status evaluation.

[0035] Preferably, in step 8, the neural network parameters θ are updated using the gradient descent algorithm. The calculation formula for the gradient descent algorithm is as follows:

[0036]

[0037] Where θ represents the model parameters, which include all the weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter updates; This represents the gradient of the loss function J(θ) with respect to the parameter θ. It is a vector pointing in the direction where the loss function grows the fastest; := represents an assignment operation, that is, updating the value of parameter θ.

[0038] Preferably, the reward function design submodule receives the action execution result from the action space definition submodule and calculates the reward value r according to the following formula:

[0039] r=w1·R+w2·E-w3·C+w4·T+w5·S

[0040] Where w1, w2, w3, w4, and w5 represent the weighting coefficients of equipment reliability reward, operating efficiency reward, maintenance cost penalty, power quality reward, and safety performance reward, respectively, and satisfy w1+w2+w3+w4+w5=1; R, E, C, T, and S represent the quantitative index values ​​of equipment reliability, operating efficiency, maintenance cost, power quality, and safety performance, respectively.

[0041] Preferably, the training data for the intelligent control model is obtained through the following steps:

[0042] Based on preset equipment operating status scenarios and evaluation indicators, simulated data containing multiple equipment status types and corresponding evaluation results is generated; the simulated data includes operating parameters, performance indicators, maintenance records, power quality data and safety performance data of the equipment in normal, abnormal and fault states.

[0043] For each type of equipment status, different operating environment conditions and external influencing factors are simulated to further enrich the simulation dataset;

[0044] The various types of data in the simulated dataset are encoded and standardized.

[0045] Preferably, the method of encoding the data in the simulated training dataset includes: performing one-hot encoding transformation on discrete features;

[0046] The method for standardizing the data in the simulated training dataset includes: using the Z-score standardization method to standardize continuous features, specifically: subtracting the mean of each numerical data point and dividing by its standard deviation, so that the processed data conforms to a standard normal distribution. The standardization formula is: Z = (X - μ) / σ, where X is the original data, μ is the mean of the original data, σ is the standard deviation of the original data, and Z is the standardized data.

[0047] Preferably, a method for evaluating the condition of power distribution equipment includes:

[0048] S1: Real-time acquisition of operating data from power distribution equipment;

[0049] S2: Define the state space of the power distribution equipment, and organize and classify the collected data according to the dimensions of electrical parameters, physical state, operating environment and maintenance history.

[0050] S3: Determine the set of actions that may be taken when the power distribution equipment is in an abnormal state;

[0051] S4: Based on equipment reliability, operating efficiency, maintenance costs, power quality, and safety performance, design a reward function to evaluate the results of the actions taken;

[0052] S5: Formulate a strategy that determines the action to be taken in a given state based on the current device state and reward function, and continuously optimize the choice of strategy through an algorithm;

[0053] S6: Employs deep reinforcement learning algorithms and combines them with experience samples accumulated during the policy formulation process to train an intelligent control model, enabling the model to dynamically output control strategies for power distribution equipment based on real-time data.

[0054] S7: Based on a preset evaluation index system, the overall status of the power distribution equipment is quantitatively evaluated. This index system includes at least five aspects: electrical parameter stability, physical health, operating environment adaptability, maintenance history record quality, and intelligent control strategy effectiveness. By calculating the scores of each index and taking them into account, the status evaluation level of the power distribution equipment is output.

