技术领域technical field
本发明涉及电力安全管控技术领域，更具体地，涉及基于神经网络的电力人身事故、误操作事故与违章操作事故的预测方法。The invention relates to the technical field of electric power safety management and control, and more specifically, to a neural network-based prediction method for electric power personal accidents, misoperation accidents, and illegal operation accidents.
背景技术Background technique
大数据的发展可谓如日中天，也成为当今信息行业的热门话题，所谓大数据即数据量庞大到无法通过现阶段的主流软件工具进行统计，并在合理的时间内达到获取，管理，处理并整理成为帮助企业经营决策的更有价值更有效的信息。The development of big data can be said to be in full swing, and it has become a hot topic in today's information industry. The so-called big data means that the amount of data is too large to be counted by the current mainstream software tools, and it can be acquired, managed, processed and organized within a reasonable time. More valuable and effective information to help enterprises make business decisions.
从电力企业近年来安全生产的数据来看，人身伤亡事故数和死亡人数呈逐年上升的趋势，电力生产中人身事故、特大和重大设备事故等恶性事故也时有发生，在人因安全方面依然形势严峻。同时，违章问题也一直是电力企业生产安全管理的一个重难点。相关研究表明，违章行为可导致人身伤害的主观感觉的作用、违章行为会受到法规惩处的作用、违章行为与遵章行为满足生理心理需要作用的差值等；侥幸心理、从众心理、草率心理、省能心理、逆反心理、散漫心理等也是产生违章作业的原因。此外，电气误操作问题一直受到高度重视。运行值班人员误操作行为的原因包括人为因素、自然环境因素、设备因素、人因失误以及认知和技能因素、环境因素、生理/心理/性格因素等。Judging from the safety production data of electric power enterprises in recent years, the number of personal injury accidents and death toll is increasing year by year, and vicious accidents such as personal accidents, extraordinarily large and major equipment accidents also occur from time to time in electric power production, and there is still a situation in terms of human safety. severe. At the same time, the problem of violation of regulations has always been a major difficulty in the production safety management of electric power enterprises. Relevant studies have shown that the role of the subjective feeling that violations can lead to personal injury, the role that violations will be punished by laws and regulations, the difference between violations and compliance that meet physiological and psychological needs, etc.; Energy-saving psychology, rebellious psychology, sloppy psychology, etc. are also the reasons for illegal operations. In addition, the issue of electrical misuse has been given high priority. The reasons for the misoperation of personnel on duty include human factors, natural environment factors, equipment factors, human errors, cognitive and skill factors, environmental factors, physiological/psychological/character factors, etc.
然而目前电力人身事故、误操作与违章研究方法及其成果大多相近，大多以定性分析为主。例如：采用层次分析法，将各目标因素按照重要性的大小，形成关联层次，建立排序判断矩阵，并作为决策的依据，存在一致性和计算量大等问题。采用基于模糊函数的层次分析法，能够有效解决上述问题，但难以揭示电力人身和误操作事故事件的规律和特征。采用基于三角模糊函数的层次分析法，改进了以往单纯使用层次分析法的不足，在一定程度上有助于揭示电力人因事故事件的发生机理，但缺乏基础数据。However, most of the research methods and results of electric personal accidents, misuse and violations are similar, and most of them are qualitative analysis. For example: using the AHP to form a correlation hierarchy for each target factor according to its importance, establish a ranking judgment matrix, and use it as the basis for decision-making, there are problems such as consistency and large amount of calculation. Using AHP based on fuzzy functions can effectively solve the above problems, but it is difficult to reveal the laws and characteristics of power personal accidents and misoperation accidents. The AHP based on the triangular fuzzy function has improved the shortcomings of the simple use of the AHP in the past. To a certain extent, it is helpful to reveal the mechanism of electric human-caused accidents, but there is a lack of basic data.
发明内容Contents of the invention
本发明为解决以上现有技术的缺陷，提供了一种基于神经网络的电力人身事故、误操作事故与违章操作事故的预测方法。In order to solve the above defects of the prior art, the present invention provides a neural network-based prediction method for electrical personal accidents, misoperation accidents and operation violation accidents.
