技术领域Technical field
本发明属于电力工程领域，具体是一种基于人工智能的发电故障诊断方法。The invention belongs to the field of electric power engineering, and is specifically a power generation fault diagnosis method based on artificial intelligence.
背景技术Background technique
发电故障诊断是对发电系统中可能出现的故障进行故障源定位、故障检测和故障分类的过程，其目的是通过对运行数据、监测信号和维护记录的分析，提前发现潜在的故障，采取相应的维修和保养措施，避免设备的严重故障和停机事故，提高发电系统的可靠性和可用性。但是现有的发电故障诊断，存在发电系统运行数据复杂，难以进行有效的数据存储和查找，对故障诊断和运维效率产生负面影响的技术问题；存在发电故障诊断过程中，对原始数据进行特征提取和选择时难以提取最具代表性的数据特征，无法充分理解特征与发电故障之间的关系，影响诊断结果准确性的技术问题；存在目前发电故障诊断模型是基于静态模型进行建模和预测故障的发生，无法实时感知和适应变化的工作环境，导致诊断结果滞后，发电系统维修成本增加的技术问题；存在黑盒子算法缺乏可解释性，无法解释发电系统内部运行方式，工程师和操作人员难以发现算法漏洞和改进点，无法进行深入分析和优化并做出决策的技术问题。Power generation fault diagnosis is the process of locating fault sources, detecting faults and classifying faults that may occur in the power generation system. Its purpose is to discover potential faults in advance and take corresponding measures through the analysis of operating data, monitoring signals and maintenance records. Repair and maintenance measures can avoid serious equipment failures and shutdown accidents and improve the reliability and availability of power generation systems. However, the existing power generation fault diagnosis has technical problems such as complex power generation system operating data, which makes it difficult to effectively store and search data, and has a negative impact on fault diagnosis and operation and maintenance efficiency. In the process of power generation fault diagnosis, it is necessary to characterize the original data. It is difficult to extract the most representative data features when extracting and selecting, and it is impossible to fully understand the relationship between features and power generation faults, which affects the accuracy of diagnosis results. There are technical issues such as current power generation fault diagnosis models are based on static models for modeling and prediction. The occurrence of faults makes it impossible to sense and adapt to the changing working environment in real time, resulting in technical problems such as delayed diagnostic results and increased maintenance costs for the power generation system; there is a black box algorithm that lacks interpretability and cannot explain the internal operation of the power generation system, making it difficult for engineers and operators Discover algorithm loopholes and improvement points, and technical issues that prevent in-depth analysis and optimization and decision-making.
发明内容Contents of the invention
针对上述情况，为克服现有技术的缺陷，本发明提供了一种基于人工智能的发电故障诊断方法，针对发电系统运行数据复杂，难以进行有效的数据存储和查找，对故障诊断和运维效率产生负面影响的技术问题，采用一致性哈希算法，将存储空间和数据映射到哈希环上，通过对哈希环的划分来确定数据在存储空间的节点位置，减少数据查找范围，实现数据的负载均衡和高效的数据存储，有效提高系统的数据处理效率、容错性和扩展性；针对发电故障诊断过程中，对原始数据进行特征提取和选择时难以提取最具代表性的数据特征，难以充分理解特征与发电故障之间的关系，影响诊断结果准确性的技术问题，采用聚类算法计算特征维度的中心点将数据样本进行分组，并用非线性特征映射函数将原始数据映射到高维特征空间，通过非线性特征映射函数的特征空间进行特征提取，能够减少特征维度，提高特征的区分性和代表性，得到最具代表性的数据特征；针对目前发电故障诊断模型是基于静态模型进行建模和预测故障的发生，无法实时感知和适应变化的工作环境，导致诊断结果滞后，发电系统维修成本增加的技术问题，采用寻找最优策略算法，将发电故障诊断的决策过程作为智能体的行为与环境状态进行交互并根据反馈奖励函数进行学习，根据实时观测到的反馈奖励函数优化诊断过程，在实时性要求高的场景下进行快速反应，减少发电系统维修成本；针对黑盒子算法缺乏可解释性，无法解释内部运行方式，工程师和操作人员难以发现算法漏洞和改进点，无法进行深入分析和优化并做出决策的技术问题，采用规则提取算法提取易于理解的决策规则，根据特征重要性分析和决策树结构解释模型中的决策过程，并用局部解释性模型解释单个样本点的诊断结果，利用可视化手段展开发电故障诊断模型的诊断结果，帮助工程师和操作人员理解发电故障诊断模型的诊断结果，提高工作效率。In view of the above situation, in order to overcome the shortcomings of the existing technology, the present invention provides a power generation fault diagnosis method based on artificial intelligence. In view of the complex operating data of the power generation system, it is difficult to perform effective data storage and search, which has a negative impact on fault diagnosis and operation and maintenance efficiency. For technical issues that have negative impacts, the consistent hash algorithm is used to map the storage space and data to the hash ring. The node location of the data in the storage space is determined by dividing the hash ring, reducing the data search scope and realizing the data Load balancing and efficient data storage effectively improve the data processing efficiency, fault tolerance and scalability of the system; in the process of power generation fault diagnosis, it is difficult to extract the most representative data features when extracting and selecting features from the original data. Fully understand the relationship between features and power generation faults, and the technical issues that affect the accuracy of diagnosis results. Use a clustering algorithm to calculate the center point of the feature dimension to group the data samples, and use a nonlinear feature mapping function to map the original data to high-dimensional features. Space, feature extraction through the feature space of nonlinear feature mapping functions can reduce feature dimensions, improve the distinction and representativeness of features, and obtain the most representative data features; the current power generation fault diagnosis model is based on a static model. Model and predict the occurrence of faults, unable to perceive and adapt to the changing working environment in real time, leading to technical problems such as lag in diagnosis results and increased maintenance costs of the power generation system. The optimal strategy algorithm is adopted to regard the decision-making process of power generation fault diagnosis as the behavior of an intelligent agent. Interact with the environmental state and learn based on the feedback reward function, optimize the diagnosis process based on the real-time observed feedback reward function, react quickly in scenarios with high real-time requirements, and reduce power generation system maintenance costs; for the lack of explainability of black box algorithms It is difficult for engineers and operators to discover algorithm loopholes and improvement points, and it is difficult for engineers and operators to conduct in-depth analysis and optimization and make decisions. The rule extraction algorithm is used to extract easy-to-understand decision-making rules, and the analysis is based on the importance of features. and a decision tree structure to explain the decision-making process in the model, and use a local explanatory model to explain the diagnosis results of a single sample point, and use visualization means to expand the diagnosis results of the power generation fault diagnosis model to help engineers and operators understand the diagnosis results of the power generation fault diagnosis model. Improve work efficiency.
本发明采取的技术方案如下：本发明提供的一种基于人工智能的发电故障诊断方法，该方法包括以下步骤：The technical solution adopted by the present invention is as follows: The present invention provides an artificial intelligence-based power generation fault diagnosis method, which includes the following steps:
步骤S1：数据收集与处理，具体为采集发电故障数据，并进行数据清洗和数据预处理，得到预处理后的发电故障数据，并对预处理后的发电故障数据进行数据标注；Step S1: Data collection and processing, specifically collecting power generation failure data, performing data cleaning and data preprocessing, obtaining preprocessed power generation failure data, and annotating the preprocessed power generation failure data;
步骤S2：数据存储与查找，具体为采用一致性哈希算法，将存储空间和预处理后的发电故障数据映射到哈希环上，通过对哈希环的划分来确定预处理后的发电故障数据在存储空间的节点位置；Step S2: Data storage and search. Specifically, the consistent hash algorithm is used to map the storage space and preprocessed power generation fault data to the hash ring, and the preprocessed power generation fault is determined by dividing the hash ring. The node location of the data in the storage space;
步骤S3：特征提取与选择，具体为采用聚类算法计算特征维度的中心点，根据特征维度的中心点对预处理后的发电故障数据进行分组，并用非线性特征映射函数将原始数据映射到高维特征空间，通过非线性特征映射函数的高维特征空间进行特征提取；Step S3: Feature extraction and selection, specifically using a clustering algorithm to calculate the center point of the feature dimension, grouping the preprocessed power generation fault data according to the center point of the feature dimension, and using a nonlinear feature mapping function to map the original data to high-level features. Dimensional feature space, feature extraction is performed through the high-dimensional feature space of the nonlinear feature mapping function;
步骤S4：建立发电故障诊断模型，具体为采用寻找最优策略算法，将发电故障诊断的决策过程作为智能体的行为与环境状态进行交互并根据反馈奖励函数进行学习，根据实时观测到的反馈奖励函数优化诊断过程，在实时性要求高的场景下进行快速反应；Step S4: Establish a power generation fault diagnosis model, specifically using the optimal strategy algorithm, taking the decision-making process of power generation fault diagnosis as the behavior of the agent, interacting with the environmental state, and learning based on the feedback reward function, and rewarding based on the real-time observed feedback. The function optimizes the diagnosis process and responds quickly in scenarios with high real-time requirements;
步骤S5：发电故障诊断结果解释，具体为使用决策树结构分析发电故障诊断模型中每个特征样本的重要性，并用局部解释性模型解释单个样本点的故障诊断结果，采用决策规则提取算法提取易于理解的决策规则，解释发电故障诊断模型中的决策过程，并利用可视化手段展开发电故障诊断模型的结果，帮助工程师和操作人员理解发电故障诊断模型的结果。Step S5: Interpretation of power generation fault diagnosis results, specifically using a decision tree structure to analyze the importance of each feature sample in the power generation fault diagnosis model, using a local explanatory model to explain the fault diagnosis results of a single sample point, and using a decision rule extraction algorithm to extract easy Understand the decision-making rules, explain the decision-making process in the power generation fault diagnosis model, and use visual means to unfold the results of the power generation fault diagnosis model to help engineers and operators understand the results of the power generation fault diagnosis model.
