Monitoring method for power equipment operation condition based on data warehouse

in the process of monitoring the operation state of power equipment, it is easy to be affected by the external environment, resulting in a large amount of data of the operation state of power equipment and the monitoring effect. A monitoring method of power equipment operation condition based on data warehouse is proposed. By analyzing the constraints of power equipment operation condition monitoring, a data warehouse is established; The transition probability and stability probability are determined by Markov chain model, and the characteristic values of operation state level variables are obtained by calculating the comprehensive correlation value and sensitivity; The entropy weight coefficient method calculates the weight of each performance index of the monitoring object, constructs the operation state monitoring model of power equipment, and realizes the monitoring of the operation state of power equipment. The experimental results show that the improved method can effectively improve the monitoring efficiency and shorten the monitoring time.


Introduction
With the rapid development of smart grid, higher requirements are put forward for the operation state of each equipment in the grid.Power equipment needs to establish a panoramic perception network that fully covers multiple links of the power system, get rid of the dilemma of incomplete data acquisition [1] in the past and incomplete data analysis and application, and make overall analysis and comprehensive evaluation based on massive data and correlation analysis [2] .Power equipment is the most basic component of power system, and its operation is directly related to the safety and reliability of power system.The operation, maintenance and replacement of power equipment also play an important role in the economic and efficient operation of power system [3] .In the past, the method of collecting the operation status information of power equipment was mainly through periodic manual inspection combined with regular preventive experiments [4] .Observing and recording the appearance of equipment, comparing test phenomena, instruments, etc. these methods belong to detection and test, which can not realize the real-time acquisition and display of the operation state of power equipment, nor meet the development needs of smart grid.Through the efforts of many researchers for many years, the operation condition monitoring method of power equipment has been widely used.However, for its operation condition monitoring data, the problems of large amount of data and lack of information lead to the poor use effect of condition monitoring [5] .In this regard, in the context of big data, how to use these data to provide strong technical support and new development ideas for power grid safety production, improve energy efficiency and provide high-quality services is an urgent problem to be solved.

Constraints
During the operation of power equipment, we should not only meet our own constraints, but also consider the balance constraints when interacting with the large power grid, mainly including the output constraints of power equipment, that is, to ensure that the power equipment operates at the upper and lower limits of output power [6] ; The node power flow constraint of power equipment ensures the power of any node of power equipment all the time, and the most important thing is to meet the power first win.
(1) Node power flow constraint In a single node, the active power injected by each power equipment shall be equal to the sum of the active power of the load at this point and the active power of this node and other nodes; Similarly, the reactive power at the node is also equal.On this basis, it is extended to all nodes.
(2) Power balance constraint In each power equipment operation device, the power shall be conserved, that is, the power balance constraint shall be met: Where, i gen P , is the power of each power equipment in each period, loss P is the loss power of power equipment, and ) (t P lxty is the load power. (3) Power equipment output constraints The power equipment operates within the rated range to ensure the efficient operation of power equipment.The constraints are as follows: For power equipment, it is necessary to meet the load point state constraints in each time period and ensure that the power equipment operates within the required range.After the nodes in the power equipment are connected to the power supply [7] , excessive access capacity may cause local operation damage.The power equipment at such nodes shall meet the upper and lower limit constraints of amplitude: Interaction capacity constraint of power equipment: due to the limited carrying capacity of power equipment, it causes too much pressure on power equipment when meeting the load requirements, so the following conditions need to be met:

Data warehouse construction
The mainstream data warehouse is mainly divided into power equipment information, dimensional data warehouse and multiple independent data warehouse.This paper mainly adopts independent data warehouse.Independent data warehouse is an analytical data storage, which is not designed in normal environment.It only focuses on the subject area.One or more power equipment can meet a database called data mart [8] .Data marts may adopt dimension design, er relationship model or other forms of design.Use analytical tools to directly query the power equipment, and then feed back the results to the end user.The main reason for adopting independent data mart is that it does not need cross functional analysis and can be put into production soon.