为实现以上发明目的，采用的技术方案是：For realizing above-mentioned purpose of the invention, the technical scheme that adopts is:
基于神经网络的电力人身事故、误操作事故与违章操作事故的预测方法，包括以下步骤：A neural network-based prediction method for electrical personal accidents, misuse accidents, and illegal operation accidents includes the following steps:
S1.采集各种电力人身事故、误操作事故、违章操作事故的历史数据；S1. Collect historical data of various electrical personal accidents, misuse accidents, and illegal operation accidents;
S2.搭建神经网络，利用各种电力人身事故、误操作事故、违章操作事故的历史数据对神经网络进行训练；S2. Build a neural network and train the neural network using historical data of various electrical personal accidents, misuse accidents, and illegal operation accidents;
S3.将测试数据输入至训练好的神经网络中，得到预测结果。S3. Input the test data into the trained neural network to obtain the prediction result.
与现有技术相比，本发明的有益效果是：Compared with prior art, the beneficial effect of the present invention is:
本发明利用历史数据，研究电力人身事故、误操作事故、违章操作事故的特征，采用神经网络探究事故发生规律，分析检测各种事故状态，研究电力人身事故、误操作事故、违章操作事故与人、机、料、法、环等特征要素关联性和灵敏性的大数据分析方法，揭示电力人身事故、误操作事故、违章操作事故的发生规律，明确电力人身事故、误操作事故、违章操作事故分析所需的关键信息，确定导致电力人身事故、误操作事故、违章操作事故的高风险因素，制定电力人身事故、误操作事故、违章操作事故的预控策略。The invention uses historical data to study the characteristics of electric power personal accidents, misuse accidents, and operation violation accidents, uses neural networks to explore the occurrence rules of accidents, analyzes and detects various accident states, and studies the relationship between personal accidents, misuse accidents, and operation violation accidents related to people and machines. A big data analysis method based on the correlation and sensitivity of characteristic elements such as material, law, environment, etc., reveals the occurrence rules of electric power personal accidents, misuse accidents, and illegal operation accidents, and clarifies the analysis of electric personal accidents, misuse accidents, and illegal operation accidents. Key information, determine the high-risk factors that lead to electric power personal accidents, misuse accidents, and illegal operation accidents, and formulate pre-control strategies for electric power personal accidents, misuse accidents, and illegal operation accidents.
附图说明Description of drawings
图1为方法的流程图示意图。Figure 1 is a schematic flow chart of the method.
图2为训练神经网络的示意图。Figure 2 is a schematic diagram of training a neural network.
具体实施方式Detailed ways
附图仅用于示例性说明，不能理解为对本专利的限制；The accompanying drawings are for illustrative purposes only and cannot be construed as limiting the patent;
以下结合附图和实施例对本发明做进一步的阐述。The present invention will be further elaborated below in conjunction with the accompanying drawings and embodiments.
实施例1Example 1
如图1所示，本发明提供的方法包括有以下步骤：As shown in Figure 1, the method provided by the present invention includes the following steps:
S1.采集各种电力人身事故、误操作事故、违章操作事故的历史数据；S1. Collect historical data of various electrical personal accidents, misuse accidents, and illegal operation accidents;
S2.搭建神经网络，利用各种电力人身事故、误操作事故、违章操作事故的历史数据对神经网络进行训练；S2. Build a neural network and train the neural network using historical data of various electrical personal accidents, misuse accidents, and illegal operation accidents;
S3.将测试数据输入至训练好的神经网络中，得到预测结果。S3. Input the test data into the trained neural network to obtain the prediction result.