进一步地，在步骤S1中，所述数据收集与处理，包括以下步骤：Further, in step S1, the data collection and processing includes the following steps:
步骤S11：数据采集，通过数据库查询、传感器和监控设备获取需要的发电故障数据，所述发电故障数据包括发电机运行数据、温度数据、故障日志和报警记录和外部环境监测数据；Step S11: Data collection, obtain the required power generation fault data through database query, sensors and monitoring equipment. The power generation fault data includes generator operating data, temperature data, fault logs and alarm records and external environment monitoring data;
步骤S12：数据预处理，对采集的发电故障数据进行数据清洗和数据预处理，包括消除噪声、填补缺失值和处理异常值，得到预处理后的发电故障数据；Step S12: Data preprocessing, perform data cleaning and data preprocessing on the collected power generation fault data, including eliminating noise, filling in missing values and processing outliers, to obtain preprocessed power generation fault data;
步骤S13：数据标注，将已经发生的故障标记为故障类别，正常运行的数据标记为正常类别，得到训练模型的目标标签。Step S13: Data labeling, mark the faults that have occurred as fault categories, mark the normal operating data as normal categories, and obtain the target label of the training model.
进一步地，在步骤S2中，所述数据存储与查找，包括以下步骤：Further, in step S2, the data storage and search include the following steps:
步骤S21：映射哈希值，通过哈希函数SHA-3将输入的预处理后的发电故障数据映射为固定长度的哈希值，将固定长度的哈希值用作索引进行数据的存储与查找，包括以下步骤：Step S21: Map the hash value. Use the hash function SHA-3 to map the input preprocessed power generation fault data into a fixed-length hash value. Use the fixed-length hash value as an index to store and search the data. , including the following steps:
步骤S211：定义输入是消息M，输出是哈希值H，初始化一个1600位的状态数组S1，并将1600位的状态数组S1划分为5*5的矩阵；Step S211: Define the input as message M, the output as hash value H, initialize a 1600-bit status array S1 , and divide the 1600-bit status array S1 into a 5*5 matrix;
步骤S212：对消息M进行填充，并将填充后的消息M划分为1600位的块；Step S212: Fill the message M and divide the filled message M into 1600-bit blocks;
步骤S213：将每个1600位的块扩展为一个扩展矩阵A，将扩展矩阵A与状态数组S1进行逐位或操作；Step S213: Expand each 1600-bit block into an expansion matrix A, and perform a bitwise OR operation on the expansion matrix A and the state array S1 ;
步骤S214：重复执行12次步骤S213，得到矩阵E；Step S214: Repeat step S213 12 times to obtain matrix E;
步骤S215：将矩阵E展开为一个比特串，取比特串的前缀部分作为最终的哈希值H；Step S215: Expand the matrix E into a bit string, and take the prefix part of the bit string as the final hash value H;
步骤S22：构建哈希环，将哈希空间映射到环上，形成哈希环，每个节点在哈希环上占据一个位置，节点的位置由哈希函数SHA-3计算得到；Step S22: Construct a hash ring, map the hash space to the ring, and form a hash ring. Each node occupies a position on the hash ring, and the position of the node is calculated by the hash function SHA-3;
步骤S23：更新哈希环，当有新的节点加入系统和旧的节点离开系统时，更新哈希环，加入的节点通过哈希值H在哈希环上找到自己的位置，离开的节点则被重新分配到哈希环上的其他位置；Step S23: Update the hash ring. When a new node joins the system and an old node leaves the system, the hash ring is updated. The joining node finds its position on the hash ring through the hash value H, and the leaving node is reallocated to other locations on the hash ring;
步骤S24：数据存储，将要储存的数据进行哈希计算，根据计算结果在哈希环上找到对应的节点位置，将数据存储在相应的节点上，进行数据的备份和冗余存储；Step S24: Data storage, perform hash calculation on the data to be stored, find the corresponding node position on the hash ring according to the calculation result, store the data on the corresponding node, and perform data backup and redundant storage;
步骤S25：数据查找，当需要查找某个数据时，先对数据进行哈希计算，通过哈希值H找到数据在哈希环上的对应位置，再通过一致性哈希算法找到负责存储该数据的节点；Step S25: Data search. When you need to find a certain data, first perform a hash calculation on the data, find the corresponding position of the data on the hash ring through the hash value H, and then use the consistent hash algorithm to find the location responsible for storing the data. node;
步骤S26：数据迁移，当出现节点增加、节点移除和节点发生故障的情况时，用一致性哈希算法将节点上的数据迁移到其他节点上。Step S26: Data migration. When a node is added, a node is removed, or a node fails, the data on the node is migrated to other nodes using a consistent hash algorithm.
进一步地，在步骤S3中，所述特征提取与选择，包括以下步骤：Further, in step S3, the feature extraction and selection includes the following steps:
步骤S31：数据准备，将预处理后的发电故障数据分为训练数据集和测试数据集，定义训练数据集为原始数据，所述原始数据包括预处理后的发电故障数据和训练模型的目标标签；Step S31: Data preparation, divide the preprocessed power generation fault data into a training data set and a test data set, and define the training data set as original data. The original data includes the preprocessed power generation fault data and the target label of the training model. ;
步骤S32：计算特征维度的中心点，计算每个特征维度上的所有样本点的平均值，将所有样本点的平均值作为对应的特征维度的中心点，所用公式如下：Step S32: Calculate the center point of the feature dimension, calculate the average of all sample points on each feature dimension, and use the average of all sample points as the center point of the corresponding feature dimension. The formula used is as follows:
u[j]=（x[1][j]+x[2][j]+……+x[n][j]）/n；u[j]=（x[1][j]+x[2][j]+……+x[n][j])/n;
c[j]=u[j]；c[j]=u[j];
式中，x[i][j]是表示第i个样本点在第j个特征维度上的取值，u[j]是第j个特征维度上所有样本点的平均值，c[j]是第j个特征维度的中心点，n是样本点的数量，i是[1，n]之间的整数；In the formula, x[i][j] represents the value of the i-th sample point in the j-th feature dimension, u[j] is the average of all sample points in the j-th feature dimension, c[j] is the center point of the j-th feature dimension, n is the number of sample points, and i is an integer between [1, n];
步骤S33：计算映射结果，使用非线性特征映射函数将每个样本点与其特征维度的中心点的距离映射到高维特征空间，从而引入非线性关系，得到样本点在高维特征空间中的映射结果，所用公式如下：Step S33: Calculate the mapping result, and use the nonlinear feature mapping function to map the distance between each sample point and the center point of its feature dimension to the high-dimensional feature space, thereby introducing a nonlinear relationship and obtaining the mapping of the sample point in the high-dimensional feature space. As a result, the formula used is as follows:
； ;
式中，x1是原始数据的特征向量，Γ是非线性特征映射函数中心，ε是控制非线性特征映射函数宽度的参数，是原始数据的特征向量和非线性特征映射函数中心的欧式距离的平方，Ψ（x）是样本点在高维特征空间中的映射结果；In the formula, x1 is the eigenvector of the original data, Γ is the center of the nonlinear feature mapping function, ε is the parameter that controls the width of the nonlinear feature mapping function, is the square of the Euclidean distance between the feature vector of the original data and the center of the nonlinear feature mapping function, and Ψ(x) is the mapping result of the sample point in the high-dimensional feature space;
步骤S34：特征表示，将每个样本点在高维特征空间中的映射结果作为样本点的新特征，再将每个样本点的新特征与原始数据的特征进行合并，得到特征向量α；Step S34: Feature representation, using the mapping result of each sample point in the high-dimensional feature space as the new feature of the sample point, and then merging the new features of each sample point with the features of the original data to obtain the feature vector α;
步骤S35：特征提取模型训练，通过特征向量α和训练模型的目标标签训练特征提取模型，包括以下步骤：Step S35: Feature extraction model training. Training the feature extraction model through the feature vector α and the target label of the training model includes the following steps:
步骤S351：训练集准备，将特征向量α与训练模型的目标标签进行配对，组成训练集；Step S351: Training set preparation, pairing the feature vector α with the target label of the training model to form a training set;
步骤S352：特征标准化，对训练集进行特征标准化操作，将训练集的特征的值域映射到同一范围内，得到标准训练集，所用公式如下：Step S352: Feature standardization, perform feature standardization operation on the training set, map the value range of the features of the training set to the same range, and obtain a standard training set. The formula used is as follows:
y=（x-xmin）/（xmax-xmin）；y=(xxmin )/(xmax -xmin );
式中，y是特征标准化后的值，y的取值范围在[0，1]之间，x是训练集的特征，xmin是训练集的特征的最小值，xmax是训练集的特征的最大值；In the formula, y is the value of the feature after normalization, the value range of y is between [0, 1], x is the feature of the training set, xmin is the minimum value of the feature of the training set, xmax is the feature of the training set the maximum value;
步骤S353：计算拉格朗日乘子，通过序列最小优化算法得到拉格朗日乘子，所用公式如下：Step S353: Calculate the Lagrange multiplier, and obtain the Lagrange multiplier through the sequence minimum optimization algorithm. The formula used is as follows:
； ;
式中，W（α）是支持向量机的目标函数，xi和xj是训练集中的训练样本，yi是xi的标签，yj是xj的标签，αi是待求解的拉格朗日乘子，αi范围是，C是松弛因子，αi和yi满足公式/>，K（xi，xj）是核函数；In the formula, W (α) is the objective function of the support vector machine, xi and xj are the training samples in the training set, yi is the label of xi , yj is the label of xj , αi is the pull to be solved Granger multiplier, the range of αi is , C is the relaxation factor, αi and yi satisfy the formula/> , K (xi , xj ) is the kernel function;
步骤S354：构建分类决策函数，引入拉格朗日乘子作为乘法因子，再计算支持向量，得到分类决策函数；Step S354: Construct a classification decision function, introduce Lagrange multipliers as multiplication factors, and then calculate support vectors to obtain a classification decision function;
步骤S355：分类决策函数通过支持向量的权重计算特征的重要性，进行特征筛选，得到最具代表性的数据特征。Step S355: The classification decision function calculates the importance of features through the weight of the support vector, performs feature screening, and obtains the most representative data features.