Characteristic values of operating state level variables
Markov chain has the characteristics of no aftereffect, which can better describe the deterioration degree of power equipment operation state.Using Markov process as a tool, it is convenient to calculate a variety of reliability indexes of power equipment, including steady-state probability and state frequency [9] .Considering more parameters, Markov process can be applied to the description of deterioration process with multiple states and considering the duration of each state, which can more accurately express the state and transformation of power equipment.Its transition probability ij  can be by using the following formula: Where, N represents the number of states contained in the model, and ij  represents the average number of transitions between states.In order to solve the stationary probability, it is necessary to calculate the transition proportion i  entering each state and ensure that the sum of the transition proportions of all transitions in the model is 1.The solution of stationary probability not only considers the transition probability of each state [10] , but also takes into account the transition duration of each state.Therefore, the calculation formula of transition stationary probability is: Where, i T represents the average duration of the ith state, which is equal to the reciprocal of the sum of all probabilities transferred from the state.According to the description of "distance" in extension theory, the distance from a point x to interval ] , [ 0 b a x  can be expressed as: Then the definition of correlation function Where:   Where: of each level corresponding to the power equipment; ij w is the weight coefficient of each monitoring index.
Suppose that there are the n single models, the first mock exam results of n models are the best matrix, and ij  is used to show the accurate number of i models relative to the j model.The weight of the K model can be obtained as follows: Where: During monitoring, the correlation degree between each index of the power equipment to be monitored and different state levels can be reflected by the magnitude of the goodness value [11] .The calculation formula of the goodness of the power equipment 0 P to j N to be monitored is: Where: represents the comprehensive priority value of the power equipment index to be monitored for each state level, and each priority represents the degree that 0 P belongs to monitoring level j N .
After determining the goodness value, when: It shows that the solved condition monitoring model is closest to 0 N , so the condition monitoring level of the power equipment is 0 N .
By calculating the optimization value, it is considered that the current operation level of power equipment is level j N .
The sensitivity of the index can be analyzed through the eigenvalue of the grade variable, and the calculation formula is: Then the characteristic value of the level variable at this time is: In conclusion, the consistency between the monitored matter element R and each state level can be judged by the change of the eigenvalue of the level variable.

Construction of power equipment operation condition monitoring model
The input of the power equipment operation condition monitoring model is the observation vector ob X at a certain observation time of the power equipment.The output and input correlation factors of the model can be represented by a predicted value e X of the observation vector value.The input observation vector of the model can form a weight vector with m dimensions: In other words, the monitoring value output by this model is the set of m relevant factor observation vectors in the process matrix D. the residual between the input observation vector and the output monitoring vector of the power equipment monitoring model is: Select W to minimize the sum of squares of residuals.The sum of squares of residuals is: Assuming that i  is the subjective weight of the i index and ij r is the importance of this index relative to j index, the subjective weight of index i can be calculated by the following formula.

  
The entropy weight coefficient method is used to calculate the objective weight of each performance index of power equipment.When calculating the weight of each performance index of the monitoring object, the entropy weight coefficient method mainly uses the observed values of different performance indexes of the evaluation object.Under N performance evaluation indexes, m power equipment data to be evaluated form matrix . Then the objective weight i v of the i index of the monitoring power equipment can be obtained: Make full use of the time sequence characteristics of the data, and the monitoring is the value of a fixed number of time points, which is more suitable for short-term monitoring.The formula is as follows: is the moving average model.The most commonly used method is the autoregressive moving average model.According to the AIC minimum information criterion, focus on determining the P and Q values of subordinate time series, and then confirm the monitoring time points of fixed length.Limited by the monitoring principle of time series, generally, the monitoring points should not be too many.The test data set is used to verify the model and output the monitoring value.In order to represent the cumulative change of operation probability and the influence of internal and external factors of power equipment on its own operation state, the monitoring model is established as follows: Where, t is the actual operation time of power equipment, k is the lag order,     is the operation state probability of previous continuous k-time,   t X is the d-dimensional column vector, which represents various internal and external influencing factors at time t, including equivalent operation time, equipment state, gold factor, etc., k   ,..., 1 and d-dimensional row vectors,  is the coefficient matrix to be estimated, and C is the constant term.