[0055] Compared with the prior art, the beneficial effects of the present invention are:

[0056] This invention collects multi-dimensional operational data of power distribution equipment in real time through a data acquisition module, including electrical parameters, physical state, operating environment, and maintenance history, ensuring the comprehensiveness and timeliness of the evaluation data. This invention constructs an intelligent control model, modeling the operation of power distribution equipment as a Markov decision process, capable of analyzing equipment status in real time and dynamically outputting control strategies. Through an intelligent decision support mechanism, not only is response speed and decision-making efficiency improved, but the control strategy can also be adjusted promptly according to changes in equipment status, achieving optimized operation and fault prevention. Through the collaborative work of the reward function design submodule and the strategy formulation submodule, the system can comprehensively consider multiple aspects such as equipment reliability, operating efficiency, maintenance costs, power quality, and safety performance to formulate and optimize control strategies. This strategy selection mechanism based on multi-dimensional evaluation ensures the scientific nature and effectiveness of the control strategy, contributing to improving the overall operating efficiency and reliability of the power system. This invention employs a deep reinforcement learning algorithm for model training, enabling the intelligent control model to continuously learn and adapt to changes in equipment status, improving the system's adaptability and flexibility. Simultaneously, by defining the state space and action space, the system can flexibly respond to different types of power distribution equipment and operating states, exhibiting broad applicability and scalability.

[0057] The equipment status evaluation module, based on a pre-set evaluation index system, quantitatively assesses the overall status of power distribution equipment and outputs a status evaluation level. The quantitative evaluation results not only allow managers to intuitively understand the equipment status but also provide a scientific basis for equipment maintenance, repair, and replacement, helping to reduce maintenance costs and extend equipment lifespan. Attached Figure Description

[0058] Figure 1 This is a schematic diagram illustrating the working principle of a power distribution equipment condition evaluation system according to the present invention.

[0059] Figure 2 This is a flowchart of the training process for an intelligent control model.

[0060] Figure 3 This is a schematic diagram of the reward function design. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] Please see Figure 1-3The present invention provides a technical solution: a power distribution equipment condition evaluation system, the system comprising:

[0063] The data acquisition module is responsible for collecting real-time operating data from the power distribution equipment. The collected data includes electrical parameters (such as voltage, current, and power factor), physical conditions (such as temperature, vibration, and noise), operating environment (such as humidity, temperature, and pollution levels), and maintenance history data (such as maintenance records and fault records). This data is acquired through sensors, monitoring devices, or data interfaces and transmitted to the system for further processing.

[0064] The control strategy output module's core function is to construct an intelligent control model. This model models the operation of the power distribution equipment as a Markov decision process, dynamically outputting control strategies by analyzing the operational data transmitted from the data acquisition module in real time. In this process, the model considers the equipment's current state, historical data, and potential future state changes to make optimal decisions.

[0065] The construction of the intelligent control model is further subdivided into several sub-modules. The state space definition sub-module is responsible for defining the state space of the power distribution equipment, organizing and classifying the collected data according to the dimensions of electrical parameters, physical state, operating environment, and maintenance history, providing a clear input state for the model. The action space definition sub-module defines the set of actions that may be taken when the equipment is in an abnormal state, such as adjusting electrical parameters or performing maintenance operations, providing the model with an optional action space.

[0066] The reward function design submodule is closely connected to the action space definition submodule. It designs the reward function based on key indicators such as equipment reliability, operating efficiency, maintenance costs, power quality, and safety performance. This function evaluates the results of the actions taken by the model, providing positive or negative rewards to guide the model in learning and optimizing control strategies. The strategy formulation submodule, based on the reward function design, formulates a strategy for selecting actions in a given state and continuously optimizes the strategy selection through algorithms (such as Q-learning and deep reinforcement learning) to maximize cumulative rewards.

[0067] The model training submodule employs a deep reinforcement learning algorithm, combined with experience samples accumulated by the policy formulation submodule, to train the intelligent control model. Through continuous iterative learning, the model can gradually grasp the patterns in the operation of power distribution equipment, improving the accuracy and effectiveness of the control strategy.

[0068] The equipment status evaluation module connects to the data acquisition module and receives real-time operational data from the power distribution equipment. Based on a pre-set evaluation index system, this module quantitatively assesses the overall status of the power distribution equipment. The evaluation index system includes at least five aspects: electrical parameter stability, physical health, adaptability to the operating environment, good maintenance history, and the effectiveness of the control strategies output by the intelligent control model. By calculating the scores of each index and comprehensively considering them, the module outputs the status evaluation level of the power distribution equipment, providing a scientific basis for optimized operation and preventative maintenance.