进一步地，在步骤S4中，所述建立发电故障诊断模型，包括以下步骤：Further, in step S4, establishing a power generation fault diagnosis model includes the following steps:
步骤S41：状态表示，根据发电系统的特点和诊断的故障类别，设计发电系统的状态表示；Step S41: State representation. Design the state representation of the power generation system according to the characteristics of the power generation system and the diagnosed fault category;
步骤S42：定义反馈奖励函数，定义反馈奖励函数是R（s，a，s’），表示在状态s下采取动作a后转移到状态s’时的反馈奖励函数；Step S42: Define the feedback reward function. The feedback reward function is defined as R(s, a, s’), which represents the feedback reward function when taking action a in state s and then transitioning to state s’;
步骤S43：构建强化学习环境，将发电系统的状态表示和最具代表性的数据特征作为环境状态，将发电故障诊断的决策过程作为智能体的行为与环境状态进行交互并根据反馈奖励函数进行学习；Step S43: Construct a reinforcement learning environment, use the state representation of the power generation system and the most representative data features as the environmental state, use the decision-making process of power generation fault diagnosis as the behavior of the agent to interact with the environmental state, and learn based on the feedback reward function ;
步骤S44：构建发电故障诊断模型，使用深度神经网络作为发电故障诊断模型的函数近似器，学习最佳的决策策略；Step S44: Construct a power generation fault diagnosis model, use a deep neural network as a function approximator of the power generation fault diagnosis model, and learn the best decision-making strategy;
步骤S45：发电故障诊断模型训练，使用寻找最优策略算法，将发电故障诊断的决策过程作为智能体的行为通过与环境状态的交互进行发电故障诊断模型的训练和优化，包括以下步骤：Step S45: Training of the power generation fault diagnosis model. Using the optimal strategy algorithm, the decision-making process of power generation fault diagnosis is regarded as the behavior of the agent through interaction with the environmental state to train and optimize the power generation fault diagnosis model, including the following steps:
步骤S451：初始化Z值函数，将每一对状态——行动对的Z值初始化为0；Step S451: Initialize the Z value function, and initialize the Z value of each pair of state-action pairs to 0;
步骤S452：环境交互，根据最佳的决策策略选择行动，与环境状态进行交互，观察反馈奖励函数和下一状态；Step S452: Environment interaction, select actions according to the best decision-making strategy, interact with the environment state, and observe the feedback reward function and next state;
步骤S453：更新Z值函数，所用公式如下：Step S453: Update the Z value function, the formula used is as follows:
Z（s，a）=Z（s，a）+θ*（R+γ*max（Z（s’，a’））-Z（s，a））；Z(s,a)=Z(s,a)+θ*(R+γ*max(Z(s’,a’))-Z(s,a));
式中，Z（s，a）是当前状态——行动对的Z值，θ是学习率，R是当前反馈奖励函数，γ是折扣因子，s’是下一状态，a’是下一状态的最优行动；In the formula, Z (s, a) is the Z value of the current state-action pair, θ is the learning rate, R is the current feedback reward function, γ is the discount factor, s' is the next state, and a' is the next state. optimal action;
步骤S454：重复步骤S452和步骤S453，逐渐降低学习率θ，直到Z值函数收敛，求出一个最优Z值函数，得到使预期回报最大化的最优策略，并根据最优策略进行决策和行动选择；Step S454: Repeat steps S452 and S453, gradually reduce the learning rate θ until the Z-value function converges, find an optimal Z-value function, obtain the optimal strategy that maximizes expected returns, and make decisions based on the optimal strategy. action choice;
步骤S46：实时故障诊断，发电故障诊断模型经过训练优化后部署到实时环境中，根据实时采集到的数据进行故障预测和诊断，提供故障预警和维修建议。Step S46: Real-time fault diagnosis. The power generation fault diagnosis model is deployed in the real-time environment after training and optimization. It performs fault prediction and diagnosis based on the data collected in real time, and provides fault warning and maintenance suggestions.
进一步地，在步骤S5中，所述发电故障诊断结果解释，包括以下步骤：Further, in step S5, the power generation fault diagnosis result interpretation includes the following steps:
步骤S51：特征重要性分析，使用决策树结构来分析发电故障诊断模型中每个特征样本的重要性，并计算每个特征样本的基尼系数来评估特征的重要性，所用公式如下：Step S51: Feature importance analysis, use the decision tree structure to analyze the importance of each feature sample in the power generation fault diagnosis model, and calculate the Gini coefficient of each feature sample to evaluate the importance of the feature. The formula used is as follows:
； ;
式中，Gini（p）是特征样本p的基尼系数，pd是第d类特征样本在特征样本p上的占比；In the formula, Gini(p) is the Gini coefficient of feature sample p, pd is the proportion of feature samples of type d in feature sample p;
步骤S52：解释故障诊断结果，构建局部解释性模型，使用线性回归模型获取诊断结果的特征权重，并解释单个样本点的故障诊断结果，包括以下步骤：Step S52: Interpret the fault diagnosis results, build a local explanatory model, use the linear regression model to obtain the feature weight of the diagnosis results, and interpret the fault diagnosis results of a single sample point, including the following steps:
步骤S521：通过特征重要性分析从所有特征样本中选择一部分对故障诊断有重要影响的重要特征；Step S521: Select some important features that have an important impact on fault diagnosis from all feature samples through feature importance analysis;
步骤S522：使用发电故障诊断模型对单个样本点进行故障诊断，得到样本点的故障诊断结果；Step S522: Use the power generation fault diagnosis model to perform fault diagnosis on a single sample point, and obtain the fault diagnosis result of the sample point;
步骤S523：样本点附近采样，从单个样本点附近进行采样，得到一组临近样本点；Step S523: Sampling near the sample point, sampling from the vicinity of a single sample point, and obtaining a group of adjacent sample points;
步骤S524：构建局部解释性模型，使用线性回归模型，将单个样本点的特征和临近样本点的特征数据作为输入，将样本点的故障诊断结果作为输出，进行局部解释性模型训练；Step S524: Construct a local explanatory model, use a linear regression model, use the characteristics of a single sample point and the characteristic data of adjacent sample points as input, use the fault diagnosis results of the sample points as output, and perform local explanatory model training;
步骤S525：通过线性回归模型的系数获取样本点的故障诊断结果的特征权重，并解释该样本点的故障诊断结果；Step S525: Obtain the characteristic weight of the fault diagnosis result of the sample point through the coefficient of the linear regression model, and interpret the fault diagnosis result of the sample point;
步骤S53：解释决策过程，使用决策规则提取算法从发电故障诊断模型中提取易于理解的决策规则，解释发电故障诊断模型中的决策过程，包括以下步骤：Step S53: Explain the decision-making process, use the decision rule extraction algorithm to extract easy-to-understand decision rules from the power generation fault diagnosis model, and explain the decision-making process in the power generation fault diagnosis model, including the following steps:
步骤S531：特征选择，输入要解释的发电故障诊断模型，使用基尼系数从发电故障诊断模型中选择与样本点的故障诊断结果有关的特征Q；Step S531: Feature selection, input the power generation fault diagnosis model to be explained, and use the Gini coefficient to select the feature Q related to the fault diagnosis result of the sample point from the power generation fault diagnosis model;
步骤S532：决策规则生成，利用关联规则挖掘方法挖掘出故障诊断结果和特征Q之间的关联，生成决策规则；Step S532: Generate decision rules, use the association rule mining method to mine the association between the fault diagnosis results and the feature Q, and generate decision rules;
步骤S533：决策规则评估，根据样本点的故障诊断结果与训练模型的目标标签的一致性对决策规则的准确度进行评估，根据决策规则的长度和可读性对决策规则的解释度进行评估；Step S533: Decision rule evaluation: evaluate the accuracy of the decision rule based on the consistency of the fault diagnosis results of the sample points and the target label of the training model, and evaluate the interpretability of the decision rule based on the length and readability of the decision rule;
步骤S534：决策规则筛选，应用启发式策略进行决策规则筛选，利用特征Q的局部信息指导搜索方向，根据搜索方向寻找重要特征，根据重要特征对决策规则进行筛选，得到筛选后的决策规则；Step S534: Decision rule screening, apply heuristic strategy to decision rule screening, use the local information of feature Q to guide the search direction, find important features according to the search direction, screen the decision rules according to the important features, and obtain the filtered decision rules;
