Intelligent Perception of Distribution Network State Based on Multi-source Data Fusion

This article proposes an intelligent state perception method based on multi-source data fusion to address the issue of the difference between the state perception range and the actual range caused by the abnormal operation status of active distribution networks. In this paper, by combining Ant colony optimization algorithms and fusion rules, a multi-source data fusion model of the operation situation is constructed. The operation situation indicators, such as line loss rate and light load rate, are determined. Using multi-source data fusion information entropy to train samples, the range of abnormal operating states of the power grid is determined. By fusing multi-source voltage data and using the multi-source data fusion method to distinguish voltage data, four operating state perception results of voltage are obtained. Finally, the perception data is modified using Analytic Hierarchy Process to ensure that the state perception results meet the compatibility and consistency requirements with the actual results. According to the experimental results, the proposed method can accurately perceive the fluctuation of abnormal data within 40~50 ranges, which is consistent with the actual value fluctuation range, so as to obtain accurate perception results under different operating conditions.


Introduction
With the continuous growth of electricity demand in modern society, the reliability and stability of distribution networks are particularly important [1][2][3].However, due to the complex operating environment and various factors, the operating status of active distribution networks often experiences abnormal situations, resulting in certain differences between the actual state perception range and theoretical expectations [4][5][6].Therefore, in order to improve the accuracy and reliability of distribution network state perception, this paper proposes an intelligent perception method based on multi-source data fusion.
The goal of this study is to use multi-source data fusion technology to solve the problem of active distribution network operation state perception.This method combines Ant colony optimization algorithms and fusion rules [7][8] to build a multi-source data fusion model of operation situation by determining the order of data fusion to determine the line loss rate, light load rate, and other operation situation indicators.By using multi-source data fusion information entropy to train samples, the range of abnormal operating states of the power grid is determined.By fusing multi-source voltage data and utilizing multi-source data fusion methods to distinguish voltage data, four operational state perception results regarding voltage are obtained [9][10][11].Finally, the perceptual data is modified using the Analytic Hierarchy Process to meet the compatibility and consistency requirements between the perceptual results and the actual results [12][13].

The operation situation index of the ant colony algorithm and fusion rule is determined
In the process of multi-source data fusion, the same operating situation indicators in multiple data sources are combined.The ant colony algorithm and fusion rules are used to deal with multi-source data integration.By integrating different running state indices, the trusted weight can be obtained.In the ant colony optimization algorithm, if the optimization process includes m ants, each ant has to evaluate the combination of O dimensional data sources.This paper determines the data source fusion order of data fusion based on the retrieval path and constructs a running situation multi-source data fusion model, as shown in Figure 1.

Figure 1. Run the situation multi-source data fusion model
As shown in Figure 1, line loss rate is the most important parameter in the active distribution network.The line loss rate is determined by both the active power and the line loss power, so the line loss rate is the line loss rate of a group of lines.If the calculated line loss rate is a single line loss rate, special instructions should be given.The calculation formula of the line loss rate is: where i P represents the active power; p ' represents the line loss power; N i H represents the regularization residual; L represents the line set; i represents a line.
In addition to line loss rate, light load rate can also reflect the operation status of the distribution network.By analyzing the line group, the actual operating current on the line can be obtained.By comparing it with the rated current, the light load rate can be calculated using the formula.
Ant Colony Search

Real-Time Data
Historical Data data bus

Static stream data integration
Multi data fusion data where 1 i P represents the rated operating power; 1 Ri P represents the actual operating power.The device is a light load device when the load rate is less than 20%; the device is heavy equipment when the load rate exceeds 20%.
The line loss rate and light load rate are taken as the situational awareness parameters of the active distribution network.The two indexes are calculated by combining with the ant colony algorithm and fusion rules to provide the main parameters for the operation state identification and perception of the active distribution network.

Security situation perception of active power distribution network
By analyzing the operational status indicators, the operational status of future active distribution networks can be obtained.The multi-source data fusion information entropy is sued to analyze zerosequence voltage, over-limit voltage, overload voltage, and future state voltage in order to determine their threat to the active distribution network and better understand the security situation of the distribution network.