[0069] In terms of electrical parameters, the system collects detailed data on voltage, current, power factor, active power, reactive power, power quality, and harmonic content. This data is acquired in real time using high-precision sensors and monitoring devices and serves as crucial input to the intelligent control model. Voltage and current reflect the basic electrical characteristics of the equipment, while power factor and active / reactive power ratios reflect the equipment's energy efficiency and power distribution. Power quality and harmonic content are used to assess the degree of pollution the equipment causes to the power grid and the potential power quality problems it may trigger.

[0070] In terms of physical state, the system monitors the equipment's temperature, vibration, noise, partial discharge, and insulation status. Temperature is acquired using an infrared thermometer or a built-in temperature sensor to determine if overheating is occurring. Vibration and noise are measured using vibration sensors and noise meters to reflect the equipment's mechanical stability and operating environment. Partial discharge is monitored using a partial discharge detector to identify potential internal faults. Insulation status is assessed through insulation resistance or dielectric loss testing to ensure good insulation performance.

[0071] The operating environment dimension encompasses ambient temperature, humidity, atmospheric pressure, and pollution levels. This data is acquired through environmental monitoring stations or weather stations to assess the severity of the environment in which the equipment operates and its potential impact on equipment performance.

[0072] The maintenance history dimension records the equipment's historical repair records, maintenance records, fault types, fault frequencies, and repair times. This data is obtained through the equipment management system or maintenance logbook, providing the intelligent control model with the equipment's maintenance history and fault modes, which helps the model more accurately predict the equipment's future condition.

[0073] In terms of action space definition, the system sets a set of actions such as adjusting electrical parameters, switching equipment operating modes, starting standby equipment, performing equipment maintenance, notifying operators, and scheduling resources.

[0074] The present invention will be further described below with reference to Examples 1 to 3:

[0075] Example 1:

[0076] The model training submodule uses the Deep Q-Network (DQN) algorithm to train the intelligent control model. The training process includes the following steps:

[0077] Step 1: Initialize Model Parameters: Set the neural network structure in the deep reinforcement learning algorithm, including the input layer, hidden layers, and output layer. The input layer receives data from dimensions such as electrical parameters, physical state, operating environment, and maintenance history; the hidden layer is used to extract features and learn mapping relationships; the output layer outputs the Q-value corresponding to each action. Simultaneously, set hyperparameters such as the learning rate, discount factor, and exploration rate to control the training process.

[0078] Step 2: Construct an experience replay buffer: This buffer stores experience samples generated from the agent's interactions with the power distribution equipment environment. Each experience sample includes the current state, the selected action, the reward received, and the next state. The size of the buffer is set according to training requirements to ensure the model can learn sufficient experience.

[0079] Step 3: Collect training data: Prepare a training dataset containing various operating states and action combinations. This data can be obtained from actual operating data, simulation data, or historical data.

[0080] Step 4: Execute the action: Select an action based on the current state and policy. In the early stages of training, the model may take random actions to explore the environment; as training progresses, the model will gradually learn to select the optimal action based on the state. After the action is executed, observe the next state, the reward value, and a flag indicating whether the process has terminated.

[0081] Step 5: Store the experience samples: Store the experience samples obtained in Step 4 into the experience playback buffer. The buffer adopts a first-in, first-out (FIFO) strategy; when the buffer is full, the earliest stored sample will be replaced.

[0082] Step 6: Sampling from the buffer: Randomly select a batch of experience samples from the experience replay buffer for model training. This sampling method can break the temporal correlation between samples and improve the training efficiency of the model.

[0083] Step 7: Calculate the target value: For each sampled empirical sample, calculate the target Q-value using the Bellman equation. The target Q-value is a discount to the sum of the reward value obtained from the current action and the maximum Q-value of the possible actions in the next state. The calculation formula is:

[0084]

[0085] Among them, Q target Here, γ is the target Q value, r is the reward value, γ is the discount factor, and s is the target Q value. ′ The next state is a ′The action that can be selected for the next state, Q(s) ′ ,a ′ ;θ - ) is the target network parameter θ - Calculate the Q value.