步骤S535：解释决策规则，将筛选后的决策规则用可视化技术对发电故障诊断模型结果进行展开，将决策规则转化为易于理解的自然语言和图形化展示；Step S535: Explain the decision rules, use visualization technology to expand the power generation fault diagnosis model results after screening, and convert the decision rules into easy-to-understand natural language and graphical display;
步骤S54：交叉验证与评估优化，使用交叉验证来验证决策规则的泛化能力，对决策规则的解释性能进行定性分析，包括以下步骤：Step S54: Cross-validation and evaluation optimization, use cross-validation to verify the generalization ability of the decision rule, and conduct a qualitative analysis of the interpretation performance of the decision rule, including the following steps:
步骤S541：W折交叉验证，将测试数据集分为W个大小相同的子集，称为折，对于每个折，使用自己作为验证折，剩下的W-1个折作为训练折，对于每一次交叉验证的训练集，利用决策规则提取算法从训练折中提取决策规则，将提取的决策规则应用于对应的验证折，评估决策规则的泛化能力；Step S541: W-fold cross-validation, divide the test data set into W subsets of the same size, called folds. For each fold, use itself as the validation fold, and the remaining W-1 folds as training folds. For For each cross-validation training set, the decision rule extraction algorithm is used to extract decision rules from the training fold, and the extracted decision rules are applied to the corresponding validation fold to evaluate the generalization ability of the decision rules;
步骤S542：对于W折交叉验证的结果，计算准确率、召回率和F1值，对决策规则的整体性能进行评估；Step S542: For the results of W-fold cross-validation, calculate the accuracy rate, recall rate and F1 value, and evaluate the overall performance of the decision rule;
步骤S544：分析优化，对决策规则的解释性能进行定性分析，观察决策规则提取算法是否对决策规则提供清晰的解释以及决策规则是否符合领域专家的知识，收集反馈和意见并及时调整改进。Step S544: Analyze and optimize, perform a qualitative analysis on the interpretation performance of the decision rule, observe whether the decision rule extraction algorithm provides a clear explanation of the decision rule and whether the decision rule conforms to the knowledge of domain experts, collect feedback and opinions, and make timely adjustments and improvements.
采用上述方案本发明取得的有益成果如下：The beneficial results achieved by the present invention using the above scheme are as follows:
（1）针对发电系统运行数据复杂，难以进行有效的数据存储和查找，对故障诊断和运维效率产生负面影响的技术问题，采用一致性哈希算法，将存储空间和数据映射到哈希环上，通过对哈希环的划分来确定数据在存储空间的节点位置，减少数据查找范围，实现数据的负载均衡和高效的数据存储，有效提高系统的数据处理效率、容错性和扩展性；(1) In view of the technical issues that the power generation system operating data is complex, making it difficult to effectively store and search data, and negatively affecting fault diagnosis and operation and maintenance efficiency, a consistent hash algorithm is used to map the storage space and data to a hash ring. By dividing the hash ring to determine the node location of the data in the storage space, reduce the data search scope, achieve data load balancing and efficient data storage, and effectively improve the data processing efficiency, fault tolerance and scalability of the system;
（2）针对发电故障诊断过程中，对原始数据进行特征提取和选择时难以提取最具代表性的数据特征，难以充分理解特征与发电故障之间的关系，影响诊断结果准确性的技术问题，采用聚类算法计算特征维度的中心点将数据样本进行分组，并用非线性特征映射函数将原始数据映射到高维特征空间，通过非线性特征映射函数的特征空间进行特征提取，能够减少特征维度，提高特征的区分性和代表性，得到最具代表性的数据特征；(2) In the process of power generation fault diagnosis, it is difficult to extract the most representative data features when performing feature extraction and selection on original data, and it is difficult to fully understand the relationship between features and power generation faults, which affects the accuracy of diagnosis results. A clustering algorithm is used to calculate the center point of the feature dimension to group the data samples, and a nonlinear feature mapping function is used to map the original data to a high-dimensional feature space. Feature extraction through the feature space of the nonlinear feature mapping function can reduce the feature dimension. Improve the distinction and representativeness of features and obtain the most representative data features;
（3）针对目前发电故障诊断模型是基于静态模型进行建模和预测故障的发生，无法实时感知和适应变化的工作环境，导致诊断结果滞后，发电系统维修成本增加的技术问题，采用寻找最优策略算法，将发电故障诊断的决策过程作为智能体的行为与环境状态进行交互并根据反馈奖励函数进行学习，根据实时观测到的反馈奖励函数优化诊断过程，在实时性要求高的场景下进行快速反应，减少发电系统维修成本；(3) In view of the technical problems that the current power generation fault diagnosis model is based on a static model to model and predict the occurrence of faults and cannot perceive and adapt to the changing working environment in real time, resulting in delayed diagnosis results and increased maintenance costs of the power generation system, the search for the optimal The strategy algorithm uses the decision-making process of power generation fault diagnosis as the behavior of the agent to interact with the environmental state and learn based on the feedback reward function. It optimizes the diagnosis process based on the feedback reward function observed in real time and performs rapid operations in scenarios with high real-time requirements. response to reduce power generation system maintenance costs;
（4）针对黑盒子算法缺乏可解释性，无法解释内部运行方式，工程师和操作人员难以发现算法漏洞和改进点，无法进行深入分析和优化并做出决策的技术问题，采用规则提取算法提取易于理解的决策规则，根据特征重要性分析和决策树结构解释模型中的决策过程，并用局部解释性模型解释单个样本点的诊断结果，利用可视化手段展开发电故障诊断模型的诊断结果，帮助工程师和操作人员理解发电故障诊断模型的诊断结果，提高工作效率。(4) In view of the technical problems that the black box algorithm lacks interpretability and cannot explain the internal operation mode, and it is difficult for engineers and operators to discover algorithm loopholes and improvement points, and cannot conduct in-depth analysis and optimization and make decisions, the rule extraction algorithm is used to extract easy-to-use Understand the decision rules, explain the decision-making process in the model based on feature importance analysis and decision tree structure, and use a local explanatory model to explain the diagnosis results of a single sample point, and use visualization means to expand the diagnosis results of the power generation fault diagnosis model to help engineers and operators Personnel understand the diagnosis results of the power generation fault diagnosis model and improve work efficiency.
附图说明Description of the drawings
图1为本发明提供的一种基于人工智能的发电故障诊断方法的流程示意图；Figure 1 is a schematic flow chart of an artificial intelligence-based power generation fault diagnosis method provided by the present invention;
图2为步骤S2的流程示意图；Figure 2 is a schematic flow chart of step S2;
图3为步骤S3的流程示意图；Figure 3 is a schematic flow chart of step S3;
图4为步骤S4的流程示意图；Figure 4 is a schematic flow chart of step S4;
图5为步骤S5的流程示意图。Figure 5 is a schematic flow chart of step S5.
附图用来提供对本发明的进一步理解，并且构成说明书的一部分，与本发明的实施例一起用于解释本发明，并不构成对本发明的限制。The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.
具体实施方式Detailed ways
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例；基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them; based on The embodiments of the present invention and all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.
在本发明的描述中，需要理解的是，术语“上”、“下”、“前”、“后”、“左”、“右”、“顶”、“底”、“内”、“外”等指示方位或位置关系为基于附图所示的方位或位置关系，仅是为了便于描述本发明和简化描述，而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作，因此不能理解为对本发明的限制。In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inner", " The orientation or positional relationship indicated by "outside" is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description. It does not indicate or imply that the device or element referred to must have a specific orientation or a specific location. orientation, construction and operation, and therefore should not be construed as limitations of the present invention.