Based on multi-source data fusion information
There is great uncertainty in the operation state of the active distribution network, which leads to the high entropy of fusion information.Therefore, the size of entropy can be used to effectively process the training set and reduce the number of samples.When the sample distribution is balanced, internal samples are in the high-density area, and external samples are in the low-density area.During the multisource data fusion process, the information entropy can be used to express the data density.Based on this method, this paper proposes the method of reducing the entropy training sample to find the entropy value according to the distance between the two samples and to compare it with the information entropy.This is to determine the training point.The Euclidean distance between each sample during the operation of the active distribution network is: where x g x g ǃ indicate the positions of the two samples respectively.According to the results, the sample similarity is evaluated.If the similarity is high, the distance between the two is small, while the distance is large.The information entropy is introduced as a sample similarity measure, with the following formula: , then the sample belongs to the high-density area, namely the normal operation state range of the active distribution network.

Running state sensing based on voltage sequence decomposition
By using a multi-source data fusion method to decompose the voltage sequence after state recognition, operational state perception results can be obtained.Before decomposing the voltage sequence, it is necessary to first fuse multi-source voltage data, eliminate unreliable data, and make the perception results more accurate.By using a multi-source data fusion method to distinguish between trusted and untrusted voltage data, the formula is: where represent the amount of trusted and untrusted voltage data, respectively; z represents the intersection of data 1 2 g g ǃ .Through the standardization of the above formula, the relative importance of the voltage sequence is determined, the non-credible data is eliminated, and the extracted voltage sequence is: , , , According to Formula (6), four voltages, including zero-order voltage, over-limit voltage, overload voltage, and future state voltage, are obtained in the 1 ñ n t t .The operating state sensing results of these four voltages are as follows: 1) Residual voltage.The neutral point adopts the operation mode of ungrounding.When the singlephase ground or double-phase ground failure problem occurs, the zero-sequence voltage will rise sharply.Therefore, the current safety situation of the system can be judged according to the sudden change of the zero-order voltage.
2) Voltage exceeding the limit.The voltage over limit margin indicator is the current-voltage safety status and future safety development trend of the distribution network, such as short circuit faults, input, and removal of power equipment, etc.The formula for calculating the margin index of voltage exceeding the limit is: 3) Overload voltage.Facing the problem of safe overload of branch current, the overload severity index is set, which can truly reflect the current distribution network, operation status, and development trend after a period of operation.The overload severity index can be expressed as follows: where 0 L represents the overload value, this value means that in the active distribution network, once there is a safety accident, the electric device will have the danger of electric power overload.The overload severity indicator is 0 if the current exceeds 80%.
4) Future state voltage.The phase change of the future state voltage is the voltage of the distribution network after a period of operation.The phase change is very stable.The phase change of the node voltage during the prediction period is as follows:

Analytic hierarchy process for correcting perceived data
Based on determining the perception results, the weights of operational situation indicators are calculated.The existing weight analysis methods have low accuracy, so this article uses the AHP method to calculate the weight and analyze the importance of each weight.When the obtained feature vector matrix meets the consistency detection requirements, the maximum value of the matrix is normalized, and its final weight value is obtained.In the intelligent distribution network situational awareness algorithm based on state evaluation, the proportional parameter CR represents the compatibility index, and its calculation formula is as follows: CI CR RI (10) where CI represents the consistency index; RI represents the randomness index.When the CR value is less than 0.1, the resulting vector matrix meets the compatible consistency condition; when the CR value is greater than 0.1, the resulting vector matrix cannot meet the compatible consistency requirement and needs to be modified.The modified formula is: 11) where ( represents the state removal matrix; L represents the system parameter error; t represents time; J represents the control vector; A B X X ǃ represent the horizontal and tilted components respectively.When the operation state of the active distribution network changes, the above variables can be predicted, so the correction result error is small, and accurate perception data can be obtained.

Multi-frequency common-mode signal injection device
A current transformer is used to generate monitoring signals, which are injected into the active distribution network through the neutral point of the transformer.A multi-frequency common mode monitoring signal injection device is designed.In the normal operation mode of the system, the power electronic switch T1 is in the on state and T2 is in the off state, and the current transformer is working normally.In monitoring mode, T1 switches states are switched at frequency 1 f under the PWM control signal, generating a monitoring voltage of frequency 1 0 nf f r , where 0 f is the system frequency.T2 is switched in the on state to superimpose the generated monitoring voltage with the neutral point of the system transformer, thereby generating a common mode leakage current in the cable, which can be used for subsequent monitoring.