[0086] Step 8: Update model parameters: Calculate the difference between the current Q-value and the target Q-value using the mean squared error loss function, and update the neural network parameters θ. The loss function formula is:

[0087]

[0088] Where N is the number of samples.

[0089] The gradient descent algorithm is used to update the neural network parameters θ. The formula for calculating the gradient descent algorithm is as follows:

[0090]

[0091] Where θ represents the model parameters, which include all the weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter updates; This represents the gradient of the loss function J(θ) with respect to the parameter θ. It is a vector pointing in the direction where the loss function grows the fastest; := represents an assignment operation, that is, updating the value of parameter θ.

[0092] Step 9, Iterative Training: Repeat steps 4 to 8 until the preset number of training rounds is reached or other stopping conditions are met. During training, the model will gradually learn to select the optimal action based on the state to improve cumulative rewards.

[0093] Step 10: Save the model: Save the trained intelligent control model parameters to a file or database for use in subsequent power distribution equipment condition assessments. The saved model can be updated periodically or fine-tuned according to actual needs.

[0094] Through the above steps, this invention trains an intelligent control model based on a deep reinforcement learning algorithm, enabling accurate evaluation of equipment status and output of dynamic control strategies.

[0095] Example 2:

[0096] This embodiment provides a specific implementation method for a reward function design submodule. This submodule receives the action execution results from the action space definition submodule and calculates the reward value based on preset weight coefficients and quantification index values. The implementation steps and calculation method of this reward function design submodule will be described in detail below.

[0097] First, define the components of the reward function, including equipment reliability reward, operating efficiency reward, maintenance cost penalty, power quality reward, and safety performance reward. Each component corresponds to a weighting coefficient, denoted as w1, w2, w3, w4, and w5, respectively, and these weighting coefficients satisfy w1+w2+w3+w4+w5=1 to ensure the comprehensiveness and rationality of the reward value.

[0098] For each component, a quantitative indicator value is set. Specifically:

[0099] Equipment reliability quantification index (R): It is quantified by indicators such as equipment failure rate and mean time between failures (MTBF). The value range is set to [0,1], where 1 indicates that the equipment is completely reliable and 0 indicates that the equipment is completely unreliable.

[0100] The quantitative index value of operating efficiency (E) is quantified based on parameters such as the energy consumption efficiency and processing speed of the equipment. The value range is set to [0,1], where 1 represents the highest operating efficiency of the equipment and 0 represents the lowest operating efficiency of the equipment.

[0101] Maintenance cost quantification index value (C): Based on data such as equipment maintenance costs and repair frequency, it is quantified and serves as a maintenance cost penalty item. The value range is set to [0, +∞), where 0 represents no maintenance cost and the larger the value, the higher the maintenance cost.

[0102] Power quality quantification index (T): It is quantified by power quality parameters such as voltage fluctuation and frequency deviation. The value range is set to [0,1], where 1 represents the best power quality and 0 represents the worst power quality.

[0103] Safety quantification index value (S): It is quantified based on factors such as the equipment's accident rate and safety protection measures. The value range is set to [0,1], where 1 represents the highest equipment safety and 0 represents the lowest equipment safety.

[0104] After obtaining the above quantitative indicator values, the reward value r is calculated according to the following formula:

[0105] r=w1·R+w2·E-w3·C+w4·T+w5·S

[0106] Among them, equipment reliability rewards, operating efficiency rewards, power quality rewards, and safety performance rewards are positive values, while maintenance cost penalties are negative values, to reflect their positive or negative impact on the overall reward value.

[0107] Through the above implementation methods, the reward function design submodule can accurately calculate the reward value based on the action execution result, providing an effective feedback mechanism for system optimization and control.