实施例一，参阅图1，本发明提供的一种基于人工智能的发电故障诊断方法，该方法包括以下步骤：Embodiment 1. Referring to Figure 1, the present invention provides an artificial intelligence-based power generation fault diagnosis method. The method includes the following steps:
步骤S1：数据收集与处理，具体为采集发电故障数据，并进行数据清洗和数据预处理，得到预处理后的发电故障数据，并对预处理后的发电故障数据进行数据标注；Step S1: Data collection and processing, specifically collecting power generation failure data, performing data cleaning and data preprocessing, obtaining preprocessed power generation failure data, and annotating the preprocessed power generation failure data;
步骤S2：数据存储与查找，具体为采用一致性哈希算法，将存储空间和预处理后的发电故障数据映射到哈希环上，通过对哈希环的划分来确定预处理后的发电故障数据在存储空间的节点位置；Step S2: Data storage and search. Specifically, the consistent hash algorithm is used to map the storage space and preprocessed power generation fault data to the hash ring, and the preprocessed power generation fault is determined by dividing the hash ring. The node location of the data in the storage space;
步骤S3：特征提取与选择，具体为采用聚类算法计算特征维度的中心点，根据特征维度的中心点对预处理后的发电故障数据进行分组，并用非线性特征映射函数将原始数据映射到高维特征空间，通过非线性特征映射函数的高维特征空间进行特征提取；Step S3: Feature extraction and selection, specifically using a clustering algorithm to calculate the center point of the feature dimension, grouping the preprocessed power generation fault data according to the center point of the feature dimension, and using a nonlinear feature mapping function to map the original data to high-level features. Dimensional feature space, feature extraction is performed through the high-dimensional feature space of the nonlinear feature mapping function;
步骤S4：建立发电故障诊断模型，具体为采用寻找最优策略算法，将发电故障诊断的决策过程作为智能体的行为与环境状态进行交互并根据反馈奖励函数进行学习，根据实时观测到的反馈奖励函数优化诊断过程，在实时性要求高的场景下进行快速反应；Step S4: Establish a power generation fault diagnosis model, specifically using the optimal strategy algorithm, taking the decision-making process of power generation fault diagnosis as the behavior of the agent, interacting with the environmental state, and learning based on the feedback reward function, and rewarding based on the real-time observed feedback. The function optimizes the diagnosis process and responds quickly in scenarios with high real-time requirements;
步骤S5：发电故障诊断结果解释，具体为使用决策树结构分析发电故障诊断模型中每个特征样本的重要性，并用局部解释性模型解释单个样本点的故障诊断结果，采用决策规则提取算法提取易于理解的决策规则，解释发电故障诊断模型中的决策过程，并利用可视化手段展开发电故障诊断模型的结果，帮助工程师和操作人员理解发电故障诊断模型的结果。Step S5: Interpretation of power generation fault diagnosis results, specifically using a decision tree structure to analyze the importance of each feature sample in the power generation fault diagnosis model, using a local explanatory model to explain the fault diagnosis results of a single sample point, and using a decision rule extraction algorithm to extract easy Understand the decision-making rules, explain the decision-making process in the power generation fault diagnosis model, and use visual means to unfold the results of the power generation fault diagnosis model to help engineers and operators understand the results of the power generation fault diagnosis model.
实施例二，参阅图1，该实施例基于上述实施例，在步骤S1中，所述数据收集与处理，包括以下步骤：Embodiment 2. Refer to Figure 1. This embodiment is based on the above embodiment. In step S1, the data collection and processing includes the following steps:
步骤S11：数据采集，通过数据库查询、传感器和监控设备获取需要的发电故障数据，所述发电故障数据包括发电机运行数据、温度数据、故障日志和报警记录和外部环境监测数据；Step S11: Data collection, obtain the required power generation fault data through database query, sensors and monitoring equipment. The power generation fault data includes generator operating data, temperature data, fault logs and alarm records and external environment monitoring data;
步骤S12：数据预处理，对采集的发电故障数据进行数据清洗和数据预处理，包括消除噪声、填补缺失值和处理异常值，得到预处理后的发电故障数据；Step S12: Data preprocessing, perform data cleaning and data preprocessing on the collected power generation fault data, including eliminating noise, filling in missing values and processing outliers, to obtain preprocessed power generation fault data;
步骤S13：数据标注，将已经发生的故障标记为故障类别，正常运行的数据标记为正常类别，得到训练模型的目标标签。Step S13: Data labeling, mark the faults that have occurred as fault categories, mark the normal operating data as normal categories, and obtain the target label of the training model.
实施例三，参阅图1和图2，该实施例基于上述实施例，在步骤S2中，所述数据存储与查找，包括以下步骤：Embodiment 3. Refer to Figures 1 and 2. This embodiment is based on the above embodiment. In step S2, the data storage and search include the following steps:
步骤S21：映射哈希值，通过哈希函数SHA-3将输入的预处理后的发电故障数据映射为固定长度的哈希值，将固定长度的哈希值用作索引进行数据的存储与查找，包括以下步骤：Step S21: Map the hash value. Use the hash function SHA-3 to map the input preprocessed power generation fault data into a fixed-length hash value. Use the fixed-length hash value as an index to store and search the data. , including the following steps:
步骤S211：定义输入是消息M，输出是哈希值H，初始化一个1600位的状态数组S1，并将1600位的状态数组S1划分为5*5的矩阵；Step S211: Define the input as message M, the output as hash value H, initialize a 1600-bit status array S1 , and divide the 1600-bit status array S1 into a 5*5 matrix;
步骤S212：对消息M进行填充，并将填充后的消息M划分为1600位的块；Step S212: Fill the message M and divide the filled message M into 1600-bit blocks;
步骤S213：将每个1600位的块扩展为一个扩展矩阵A，将扩展矩阵A与状态数组S1进行逐位或操作；Step S213: Expand each 1600-bit block into an expansion matrix A, and perform a bitwise OR operation on the expansion matrix A and the state array S1 ;
步骤S214：重复执行12次步骤S213，得到矩阵E；Step S214: Repeat step S213 12 times to obtain matrix E;
步骤S215：将矩阵E展开为一个比特串，取比特串的前缀部分作为最终的哈希值H；Step S215: Expand the matrix E into a bit string, and take the prefix part of the bit string as the final hash value H;
步骤S22：构建哈希环，将哈希空间映射到环上，形成哈希环，每个节点在哈希环上占据一个位置，节点的位置由哈希函数SHA-3计算得到；Step S22: Construct a hash ring, map the hash space to the ring, and form a hash ring. Each node occupies a position on the hash ring, and the position of the node is calculated by the hash function SHA-3;
步骤S23：更新哈希环，当有新的节点加入系统和旧的节点离开系统时，更新哈希环，加入的节点通过哈希值H在哈希环上找到自己的位置，离开的节点则被重新分配到哈希环上的其他位置；Step S23: Update the hash ring. When a new node joins the system and an old node leaves the system, the hash ring is updated. The joining node finds its position on the hash ring through the hash value H, and the leaving node is reallocated to other locations on the hash ring;
步骤S24：数据存储，将要储存的数据进行哈希计算，根据计算结果在哈希环上找到对应的节点位置，将数据存储在相应的节点上，进行数据的备份和冗余存储；Step S24: Data storage, perform hash calculation on the data to be stored, find the corresponding node position on the hash ring according to the calculation result, store the data on the corresponding node, and perform data backup and redundant storage;
步骤S25：数据查找，当需要查找某个数据时，先对数据进行哈希计算，通过哈希值H找到数据在哈希环上的对应位置，再通过一致性哈希算法找到负责存储该数据的节点；Step S25: Data search. When you need to find a certain data, first perform a hash calculation on the data, find the corresponding position of the data on the hash ring through the hash value H, and then use the consistent hash algorithm to find the location responsible for storing the data. node;
步骤S26：数据迁移，当出现节点增加、节点移除和节点发生故障的情况时，用一致性哈希算法将节点上的数据迁移到其他节点上。Step S26: Data migration. When a node is added, a node is removed, or a node fails, the data on the node is migrated to other nodes using a consistent hash algorithm.
在上述操作中，本方案采用一致性哈希算法，将存储空间和数据映射到哈希环上，通过对哈希环的划分来确定数据在存储空间的节点位置，减少数据查找范围，实现数据的负载均衡和高效的数据存储，有效提高系统的数据处理效率、容错性和扩展性，解决了发电系统运行数据复杂，难以进行有效的数据存储和查找，对故障诊断和运维效率产生负面影响的技术问题。In the above operations, this solution uses a consistent hash algorithm to map the storage space and data to the hash ring. By dividing the hash ring, the node location of the data in the storage space is determined, reducing the data search range and realizing the data Load balancing and efficient data storage effectively improve the data processing efficiency, fault tolerance and scalability of the system. It solves the problem of complex operating data of the power generation system, which makes effective data storage and search difficult, and has a negative impact on fault diagnosis and operation and maintenance efficiency. technical issues.