在具体的实施过程中，如图2所示，所述对神经网络进行训练时，采集发生电力人身事故、误操作事故、违章操作事故时的电力生产工作、电力建设工作、条件特征、人员特征作为神经网络的输入值；而采集发生电力人身事故、误操作事故、违章操作事故时的事故等级、人身事故类型、电力人身事故等级、人身事故事件层级、责任类型作为神经网络的输出值，通过输入值、输出值对神经网络进行训练；In the specific implementation process, as shown in Figure 2, when the neural network is trained, the power production work, power construction work, condition characteristics, and personnel characteristics when electric power personal accidents, misuse accidents, and illegal operation accidents occur are collected as The input value of the neural network; and the accident grade, personal accident type, electric personal accident grade, personal accident event level, and responsibility type when electric personal accidents, misuse accidents, and illegal operation accidents occur are collected as the output value of the neural network. train the neural network;
电力生产工作可为以下中的任一种：输电、变电、配电、供电、用电、调度、运行、巡视、检修、维护、试验、抄表验收、用电检查、修理、技改、业扩；Power production work can be any of the following: power transmission, power transformation, power distribution, power supply, power consumption, scheduling, operation, inspection, overhaul, maintenance, test, meter reading acceptance, power inspection, repair, technical transformation, business expansion;
电力建设工作可为以下中的任一种：输变电配电工程施工、安装、调试和监理；Power construction work can be any of the following: construction, installation, commissioning and supervision of power transmission and distribution projects;
条件特征可为以下中的任一种：触电、倒杆、高空坠落、火灾、爆炸爆破；The condition feature can be any of the following: electric shock, inverted rod, falling from a high altitude, fire, explosion and blasting;
人员特征可为以下中的任一种：性别、年龄、学历、用工形式、工龄、工种、本工种工龄、教育培训、安全考试、资格证书、职业禁忌；Personnel characteristics can be any of the following: gender, age, education background, employment form, length of service, type of work, length of service of this type of work, education and training, safety examinations, qualification certificates, and occupational taboos;
事故等级可为以下中的任一种：特大、重大、较大、一般；The accident level can be any one of the following: extraordinarily large, major, relatively large, and general;
人身事故类型可为以下中的任一种：人身死亡、人身伤害；The type of personal accident may be any of the following: personal death, personal injury;
电力人身事故等级可为以下中的任一种：一级、二级、三级、四级、五级；The level of electrical personal accidents can be any of the following: level 1, level 2, level 3, level 4, level 5;
人身事故事件层级可为以下中的任一种：500kV、220kV、110kV、35kV、10kV、380V、220V；The personal accident level can be any of the following: 500kV, 220kV, 110kV, 35kV, 10kV, 380V, 220V;
责任类型可为以下中的任一种：领导责任、管理责任、执行责任或直接责任、间接责任。The type of responsibility can be any of the following: leadership responsibility, management responsibility, executive responsibility or direct responsibility, indirect responsibility.
其中对神经网络进行训练的具体过程如下：The specific process of training the neural network is as follows:
1)数据预处理1) Data preprocessing
将不同来源、格式、性质的数据进行有机整合，根据数据特点进行存储，必要时进行关联储存或分布式存储。进行数据筛选，去除数据中不完整的，不一致的数据，去除噪声数据，将可用的，完成的数据整合为数据集。将数据做变换，平滑，规范化，归一化等处理，保证其成为神经网络的输入数据格式。Organically integrate data from different sources, formats, and properties, store data according to data characteristics, and perform associated storage or distributed storage when necessary. Perform data screening, remove incomplete and inconsistent data in the data, remove noise data, and integrate available and completed data into a data set. Transform, smooth, normalize, and normalize the data to ensure that it becomes the input data format of the neural network.