[0108] Example 3:

[0109] This implementation describes in detail the process of acquiring and processing training data for intelligent control models, including the generation and preprocessing steps of simulation data:

[0110] Generation of simulation data:

[0111] Based on preset equipment operating state scenarios and evaluation indicators, a simulation data generation framework is constructed. This framework can simulate various operating conditions of the equipment under normal, abnormal, and fault states. For each state type, detailed records are made of the equipment's operating parameters (such as temperature, pressure, and speed), performance indicators (such as efficiency and energy consumption), maintenance records (such as maintenance time and number of maintenance operations), power quality data (such as voltage fluctuations and frequency deviations), and safety performance data (such as the number of safety incidents and the status of safety protection).

[0112] For example, for a single electric motor, simulate its state during normal operation, overload operation, and bearing failure. During normal operation, record various parameters for stable operation; during overload operation, record abnormal parameters such as increased motor temperature and energy consumption; during bearing failure, record fault characteristics such as increased vibration and noise.

[0113] To enhance the model's generalization ability, different operating environment conditions and external influencing factors are further simulated for each equipment state type. These factors include, but are not limited to, ambient temperature, humidity, power supply voltage fluctuations, and load changes. By adjusting these conditions, more diverse simulation data is generated to cover a wider range of real-world scenarios.

[0114] Taking electric motors as an example again, in addition to simulating the basic state, we can also simulate the operating state of electric motors under high temperature conditions, unstable power supply voltage, and frequent load changes, thereby obtaining a more comprehensive dataset.

[0115] Encoding and processing of analog data:

[0116] One-hot encoding is applied to the discrete features in the simulated dataset. One-hot encoding is a method that converts each value of a discrete feature into a new binary feature, where only one feature is 1 and the rest are 0. This ensures that the model can correctly process discrete data.

[0117] For example, for the discrete feature of motor fault type, suppose there are three possible values: normal, overload, and bearing fault. Through one-hot encoding, it can be converted into three binary features: [1,0,0], [0,1,0], and [0,0,1].

[0118] Standardization of simulation data:

[0119] Continuous features in the simulated dataset were standardized using Z-score normalization. Z-score normalization is a commonly used data normalization method, with the formula: Z = (X - μ) / σ, where X is the original data, μ is the mean of the original data, σ is the standard deviation of the original data, and Z is the standardized data. This method transforms the original data into data that conforms to a standard normal distribution, which helps improve the training effect of the model.

[0120] Taking the temperature data of an electric motor as an example, suppose the average temperature data over a certain period is 50℃ and the standard deviation is 10℃. For a certain temperature data point of 60℃, after Z-score standardization, its value is: Z=(60-50) / 10=1; indicating that the deviation of this data point from the mean is one standard deviation unit.

[0121] The present invention also includes a method for evaluating the condition of power distribution equipment, the method comprising:

[0122] S1: Real-time acquisition of operating data from power distribution equipment;

[0123] S2: Define the state space of the power distribution equipment, and organize and classify the collected data according to the dimensions of electrical parameters, physical state, operating environment and maintenance history.

[0124] S3: Determine the set of actions that may be taken when the power distribution equipment is in an abnormal state;

[0125] S4: Based on equipment reliability, operating efficiency, maintenance costs, power quality, and safety performance, design a reward function to evaluate the results of the actions taken;

[0126] S5: Formulate a strategy that determines the action to be taken in a given state based on the current device state and reward function, and continuously optimize the choice of strategy through an algorithm;

[0127] S6: Employs deep reinforcement learning algorithms and combines them with experience samples accumulated during the policy formulation process to train an intelligent control model, enabling the model to dynamically output control strategies for power distribution equipment based on real-time data.

[0128] S7: Based on a preset evaluation index system, the overall status of the power distribution equipment is quantitatively evaluated. This index system includes at least five aspects: electrical parameter stability, physical health, operating environment adaptability, maintenance history record quality, and intelligent control strategy effectiveness. By calculating the scores of each index and taking them into account, the status evaluation level of the power distribution equipment is output.