实施例四，参阅图1和图3，该实施例基于上述实施例，在步骤S3中，所述特征提取与选择，包括以下步骤：Embodiment 4. Refer to Figures 1 and 3. This embodiment is based on the above embodiment. In step S3, the feature extraction and selection includes the following steps:
步骤S31：数据准备，将预处理后的发电故障数据分为训练数据集和测试数据集，定义训练数据集为原始数据，所述原始数据包括预处理后的发电故障数据和训练模型的目标标签；Step S31: Data preparation, divide the preprocessed power generation fault data into a training data set and a test data set, and define the training data set as original data. The original data includes the preprocessed power generation fault data and the target label of the training model. ;
步骤S32：计算特征维度的中心点，计算每个特征维度上的所有样本点的平均值，将所有样本点的平均值作为对应的特征维度的中心点，所用公式如下：Step S32: Calculate the center point of the feature dimension, calculate the average of all sample points on each feature dimension, and use the average of all sample points as the center point of the corresponding feature dimension. The formula used is as follows:
u[j]=（x[1][j]+x[2][j]+……+x[n][j]）/n；u[j]=（x[1][j]+x[2][j]+……+x[n][j])/n;
c[j]=u[j]；c[j]=u[j];
式中，x[i][j]是表示第i个样本点在第j个特征维度上的取值，u[j]是第j个特征维度上所有样本点的平均值，c[j]是第j个特征维度的中心点，n是样本点的数量；In the formula, x[i][j] represents the value of the i-th sample point in the j-th feature dimension, u[j] is the average of all sample points in the j-th feature dimension, c[j] is the center point of the j-th feature dimension, n is the number of sample points;
步骤S33：计算映射结果，使用非线性特征映射函数将每个样本与其特征维度的中心点的距离映射到高维特征空间，从而引入非线性关系，得到样本点在高维特征空间中的映射结果，所用公式如下：Step S33: Calculate the mapping result, and use the nonlinear feature mapping function to map the distance between each sample and the center point of its feature dimension to the high-dimensional feature space, thereby introducing a nonlinear relationship and obtaining the mapping result of the sample point in the high-dimensional feature space. , the formula used is as follows:
； ;
式中，x1是原始数据的特征向量，Γ是非线性特征映射函数中心，ε是控制非线性特征映射函数宽度的参数，是原始数据的特征向量和非线性特征映射函数中心的欧式距离的平方，Ψ（x）是样本点在高维特征空间中的映射结果；In the formula, x1 is the eigenvector of the original data, Γ is the center of the nonlinear feature mapping function, ε is the parameter that controls the width of the nonlinear feature mapping function, is the square of the Euclidean distance between the feature vector of the original data and the center of the nonlinear feature mapping function, and Ψ(x) is the mapping result of the sample point in the high-dimensional feature space;
步骤S34：特征表示，将每个样本点在高维特征空间中的映射结果作为样本点的新特征，再将每个样本点的新特征与原始数据的特征进行合并，得到特征向量α；Step S34: Feature representation, using the mapping result of each sample point in the high-dimensional feature space as the new feature of the sample point, and then merging the new features of each sample point with the features of the original data to obtain the feature vector α;
步骤S35：特征提取模型训练，通过特征向量α和训练模型的目标标签训练特征提取模型，包括以下步骤：Step S35: Feature extraction model training. Training the feature extraction model through the feature vector α and the target label of the training model includes the following steps:
步骤S351：训练集准备，将特征向量α与训练模型的目标标签进行配对，组成训练集；Step S351: Training set preparation, pairing the feature vector α with the target label of the training model to form a training set;
步骤S352：特征标准化，对训练集进行特征标准化操作，将训练集的特征的值域映射到同一范围内，得到标准训练集，所用公式如下：Step S352: Feature standardization, perform feature standardization operation on the training set, map the value range of the features of the training set to the same range, and obtain a standard training set. The formula used is as follows:
y=（x-xmin）/（xmax-xmin）；y=(xxmin )/(xmax -xmin );
式中，y是特征标准化后的值，y的取值范围在[0，1]之间，x是训练集的特征，xmin是训练集的特征的最小值，xmax是训练集的特征的最大值；In the formula, y is the value of the feature after normalization, the value range of y is between [0, 1], x is the feature of the training set, xmin is the minimum value of the feature of the training set, xmax is the feature of the training set the maximum value;
步骤S353：计算拉格朗日乘子，通过序列最小优化算法得到拉格朗日乘子，所用公式如下：Step S353: Calculate the Lagrange multiplier, and obtain the Lagrange multiplier through the sequence minimum optimization algorithm. The formula used is as follows:
； ;
式中，W（α）是支持向量机的目标函数，xi和xj是训练集中的训练样本，yi是xi的标签，yj是xj的标签，αi是待求解的拉格朗日乘子，αi范围是，C是松弛因子，αi和yi满足公式/>，K（xi，xj）是核函数；In the formula, W (α) is the objective function of the support vector machine, xi and xj are the training samples in the training set, yi is the label of xi , yj is the label of xj , αi is the pull to be solved Granger multiplier, the range of αi is , C is the relaxation factor, αi and yi satisfy the formula/> , K (xi , xj ) is the kernel function;
步骤S354：构建分类决策函数，引入拉格朗日乘子作为乘法因子，再计算支持向量，得到分类决策函数；Step S354: Construct a classification decision function, introduce Lagrange multipliers as multiplication factors, and then calculate support vectors to obtain a classification decision function;
步骤S355：分类决策函数通过支持向量的权重计算特征的重要性，进行特征筛选，得到最具代表性的数据特征。Step S355: The classification decision function calculates the importance of features through the weight of the support vector, performs feature screening, and obtains the most representative data features.
在上述操作中，本方案采用聚类算法计算特征维度的中心点将数据样本进行分组，并用非线性特征映射函数将原始数据映射到高维特征空间，通过非线性特征映射函数的特征空间进行特征提取，能够减少特征维度，提高特征的区分性和代表性，得到最具代表性的数据特征，解决了发电故障诊断过程中，对原始数据进行特征提取和选择时难以提取最具代表性的数据特征，难以充分理解特征与发电故障之间的关系，影响诊断结果准确性的技术问题。In the above operation, this program uses a clustering algorithm to calculate the center point of the feature dimension to group the data samples, and uses a nonlinear feature mapping function to map the original data to a high-dimensional feature space, and perform feature extraction through the feature space of the nonlinear feature mapping function. Extraction can reduce feature dimensions, improve the distinction and representativeness of features, and obtain the most representative data features. This solves the problem of difficulty in extracting the most representative data when extracting and selecting features from original data during the power generation fault diagnosis process. Features, it is difficult to fully understand the relationship between features and power generation faults, and technical issues that affect the accuracy of diagnostic results.
实施例五，参阅图1和图4，该实施例基于上述实施例，在步骤S4中，所述建立发电故障诊断模型，包括以下步骤：Embodiment 5. Refer to Figures 1 and 4. This embodiment is based on the above embodiment. In step S4, establishing a power generation fault diagnosis model includes the following steps:
步骤S41：状态表示，根据发电系统的特点和诊断的故障类别，设计发电系统的状态表示；Step S41: State representation. Design the state representation of the power generation system according to the characteristics of the power generation system and the diagnosed fault category;
步骤S42：定义反馈奖励函数，定义反馈奖励函数是R（s，a，s’），表示在状态s下采取动作a后转移到状态s’时的反馈奖励函数；Step S42: Define the feedback reward function. The feedback reward function is defined as R(s, a, s’), which represents the feedback reward function when taking action a in state s and then transitioning to state s’;
步骤S43：构建强化学习环境，将发电系统的状态表示和最具代表性的数据特征作为环境状态，将发电故障诊断的决策过程作为智能体的行为与环境状态进行交互并根据反馈奖励函数进行学习；Step S43: Construct a reinforcement learning environment, use the state representation of the power generation system and the most representative data features as the environmental state, use the decision-making process of power generation fault diagnosis as the behavior of the agent to interact with the environmental state, and learn based on the feedback reward function ;
步骤S44：构建发电故障诊断模型，使用深度神经网络作为发电故障诊断模型的函数近似器，学习最佳的决策策略；Step S44: Construct a power generation fault diagnosis model, use a deep neural network as a function approximator of the power generation fault diagnosis model, and learn the best decision-making strategy;
步骤S45：发电故障诊断模型训练，使用寻找最优策略算法，将发电故障诊断的决策过程作为智能体的行为通过与环境状态的交互进行发电故障诊断模型的训练和优化，包括以下步骤：Step S45: Training of the power generation fault diagnosis model. Using the optimal strategy algorithm, the decision-making process of power generation fault diagnosis is regarded as the behavior of the agent through interaction with the environmental state to train and optimize the power generation fault diagnosis model, including the following steps:
步骤S451：初始化Z值函数，将每一对状态——行动对的Z值初始化为0；Step S451: Initialize the Z value function, and initialize the Z value of each pair of state-action pairs to 0;
步骤S452：环境交互，根据最佳的决策策略选择行动，与环境状态进行交互，观察反馈奖励函数和下一状态；Step S452: Environment interaction, select actions according to the best decision-making strategy, interact with the environment state, and observe the feedback reward function and next state;
步骤S453：更新Z值函数，所用公式如下：Step S453: Update the Z value function, the formula used is as follows:
Z（s，a）=Z（s，a）+θ*（R+γ*max（Z（s’，a’））-Z（s，a））；Z(s,a)=Z(s,a)+θ*(R+γ*max(Z(s’,a’))-Z(s,a));
式中，Z（s，a）是当前状态——行动对的Z值，θ是学习率，R是当前反馈奖励函数，γ是折扣因子，s’是下一状态，a’是下一状态的最优行动；In the formula, Z (s, a) is the Z value of the current state-action pair, θ is the learning rate, R is the current feedback reward function, γ is the discount factor, s' is the next state, and a' is the next state. optimal action;
步骤S454：重复步骤S452和步骤S453，逐渐降低学习率θ，直到Z值函数收敛，求出一个最优Z值函数，得到使预期回报最大化的最优策略，并根据最优策略进行决策和行动选择；Step S454: Repeat steps S452 and S453, gradually reduce the learning rate θ until the Z-value function converges, find an optimal Z-value function, obtain the optimal strategy that maximizes expected returns, and make decisions based on the optimal strategy. action choice;
步骤S46：实时故障诊断，发电故障诊断模型经过训练优化后部署到实时环境中，根据实时采集到的数据进行故障预测和诊断，提供故障预警和维修建议。Step S46: Real-time fault diagnosis. The power generation fault diagnosis model is deployed in the real-time environment after training and optimization. It performs fault prediction and diagnosis based on the data collected in real time, and provides fault warning and maintenance suggestions.