2)搭建网络2) Build a network
神经网络结构搭建如下：电力人身事故、误操作事故、违章操作事故的数据量庞大，将准备好的电力生产工作，电力建设工作，条件特征，人员特征作为输入信号X，经过对这些信号的分析，可以预测电力事故的结果类型与等级，这些状态作为输出信号Y，本发明采用4层神经网络，隐藏层节点为7个和5个，即构建双隐藏层的电力事故BP神将网络，同时考虑到双曲正切函数在该问题上有更好的效果，激活函数采用双曲正切函数，函数为：The neural network structure is built as follows: the amount of data of electric power personal accidents, misuse accidents, and illegal operation accidents is huge. The prepared electric power production work, electric power construction work, condition characteristics, and personnel characteristics are used as input signals X. After analyzing these signals, It is possible to predict the type and level of the result of the power accident, and these states are used as the output signal Y. The present invention adopts a 4-layer neural network with 7 and 5 nodes in the hidden layer, that is, constructs a double-hidden layer power accident BP network, and considers The hyperbolic tangent function has a better effect on this problem. The activation function uses the hyperbolic tangent function. The function is:
f(x)＝tanh(x)f(x)=tanh(x)
3)初始化网络，初始化权重和偏置3) Initialize the network, initialize weights and biases
利用随机值对权重和偏置进行初始化Initialize weights and biases with random values
4)正向传播4) Forward propagation
(4.1)计算输入层到隐藏层(4.1) Calculate the input layer to the hidden layer
网络搭建完后需要进行训练，对于神经网络而言，训练即为参数的学习过程，又正向传播和反向传播构成，正向传播中，输入层到隐藏层的传播过程为公式(2)所示，将数据逐条提供给输入层，并计算隐藏层各节点值及隐藏节点的输出。其中yhk为隐藏层神经元的值，ωhi为权值，xi为输入值，bhk为偏置，f为激活函数。After the network is built, it needs to be trained. For the neural network, the training is the learning process of the parameters, and it is composed of forward propagation and back propagation. In the forward propagation, the propagation process from the input layer to the hidden layer is the formula (2) As shown, the data is provided to the input layer one by one, and the value of each node in the hidden layer and the output of the hidden node are calculated. Where yhk is the value of the neuron in the hidden layer, ωhi is the weight,xi is the input value, bhk is the bias, and f is the activation function.
(4.2)计算隐藏层到输出层(4.2) Calculate the hidden layer to the output layer
隐藏层到输出层的传播过程相同于此相同，公式如下，利用公式(3)计算出网络的输出值The propagation process from the hidden layer to the output layer is the same as this, the formula is as follows, use the formula (3) to calculate the output value of the network
5)计算误差5) Calculation error
输出层得到理论输出，将其与实际输出比较，计算误差并根据误差进行反向传播，均方差为计算误差的常用方式。利用公式(4)计算误差，其中yi为输出层神经元的输出值，Yi为实际值。The output layer obtains the theoretical output, compares it with the actual output, calculates the error and backpropagates according to the error, and the mean square error is a common way to calculate the error. The error is calculated by formula (4), where yi is the output value of the neuron in the output layer, and Yi is the actual value.
6)反向传播6) Backpropagation
根据误差进行反向传播，更新隐藏层的权值和偏置，权重更新的公式如下，其中第一项为常规算法的权值修正项，表示误差的梯度方向，η为学习率，第二项为动量项，α为动量因子Back propagation is performed according to the error, and the weight and bias of the hidden layer are updated. The formula for updating the weight is as follows, where the first item is the weight correction item of the conventional algorithm, Indicates the gradient direction of the error, η is the learning rate, the second item is the momentum item, and α is the momentum factor
7)学习率自学习7) Learning rate self-learning
按照公式(6)至(10)进行学习率自学习，调整网络的学习率η，其中gt为梯度，mt，vt分别对梯度的一阶矩估计和二阶矩估计，是对mt和vt的矫正，β1，β2为参数，可分别设为0.9和0.999，∈可设置为10-8，其可防止除数为0，m0，v0初始值为0Carry out learning rate self-learning according to formulas (6) to (10), adjust the learning rate η of the network, where gt is the gradient, mt and vt are the first-order moment estimation and second-order moment estimation of the gradient respectively, is the correction for mt and vt , β1 and β2 are parameters, which can be set to 0.9 and 0.999 respectively, ∈ can be set to 10-8 , which can prevent the divisor from being 0, and the initial value of m0 and v0 is 0
mt＝β1mt-1+(1-β1)gt (6)mt ＝β1 mt-1 +(1-β1 )gt (6)
8)动量因子自学习8) Momentum factor self-learning
按照公式(11)进行动量因子自学习，调整网络的动量因子α，其中λ为正常数，用于控制动量因子的大小，显然，α的值处于0和1之间，并随着误差关于权值向量的梯度范数的变化而变化。Momentum factor self-learning according to formula (11), adjust the momentum factor α of the network, where λ is a normal number, used to control the size of the momentum factor, obviously, the value of α is between 0 and 1, and with the error about the weight The gradient norm of the value vector changes.