Experimental results and analysis
Using the above experimental conditions as the premise for experimental verification, the abnormal data observed during the operation of the actual active distribution network will be used as the experimental control group.The values of the control group fluctuate within a range of 40-50, numerical comparisons will be made referencing [14][15], and the reliability of the research methods will be verified.The number of experiments is 80. Figure 2 shows the comparison of abnormal data perception results using different methods.

Figure 2. Comparison of abnormal data perception results by different methods
From Figure 2, it can be seen that using the research method can quickly perceive the number of abnormal data, while using the methods proposed in [14][15] to perceive abnormal data remains at 60-70 and 55-65, respectively, which deviates greatly from the actual situation.The experimental results demonstrate that using the research method can effectively perceive abnormal data.It can help to achieve data correction and processing and truly reflect the operating status of the active distribution network.

Different running state-aware results
The operation state sensing research method of an active distribution network based on multi-source data fusion is used to analyze the perception results in different operating states.
1) Residual voltage.The ungrounded zero sequence voltage jump is used to identify the zero sequence voltage of the injection point.Then the safety problems existing in the active distribution network are found.At 0~2 s and 8~10 s, the node voltage value is 0, and then the active distribution network is in a normal operation state.At 2~8 s, the node has a ground fault.At this time, the active distribution network does not consider load fluctuation, and the zero-order voltage value has a large fluctuation range.
2) Voltage exceeding the limit.Safety situation information of injection points is collected.The risk of voltage exceeding the limit at each node of the system is analyzed, and voltage exceeding data is obtained without considering distribution load fluctuations.
Analysis shows that at 0-2 s, the margin value for voltage exceeding the limit at each node is 0.92, indicating that the active distribution network is in normal operation; At 2 to 12 s, the margin value for voltage exceeding the limit increases from 0.92 to 1.00, indicating that there is a risk of voltage exceeding the limit in the active distribution network.Large voltage fluctuations can cause the distribution network to malfunction and may also lead to the problem of burning electrical appliances.
3) Overload voltage.Under normal working conditions, the existence of a no-overload branch road is fully considered, and the resulting overload severity index is shown in Table 1.It can be seen from Table 1: during the 0~2 s period, the active distribution network is operating normally, with no overload branch, which is in a normal operation state; during the 2~6 s period, the active distribution network has branch overload, and the overload value is greater than other branches, there is overload warning; and the overload severity value of the active distribution network is 0, indicating that the active distribution network is restored to the normal operation state.Number of experiments reference [14] reference [15]   From Figure 3, it can be seen that during 2-7 s, due to load fluctuation input and interruption, the voltage phase change exceeded the stable range, causing it to enter an unstable state.At this time, the operation of the active distribution network was greatly threatened by dynamic factors, resulting in abnormal operation.

Conclusion
This article studies the intelligent perception method of distribution network state based on multi-source data fusion.It solves the problem of the difference between the state perception range and the actual range caused by the abnormal operation status of active distribution networks.The experimental results validate the effectiveness and feasibility of the proposed method.The perceived abnormal data fluctuates within 40 to 50 ranges, which is consistent with the actual value fluctuation range and can obtain accurate perception results under different operating conditions.This indicates that the intelligent perception method of distribution network state based on multi-source data fusion can improve the accuracy and reliability of distribution network state perception.
The results of this study are of great significance for the operation and maintenance of distribution networks.Accurately perceiving the operational status of the distribution network can provide reliable data support, help operators detect abnormal situations in a timely manner, and take corresponding measures, thereby improving the stability and reliability of the distribution network.At the same time, the proposed multi-source data fusion method also provides useful ideas and methods for data fusion problems in other fields.It has important application prospects and research value in improving the accuracy and reliability of distribution network state perception.

U
represents the voltage value of the j node; up low U U ǃ represent the upper and lower voltage limits respectively.

Figure 3 .
Figure 3. Future state voltage safety situation diagram

Table 1 .
The severity indicator of the overload