在上述操作中，本方案采用寻找最优策略算法，将发电故障诊断的决策过程作为智能体的行为与环境状态进行交互并根据反馈奖励函数进行学习，根据实时观测到的反馈奖励函数优化诊断过程，在实时性要求高的场景下进行快速反应，减少发电系统维修成本，解决了目前发电故障诊断模型是基于静态模型进行建模和预测故障的发生，无法实时感知和适应变化的工作环境，导致诊断结果滞后，发电系统维修成本增加的技术问题。In the above operations, this solution uses the search for optimal strategy algorithm, taking the decision-making process of power generation fault diagnosis as the behavior of the agent interacting with the environmental state and learning based on the feedback reward function, and optimizing the diagnosis process based on the feedback reward function observed in real time. , react quickly in scenarios with high real-time requirements, reduce power generation system maintenance costs, and solve the problem that the current power generation fault diagnosis model is based on a static model to model and predict the occurrence of faults, and cannot perceive and adapt to the changing working environment in real time, resulting in Technical problems such as delayed diagnostic results and increased maintenance costs for the power generation system.
实施例六，参阅图1和图5，该实施例基于上述实施例，在步骤S5中，所述发电故障诊断结果解释，包括以下步骤：Embodiment 6. Refer to Figures 1 and 5. This embodiment is based on the above embodiment. In step S5, the explanation of the power generation fault diagnosis result includes the following steps:
步骤S51：特征重要性分析，使用决策树结构来分析发电故障诊断模型中每个特征样本的重要性，并计算每个特征样本的基尼系数来评估特征的重要性，所用公式如下：Step S51: Feature importance analysis, use the decision tree structure to analyze the importance of each feature sample in the power generation fault diagnosis model, and calculate the Gini coefficient of each feature sample to evaluate the importance of the feature. The formula used is as follows:
； ;
式中，Gini（p）是特征样本p的基尼系数，pd是第d类特征样本在特征样本p上的占比；In the formula, Gini(p) is the Gini coefficient of feature sample p, pd is the proportion of feature samples of type d in feature sample p;
步骤S52：解释故障诊断结果，构建局部解释性模型，使用线性回归模型获取诊断结果的特征权重，并解释单个样本点的诊断结果，包括以下步骤：Step S52: Interpret the fault diagnosis results, build a local explanatory model, use the linear regression model to obtain the feature weight of the diagnosis results, and interpret the diagnosis results of a single sample point, including the following steps:
步骤S521：通过特征重要性分析从所有特征样本中选择一部分对故障诊断有重要影响的重要特征；Step S521: Select some important features that have an important impact on fault diagnosis from all feature samples through feature importance analysis;
步骤S522：使用发电故障诊断模型对单个样本点进行故障诊断，得到样本点的故障诊断结果；Step S522: Use the power generation fault diagnosis model to perform fault diagnosis on a single sample point, and obtain the fault diagnosis result of the sample point;
步骤S523：样本点附近采样，从单个样本点附近进行采样，得到一组临近样本点；Step S523: Sampling near the sample point, sampling from the vicinity of a single sample point, and obtaining a group of adjacent sample points;
步骤S524：构建局部解释性模型，使用线性回归模型，将选定的单个样本点的特征和临近样本点的特征数据作为输入，将样本点的故障诊断结果作为输出，进行局部解释性模型训练；Step S524: Construct a local explanatory model, use a linear regression model, use the characteristics of the selected single sample point and the characteristic data of adjacent sample points as input, use the fault diagnosis results of the sample points as output, and perform local explanatory model training;
步骤S525：通过线性回归模型的系数获取样本点的故障诊断结果的特征权重，并解释该样本点的故障诊断结果；Step S525: Obtain the characteristic weight of the fault diagnosis result of the sample point through the coefficient of the linear regression model, and interpret the fault diagnosis result of the sample point;
步骤S53：解释决策过程，使用决策规则提取算法从复杂的模型中提取易于理解的决策规则，解释发电故障诊断模型中的决策过程，包括以下步骤：Step S53: Explain the decision-making process, use the decision rule extraction algorithm to extract easy-to-understand decision rules from the complex model, and explain the decision-making process in the power generation fault diagnosis model, including the following steps:
步骤S531：特征选择，输入要解释的发电故障诊断模型，使用基尼系数从发电故障诊断模型中选择与样本点的故障诊断结果有关的特征Q；Step S531: Feature selection, input the power generation fault diagnosis model to be explained, and use the Gini coefficient to select the feature Q related to the fault diagnosis result of the sample point from the power generation fault diagnosis model;
步骤S532：决策规则生成，利用关联规则挖掘方法挖掘出故障诊断结果和特征Q之间的关联，生成决策规则；Step S532: Generate decision rules, use the association rule mining method to mine the association between the fault diagnosis results and the feature Q, and generate decision rules;
步骤S533：决策规则评估，根据样本点的故障诊断结果与训练模型的目标标签的一致性对决策规则的准确度进行评估，根据决策规则的长度和可读性对决策规则的解释度进行评估；Step S533: Decision rule evaluation: evaluate the accuracy of the decision rule based on the consistency of the fault diagnosis results of the sample points and the target label of the training model, and evaluate the interpretability of the decision rule based on the length and readability of the decision rule;
步骤S534：决策规则筛选，应用启发式策略进行决策规则筛选，利用特征Q的局部信息指导搜索方向，根据搜索方向寻找重要特征，根据重要特征对决策规则进行筛选，得到筛选后的决策规则；Step S534: Decision rule screening, apply heuristic strategy to decision rule screening, use the local information of feature Q to guide the search direction, find important features according to the search direction, screen the decision rules according to the important features, and obtain the filtered decision rules;
步骤S535：解释决策规则，将筛选后的决策规则用可视化技术对发电故障诊断模型结果进行展开，将决策规则转化为易于理解的自然语言和图形化展示；Step S535: Explain the decision rules, use visualization technology to expand the power generation fault diagnosis model results after screening, and convert the decision rules into easy-to-understand natural language and graphical display;
步骤S54：交叉验证与评估优化，使用交叉验证来验证决策规则的泛化能力，对决策规则的解释性能进行定性分析，包括以下步骤：Step S54: Cross-validation and evaluation optimization, use cross-validation to verify the generalization ability of the decision rule, and conduct a qualitative analysis of the interpretation performance of the decision rule, including the following steps:
步骤S541：W折交叉验证，将测试数据集分为W个大小相同的子集，称为折，对于每个折，使用自己作为验证折，剩下的W-1个折作为训练折，对于每一次交叉验证的训练集，利用决策规则提取算法从训练折中提取决策规则，将生成的决策规则应用于对应的验证折，评估决策规则的泛化能力；Step S541: W-fold cross-validation, divide the test data set into W subsets of the same size, called folds. For each fold, use itself as the validation fold, and the remaining W-1 folds as training folds. For For each cross-validation training set, the decision rule extraction algorithm is used to extract decision rules from the training fold, and the generated decision rules are applied to the corresponding validation folds to evaluate the generalization ability of the decision rules;
步骤S542：对于W折交叉验证的结果，计算准确率、召回率和F1值，对决策规则的整体性能进行评估；Step S542: For the results of W-fold cross-validation, calculate the accuracy rate, recall rate and F1 value, and evaluate the overall performance of the decision rule;
步骤S544：分析优化，对决策规则的解释性能进行定性分析，观察决策规则提取算法是否对决策规则提供清晰的解释以及决策规则是否符合领域专家的知识，收集反馈和意见并及时调整改进。Step S544: Analyze and optimize, perform a qualitative analysis on the interpretation performance of the decision rule, observe whether the decision rule extraction algorithm provides a clear explanation of the decision rule and whether the decision rule conforms to the knowledge of domain experts, collect feedback and opinions, and make timely adjustments and improvements.
在上述操作中，本方案采用规则提取算法提取易于理解的决策规则，根据特征重要性分析和决策树结构解释模型中的决策过程，并用局部解释性模型解释单个样本点的诊断结果，利用可视化手段展开发电故障诊断模型的诊断结果，帮助工程师和操作人员理解发电故障诊断模型的诊断结果，提高工作效率，解决了黑盒子算法缺乏可解释性，无法解释内部运行方式，工程师和操作人员难以发现算法漏洞和改进点，无法进行深入分析和优化并做出决策的技术问题。In the above operations, this program uses a rule extraction algorithm to extract easy-to-understand decision rules, explains the decision-making process in the model based on feature importance analysis and decision tree structure, and uses a local explanatory model to explain the diagnostic results of a single sample point, using visualization means Expand the diagnosis results of the power generation fault diagnosis model to help engineers and operators understand the diagnosis results of the power generation fault diagnosis model, improve work efficiency, and solve the problem that black box algorithms lack interpretability and cannot explain the internal operation mode, making it difficult for engineers and operators to discover algorithms. Bugs and improvement points, technical issues that cannot be analyzed and optimized in depth and decisions made.