9)重复迭代9) Repeat iterations
返回步骤(5)进行重复迭代，直至网络的误差小于预先设定的值或迭代次数大于给定值Return to step (5) for repeated iterations until the error of the network is less than the preset value or the number of iterations is greater than the given value
10)结束训练10) End training
或者在未达到预先设定次数时以收敛，则结束训练。Or it will converge when the preset number of times is not reached, and the training will end.
显然，本发明的上述实施例仅仅是为清楚地说明本发明所作的举例，而并非是对本发明的实施方式的限定。对于所属领域的普通技术人员来说，在上述说明的基础上还可以做出其它不同形式的变化或变动。这里无需也无法对所有的实施方式予以穷举。凡在本发明的精神和原则之内所作的任何修改、等同替换和改进等，均应包含在本发明权利要求的保护范围之内。Apparently, the above-mentioned embodiments of the present invention are only examples for clearly illustrating the present invention, rather than limiting the implementation of the present invention. For those of ordinary skill in the art, other changes or changes in different forms can be made on the basis of the above description. It is not necessary and impossible to exhaustively list all the implementation manners here. All modifications, equivalent replacements and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the claims of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810012435.6ACN108564193A (en) | 2018-01-05 | 2018-01-05 | Prediction method of electric power personal accidents, misuse accidents and illegal operation accidents based on neural network |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810012435.6ACN108564193A (en) | 2018-01-05 | 2018-01-05 | Prediction method of electric power personal accidents, misuse accidents and illegal operation accidents based on neural network |
| Publication Number | Publication Date |
|---|---|
| CN108564193Atrue CN108564193A (en) | 2018-09-21 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810012435.6APendingCN108564193A (en) | 2018-01-05 | 2018-01-05 | Prediction method of electric power personal accidents, misuse accidents and illegal operation accidents based on neural network |
| Country | Link |
|---|---|
| CN (1) | CN108564193A (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111723970A (en)* | 2020-05-06 | 2020-09-29 | 国网浙江省电力有限公司衢州供电公司 | A method for predicting hidden dangers of electrical work |
| CN111862557A (en)* | 2019-04-24 | 2020-10-30 | 广州煜煊信息科技有限公司 | A family accident handling method based on the Internet of things monitoring method |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014093670A1 (en)* | 2012-12-12 | 2014-06-19 | University Of North Dakota | Analyzing flight data using predictive models |
| CN104462727A (en)* | 2014-12-31 | 2015-03-25 | 中国科学院遥感与数字地球研究所 | Oil spilling simulation parameter optimization method based on dynamic remote sensing data driving |
| CN104992299A (en)* | 2015-07-23 | 2015-10-21 | 河南行知专利服务有限公司 | Power grid risk analysis and early warning method |
| KR101830295B1 (en)* | 2017-06-23 | 2018-03-29 | 주식회사 주빅스 | INTELLIGENT ACCIDENT PREVENTION TRAINING SELECTION DECISION SYSTEM OF TOXIC SUBSTANCE BUSINESS USING IoT SENSOR |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014093670A1 (en)* | 2012-12-12 | 2014-06-19 | University Of North Dakota | Analyzing flight data using predictive models |
| CN104462727A (en)* | 2014-12-31 | 2015-03-25 | 中国科学院遥感与数字地球研究所 | Oil spilling simulation parameter optimization method based on dynamic remote sensing data driving |
| CN104992299A (en)* | 2015-07-23 | 2015-10-21 | 河南行知专利服务有限公司 | Power grid risk analysis and early warning method |
| KR101830295B1 (en)* | 2017-06-23 | 2018-03-29 | 주식회사 주빅스 | INTELLIGENT ACCIDENT PREVENTION TRAINING SELECTION DECISION SYSTEM OF TOXIC SUBSTANCE BUSINESS USING IoT SENSOR |
| Title |
|---|
| 唐开元等: "《小波网络理论及其在经济建模中的应用》", 31 December 2008* |
| 徐荣贞等: "《金融生态视角下系统性风险研究》", 31 December 2017* |
| 湛从昌: "《液压可靠性最优化与智能故障诊断》", 31 December 2015* |