[0129] The implementation of this method is the same as described in the above embodiments, and will not be repeated in the specification.

[0130] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0131] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A power distribution equipment condition assessment system, characterized in that, The system includes: The data acquisition module is used to collect real-time operating data of power distribution equipment, including electrical parameters, physical status, operating environment, and maintenance history data. A control strategy output module is used to construct an intelligent control model, which models the operation process of the power distribution equipment as a Markov decision process and dynamically outputs the control strategy for the power distribution equipment by analyzing the operating data of the power distribution equipment in real time. The control strategy output module further includes: The state space definition submodule is used to define the state space of the power distribution equipment, including the electrical parameter dimension, physical state dimension, operating environment dimension, and maintenance history dimension. The Action Space Definition submodule is used to define the set of actions that may be taken when the power distribution equipment is in an abnormal state; The reward function design submodule, connected to the action space definition submodule, is used to design reward functions based on equipment reliability, operating efficiency, maintenance costs, power quality, and safety performance, and to evaluate action results. The strategy formulation submodule, connected to the reward function design submodule, is used to formulate a strategy for selecting actions in a given state, and continuously optimizes the strategy selection through an algorithm to maximize the cumulative reward. The model training submodule, which employs a deep reinforcement learning algorithm, is connected to the policy formulation submodule and is used to train the intelligent control model. The equipment status evaluation module, connected to the data acquisition module, is used to receive real-time collected operating data of the power distribution equipment and to quantitatively evaluate the overall status of the power distribution equipment based on a preset evaluation index system, and output the status evaluation level. The evaluation index system includes at least five aspects: electrical parameter stability, physical health, operating environment adaptability, maintenance history record quality, and the effectiveness of the control strategy output by the intelligent control model.

2. The power distribution equipment condition evaluation system according to claim 1, characterized in that: The electrical parameters include voltage, current, power factor, active power, reactive power, power quality, and harmonic content. The physical state dimensions include the device's temperature, vibration, noise, partial discharge status, and insulation status; The operating environment dimensions include ambient temperature, ambient humidity, atmospheric pressure, and pollution level; The maintenance history dimension includes the equipment's historical repair records, maintenance records, fault types, fault frequencies, and repair times.

3. The power distribution equipment condition evaluation system according to claim 1, characterized in that, The action space definition submodule takes the following set of actions: adjusting electrical parameters, switching equipment operating modes, starting backup equipment, performing equipment maintenance, notifying operators, and scheduling resources.

4. The power distribution equipment condition evaluation system according to claim 1, characterized in that: The model training submodule uses the Deep Q-Network (DQN) algorithm to train the intelligent control model.

5. The power distribution equipment condition evaluation system according to claim 4, characterized in that, The steps for training the intelligent control model by the model training submodule include: Step 1: Initialize model parameters: Set the neural network structure in the deep reinforcement learning algorithm, including the input layer, hidden layer and output layer, as well as the learning rate, discount factor and exploration rate; Step 2: Construct an experience replay buffer: This buffer is used to store experience samples generated by the interaction between the agent and the power distribution equipment environment, including states, actions, rewards, and the next state. Step 3: Collect training data: Prepare the training dataset as input for the intelligent control model; Step 4: Execute the action: Select an action based on the current state and strategy, execute the action through actual control or simulation environment, and observe the next state, reward value, and whether to terminate. Step 5: Store the experience samples: Store the experience samples obtained in Step 4 into the experience replay buffer; Step 6: Sampling from the buffer: Randomly select a batch of experience samples from the experience replay buffer for model training; Step 7: Calculate the target value: For each sampled empirical sample, calculate the target Q value using the Bellman equation. The formula is as follows: Q target =r+γmax a′ Q(s ′ ,a ′ ;θ - ) Among them, Q target Here, γ is the target Q value, r is the reward value, γ is the discount factor, and s is the target Q value. ′ The next state is a ′ The action that can be selected for the next state, Q(s) ′ ,a ′ ;θ - ) is the target network parameter θ - Calculated Q value; Step 8: Update model parameters: Calculate the difference between the current Q-value and the target Q-value using the mean squared error loss function, and update the neural network parameters θ. The loss function formula is: Where N is the number of samples; Step 9, Iterative Training: Repeat steps 4 to 8 until the preset number of training rounds is reached; Step 10: Save the model: Save the parameters of the trained intelligent control model for subsequent power distribution equipment status evaluation.