需要说明的是，在本文中，诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that these entities or operations are mutually exclusive. any such actual relationship or sequence exists between them. Furthermore, the terms "comprises," "comprises," or any other variations thereof are intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus that includes a list of elements includes not only those elements, but also those not expressly listed other elements, or elements inherent to the process, method, article or equipment.
尽管已经示出和描述了本发明的实施例，对于本领域的普通技术人员而言，可以理解在不脱离本发明的原理和精神的情况下可以对这些实施例进行多种变化、修改、替换和变型，本发明的范围由所附权利要求及其等同物限定。Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art will understand that various changes, modifications, and substitutions can be made to these embodiments without departing from the principles and spirit of the invention. and modifications, the scope of the invention is defined by the appended claims and their equivalents.
以上对本发明及其实施方式进行了描述，这种描述没有限制性，附图中所示的也只是本发明的实施方式之一，实际的结构并不局限于此。总而言之如果本领域的普通技术人员受其启示，在不脱离本发明创造宗旨的情况下，不经创造性的设计出与该技术方案相似的结构方式及实施例，均应属于本发明的保护范围。The present invention and its embodiments have been described above. This description is not limiting. What is shown in the drawings is only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person of ordinary skill in the art is inspired by the invention and without departing from the spirit of the invention, can devise structural methods and embodiments similar to the technical solution without inventiveness, they shall all fall within the protection scope of the invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311083408.5ACN116821774B (en) | 2023-08-28 | 2023-08-28 | Power generation fault diagnosis method based on artificial intelligence |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311083408.5ACN116821774B (en) | 2023-08-28 | 2023-08-28 | Power generation fault diagnosis method based on artificial intelligence |
| Publication Number | Publication Date |
|---|---|
| CN116821774A CN116821774A (en) | 2023-09-29 |
| CN116821774Btrue CN116821774B (en) | 2023-11-28 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311083408.5AExpired - Fee RelatedCN116821774B (en) | 2023-08-28 | 2023-08-28 | Power generation fault diagnosis method based on artificial intelligence |
| Country | Link |
|---|---|
| CN (1) | CN116821774B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117273392B (en)* | 2023-11-16 | 2024-02-09 | 四川省致链数字科技有限公司 | Furniture production decision method and device, electronic equipment and storage medium |
| CN117390536B (en)* | 2023-12-11 | 2024-04-02 | 深圳市宝腾互联科技有限公司 | Operation and maintenance management method and system based on artificial intelligence |
| CN118152172B (en)* | 2024-05-13 | 2024-08-23 | 南京新联电子股份有限公司 | Method and system for self-checking and self-recovering energy controller software faults |
| CN118964336B (en)* | 2024-10-17 | 2024-12-27 | 上海爱可生信息技术股份有限公司 | Database fault diagnosis decision method, system and readable storage medium |
| CN119338330B (en)* | 2024-12-18 | 2025-03-18 | 安徽中科维德数字科技有限公司 | PVB production process energy efficiency optimization system based on Apriori algorithm and Bian Yun cooperation |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015195205A1 (en)* | 2014-05-16 | 2015-12-23 | Transvoyant Llc | Computer-implemented systems and methods of analyzing data in an ad-hoc network for predictive decision-making |
| CN109739849A (en)* | 2019-01-02 | 2019-05-10 | 山东省科学院情报研究所 | A kind of network sensitive information of data-driven excavates and early warning platform |
| CN110262465A (en)* | 2019-07-11 | 2019-09-20 | 电子科技大学 | A kind of winged control method for diagnosing faults based on error code classification |
| AU2020101535A4 (en)* | 2020-07-28 | 2020-09-03 | Guangdong Ocean University | Fault Diagnosis Method for Ship Power System Based on Integrated Learning |
| WO2020247949A1 (en)* | 2019-06-07 | 2020-12-10 | The Regents Of The University Of California | General form of the tree alternating optimization (tao) for learning decision trees |
| CN114386312A (en)* | 2021-12-08 | 2022-04-22 | 华能沁北发电有限责任公司 | Equipment fault diagnosis method |
| CN116543215A (en)* | 2023-05-10 | 2023-08-04 | 北京工业大学 | A Brain Network Classification Method Based on Deep Hash Mutual Learning |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8001070B2 (en)* | 2007-12-17 | 2011-08-16 | Honeywell International Inc. | Object oriented rule-based system and method |
| TWI746038B (en)* | 2020-07-02 | 2021-11-11 | 阿證科技股份有限公司 | Neural network-like artificial intelligence decision-making core system |
| AU2022396273A1 (en)* | 2021-11-23 | 2024-05-30 | Strong Force TX Portfolio 2018, LLC | Transaction platforms where systems include sets of other systems |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015195205A1 (en)* | 2014-05-16 | 2015-12-23 | Transvoyant Llc | Computer-implemented systems and methods of analyzing data in an ad-hoc network for predictive decision-making |
| CN109739849A (en)* | 2019-01-02 | 2019-05-10 | 山东省科学院情报研究所 | A kind of network sensitive information of data-driven excavates and early warning platform |
| WO2020247949A1 (en)* | 2019-06-07 | 2020-12-10 | The Regents Of The University Of California | General form of the tree alternating optimization (tao) for learning decision trees |
| CN110262465A (en)* | 2019-07-11 | 2019-09-20 | 电子科技大学 | A kind of winged control method for diagnosing faults based on error code classification |
| AU2020101535A4 (en)* | 2020-07-28 | 2020-09-03 | Guangdong Ocean University | Fault Diagnosis Method for Ship Power System Based on Integrated Learning |
| CN114386312A (en)* | 2021-12-08 | 2022-04-22 | 华能沁北发电有限责任公司 | Equipment fault diagnosis method |
| CN116543215A (en)* | 2023-05-10 | 2023-08-04 | 北京工业大学 | A Brain Network Classification Method Based on Deep Hash Mutual Learning |
| Title |
|---|
| A distributed semantic similar search for high-dimensional resources in low-dimensional content addressable network;Qingyuan Hu;《 2013 IEEE 24th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC)》;全文* |
| 基于Fisher准则函数的相关向量机模拟电路故障诊断;杨颖涛;王跃钢;邓卫强;;计算机应用研究(第06期);全文* |
| 基于神经网络和社区发现的高维数据推荐系统;唐新宇;张新政;刘保利;;计算机应用与软件(第07期);全文* |
| Publication number | Publication date |
|---|---|
| CN116821774A (en) | 2023-09-29 |
| Publication | Publication Date | Title |
|---|---|---|
| CN116821774B (en) | Power generation fault diagnosis method based on artificial intelligence | |
| CN116665130B (en) | Multivariate time series anomaly detection method for dam safety monitoring based on spatio-temporal graph | |
| CN118133203A (en) | A method for fault diagnosis of electric energy metering detection information | |
| CN108932480A (en) | The study of distributing optical fiber sensing signal characteristic and classification method based on 1D-CNN | |
| CN107358347A (en) | Equipment cluster health state evaluation method based on industrial big data | |
| CN118157132B (en) | Data mining method and device for voltage monitoring system based on neural network | |
| CN115017791A (en) | Tunnel surrounding rock level identification method and device | |
| CN117668774A (en) | Metering equipment performance monitoring system | |
| Ye et al. | A deep learning-based method for automatic abnormal data detection: Case study for bridge structural health monitoring | |
| CN119416131B (en) | Data anomaly tracing and positioning method, system, equipment and storage medium | |
| CN110555478A (en) | Fan multi-fault diagnosis method based on depth measurement network of difficult sample mining | |
| CN118503730B (en) | Power big data processing method based on data clustering algorithm | |
| CN116614177A (en) | Optical fiber state multidimensional parameter monitoring system | |
| CN104569666A (en) | Power transformer fault prediction method based on electricity-graph model | |
| CN118820862B (en) | Knowledge graph-based fault diagnosis method for indicator diagram | |
| CN115048316A (en) | Semi-supervised software code defect detection method and device | |
| CN120196879B (en) | Abnormal point detection method for time-sequence marine ecological environment early warning monitoring data | |
| CN120146319B (en) | Power line health state assessment and prediction method and system based on big data | |
| CN118940002A (en) | Wind turbine fault diagnosis and repair method and related device based on knowledge graph | |
| CN119916140B (en) | A method for quickly locating distribution network faults based on artificial intelligence | |
| CN119396993A (en) | A method and system for power grid operation and maintenance deployment based on knowledge graph | |
| CN115438897A (en) | Industrial process product quality prediction method based on BLSTM neural network | |
| CN119293933A (en) | Intelligent pavement maintenance scheme design method and device based on dynamic knowledge graph | |
| CN119046693A (en) | Migration method of slope displacement prediction model | |
| CN116861330A (en) | Equipment health state assessment method based on rapid graph transformation network |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right | Effective date of registration:20231107 Address after:272069 No. 6, Hehua Road, taibaihu new area, Jining City, Shandong Province Applicant after:SHANDONG POLYTECHNIC College Address before:272069 Inside Shandong Vocational College of Technology, South Shou, Hehua Road, Beihu New District, Jining City, Shandong Province Applicant before:SHANDONG LIGONG HAOMING NEW ENERGY CO.,LTD. | |
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee | Granted publication date:20231128 |