| 赵林明等: "《多层前向人工神经网络》", 31 December 1999* |
| 郭艳玲: "《线切割机床虚拟制造系统》", 31 December 2001* |
| 陈丽华等: "《人工神经网络及其在水质信息检测中的应用》", 31 December 2011* |
| 高晓旭: "《构建本质安全型煤矿理论与实践》", 31 December 2016* |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111862557A (en)* | 2019-04-24 | 2020-10-30 | 广州煜煊信息科技有限公司 | A family accident handling method based on the Internet of things monitoring method |
| CN111862557B (en)* | 2019-04-24 | 2022-08-16 | 广州煜煊信息科技有限公司 | Household accident handling method based on Internet of things monitoring mode |
| CN111723970A (en)* | 2020-05-06 | 2020-09-29 | 国网浙江省电力有限公司衢州供电公司 | A method for predicting hidden dangers of electrical work |
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Sound quality prediction of vehicle interior noise and mathematical modeling using a back propagation neural network (BPNN) based on particle swarm optimization (PSO) | |
| CN102496069B (en) | Cable multimode safe operation evaluation method based on fuzzy analytic hierarchy process (FAHP) | |
| Wu et al. | A safety investment optimization model for power grid enterprises based on System Dynamics and Bayesian network theory | |
| Wang et al. | Coal mine safety production forewarning based on improved BP neural network | |
| CN113327047B (en) | Power marketing service channel decision method and system based on fuzzy comprehensive model | |
| CN109447330B (en) | Power distribution network risk early warning method considering power grid elasticity and adaptability | |
| CN105023067A (en) | Analytic hierarchy process-fuzzy comprehensive evaluation-based chemical project environmental risk evaluation system | |
| CN111598352A (en) | Concrete beam type bridge comprehensive evaluation method based on Bayesian network | |
| Chen et al. | Knowledge graph improved dynamic risk analysis method for behavior-based safety management on a construction site | |
| CN106355540A (en) | Small- and medium-sized reservoir dam safety evaluating method based on GRA-BP (grey relational analysis and back propagation) neural network | |
| CN105095963A (en) | Method for accurately diagnosing and predicting fault of wind tunnel equipment | |
| Jing et al. | Application of fuzzy analytic hierarchy process in the risk assessment of dangerous small-sized reservoirs | |
| CN104700321A (en) | Analytical method of state running tendency of transmission and distribution equipment | |
| CN106096830A (en) | Relay protection method for evaluating state based on broad sense evidence theory and system | |
| CN103177288A (en) | Transformer fault diagnosis method based on genetic algorithm optimization neural network | |
| CN104951588A (en) | Aided design method for mine ventilation systems | |
| CN109740727A (en) | A kind of hydraulic turbine shaft state monitoring method neural network based and system | |
| CN107194574A (en) | A kind of grid security risk assessment method based on load loss | |
| CN105303331A (en) | Transformer repair risk decision-making method | |
| CN110059913A (en) | A kind of quantitative estimation method counted and the power failure of future-state is planned | |
| CN110889587A (en) | Risk assessment method for distribution network lines | |
| Chen et al. | A comprehensive multi‐hazard risk assessment model for an urban agglomeration with multiple factors | |
| Zhou et al. | Establishment and assessment of urban meteorological disaster emergency response capability based on modeling methods | |
| CN118627912A (en) | A method for assessing urban pipeline safety risks | |
| CN103198362A (en) | Method for coal mine safety evaluation |
| 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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication | Application publication date:20180921 |