6. The power distribution equipment condition evaluation system according to claim 5, characterized in that: In step 8, the neural network parameters θ are updated using the gradient descent algorithm. The calculation formula for the gradient descent algorithm is as follows: Where θ represents the model parameters, which include all the weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter updates; This represents the gradient of the loss function J(θ) with respect to the parameter θ. It is a vector pointing in the direction where the loss function grows the fastest; := represents an assignment operation, that is, updating the value of parameter θ.

7. The power distribution equipment condition evaluation system according to claim 1, characterized in that: The reward function design submodule receives the action execution results from the action space definition submodule and calculates the reward value r according to the following formula: r=w1·R+w2·E-w3·C+w4·T+w5·S Where w1, w2, w3, w4, and w5 represent the weighting coefficients of equipment reliability reward, operating efficiency reward, maintenance cost penalty, power quality reward, and safety performance reward, respectively, and satisfy w1+w2+w3+w4+w5=1; R, E, C, T, and S represent the quantitative index values ​​of equipment reliability, operating efficiency, maintenance cost, power quality, and safety performance, respectively.

8. The power distribution equipment condition evaluation system according to claim 5, characterized in that, The training data for the intelligent control model is obtained through the following steps: Based on preset equipment operating status scenarios and evaluation indicators, simulated data containing multiple equipment status types and corresponding evaluation results is generated; the simulated data includes operating parameters, performance indicators, maintenance records, power quality data and safety performance data of the equipment in normal, abnormal and fault states. For each type of equipment status, different operating environment conditions and external influencing factors are simulated to further enrich the simulation dataset; The various types of data in the simulated dataset are encoded and standardized.

9. A power distribution equipment condition evaluation system according to claim 8, characterized in that: The method of encoding the data in the simulated training dataset includes: performing one-hot encoding transformation on discrete features; The method for standardizing the data in the simulated training dataset includes: using the Z-score standardization method to standardize continuous features, specifically: subtracting the mean of each numerical data point and dividing by its standard deviation, so that the processed data conforms to a standard normal distribution. The standardization formula is: Z = (X - μ) / σ, where X is the original data, μ is the mean of the original data, σ is the standard deviation of the original data, and Z is the standardized data.

10. A method for evaluating the condition of power distribution equipment, characterized in that, The method includes: S1: Real-time acquisition of operating data from power distribution equipment; S2: Define the state space of the power distribution equipment, and organize and classify the collected data according to the dimensions of electrical parameters, physical state, operating environment and maintenance history. S3: Determine the set of actions that may be taken when the power distribution equipment is in an abnormal state; S4: Based on equipment reliability, operating efficiency, maintenance costs, power quality, and safety performance, design a reward function to evaluate the results of the actions taken; S5: Formulate a strategy that determines the action to be taken in a given state based on the current device state and reward function, and continuously optimize the choice of strategy through an algorithm; S6: Employs deep reinforcement learning algorithms and combines them with experience samples accumulated during the policy formulation process to train an intelligent control model, enabling the model to dynamically output control strategies for power distribution equipment based on real-time data. S7: Based on a preset evaluation index system, the overall status of the power distribution equipment is quantitatively evaluated. This index system includes at least five aspects: electrical parameter stability, physical health, operating environment adaptability, maintenance history record quality, and intelligent control strategy effectiveness. By calculating the scores of each index and taking them into account, the status evaluation level of the power distribution equipment is output.

Images