Analysis of experimental results
This experiment is based on Hadoop platform.Users can easily and quickly carry out data mining and Analysis on it.This paper uses a small S822lc server with 16 core power8 processor, main frequency of 3.3GHz, 128gddr4 memory, NVIDIA Tesla K80 24GB GPU Accelerator card and 10TB disk array, which provides a guarantee for the establishment of data warehouse.Under the condition of the same amount of data, the experimental comparative analysis is carried out with the method in this paper, the method in literature [5] and the method in literature [4] as the comparison, and the monitoring time and accuracy as the index.
(1) Monitoring time / (min): Give priority to the study of the relationship between the amount of data and the operation condition monitoring time of power equipment, and select two traditional methods as the comparison method.The specific experimental results are shown in Figure 1:  1, with the continuous increase of the amount of power equipment operation state data, the operation state monitoring time of power equipment also continues to increase.Among them, the upward trend of monitoring completion time of methods in literature [5] and literature [4] is more obvious, mainly because the two methods fail to use data warehouse technology, resulting in inaccurate data collection of power equipment operation status and increased monitoring time.The monitoring time of this method is significantly lower than that of the other two methods, which comprehensively verifies the advantages of the proposed method.
(2) Monitoring accuracy / (%): The monitoring accuracy is selected as the main basis to measure whether the monitoring results of power equipment operation status are accurate.The specific results are shown in Table 1: Table 1.comparison results of monitoring accuracy of different methods

Number of experiments / (Times)
Monitoring accuracy / (%) Paper method Literature [5] methods Literature [4] 1 that when the method of literature [5] is adopted, the monitoring accuracy is about 87.1%, and the accuracy decreases with the increase of the number of experiments; When using the method of literature [4] , the detection accuracy is about 89.3%, and the accuracy decreases with the increase of the number of experiments, but compared with the method of literature [5] , the accuracy is improved by 2.2%; When using this method, the monitoring accuracy is 96%, which decreases with the occurrence of accuracy.However, compared with the methods in literature [5] and literature [4] , the monitoring accuracy of this method is improved by 12.9% and 6.7% respectively, indicating that this method can obtain more accurate monitoring results.

Conclusion
(1)In the process of monitoring the operation state of power equipment, it is easy to be affected by the external environment, resulting in a large amount of data of the operation state of power equipment and the monitoring effect.A monitoring method of power equipment operation condition based on data warehouse is proposed.The experimental results show that: (2)When using this method, the detection time is shortened by 120 and 61S respectively compared with the methods in literature [5] and literature [4] ; (3)When using this method, the monitoring accuracy is improved by 12.9% and 6.7% respectively compared with the methods in literature [5] and literature [4] , indicating that this method can obtain more accurate monitoring results.
comprehensive weight value ij w of the monitoring index, the comprehensive correlation value   x i N k of the corresponding grade J of the monitoring power equipment n is:

d
standardized treatment.Among them, the positive performance index of power equipment adopts as the jth observation value of i index of power equipment.

Figure 1 .
Figure 1.Comparison results of monitoring timeAccording to the experimental data in Figure1, with the continuous increase of the amount of power equipment operation state data, the operation state monitoring time of power equipment also continues to increase.Among them, the upward trend of monitoring completion time of methods in literature[5] and literature[4] is more obvious, mainly because the two methods fail to use data warehouse technology, resulting in inaccurate data collection of power equipment operation status and increased monitoring time.The monitoring time of this method is significantly lower than that of the other two methods, which comprehensively verifies the advantages of the proposed method.(2)Monitoring accuracy / (%): The monitoring accuracy is selected as the main basis to measure whether the monitoring results of power equipment operation status are accurate.The specific results are shown in Table1:Table1.comparison results of monitoring accuracy of different methods