Direct current bus interference anomaly detection method based on dynamic local outlier factor algorithm

Abstract

The invention discloses a DC bus interference anomaly detection method based on a dynamic local outlier factor algorithm. The DC bus interference anomaly detection method is especially suitable for real-time operation state monitoring of a complex power system. Aiming at the limitation that a traditional anomaly detection method is high in calculation complexity and limited in detection precision in high-dimensional time series data processing, optimization is carried out on the basis of an outlier detection theory, a dynamic mode updating algorithm is provided, the algorithm evaluates outlier characteristics of time series data through local density, and the detection precision of the time series data is improved. The global calculation overhead is obviously reduced, and the real-time performance and the detection efficiency of the system are improved. Compared with a traditional method, the method consumes manual intervention, is high in detection precision and strong in early warning, can be widely applied to the field of health assessment and intelligent operation maintenance of power equipment, and shows a good engineering application prospect.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault diagnosis, and in particular relates to a DC bus interference anomaly detection method based on a dynamic local outlier factor algorithm. Background Art

[0002] With the rapid development of smart grids and power systems, DC power supply systems are becoming increasingly important in various industrial and engineering applications. In particular, in the fields of power transmission, renewable energy access, and microgrids, DC power supply systems are gradually becoming mainstream. However, the operating environment of complex DC power supply systems is full of uncertainty, and various faults or abnormalities may occur in the system, which will affect the stability and safety of the system. Therefore, how to achieve real-time monitoring of the system operation status, especially abnormality detection and fault diagnosis, has become an urgent problem to be solved.

[0003] At present, anomaly detection methods based on local outlier factors (LOF) are widely used in the recognition of anomaly patterns in time series data. These methods can effectively mine outliers in data and improve the accuracy of anomaly detection to a certain extent. However, there are still some challenges when processing time series data of complex DC power supply systems. First, due to the huge amount of data in DC power supply systems and the strong temporal and nonlinear characteristics of the data, traditional local outlier factor methods often face the problem of high computational complexity, especially in the real-time detection of large-scale data streams, which cannot meet the requirements of high efficiency. Secondly, traditional methods have poor adaptability when processing dynamic data streams and cannot quickly respond to sudden changes or changes in system status. Furthermore, existing methods usually fail to fully consider the operating environment characteristics of the system, such as voltage fluctuations, load changes and other factors, which affects the accuracy and robustness of anomaly detection.

[0004] Therefore, an improved anomaly detection algorithm is urgently needed for the time series data of complex DC power supply systems. The algorithm should be able to improve the ability to capture abnormal patterns while ensuring high real-time performance and be able to adapt to changes in dynamic data streams. By optimizing the model structure, simplifying the calculation process and introducing a new feature recognition mechanism, the detection accuracy can be improved while reducing the calculation complexity, so as to better meet the needs of complex DC power supply systems in actual engineering. Summary of the invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a DC bus interference anomaly detection method based on a dynamic local outlier factor algorithm. It effectively solves the shortcomings of traditional algorithms in real-time data processing and improves the ability to identify complex load switching patterns. It can accurately identify a variety of abnormal working conditions and has wider applicability and engineering application value.

[0006] Technical solution: A DC bus interference anomaly detection method based on a dynamic local outlier factor algorithm of the present invention includes:

[0007] Step 1: Build an experimental platform, including a DC power supply, a DC / DC converter, and a programmable electronic load to simulate actual operating conditions. Place sensors at the power input and load output to collect key electrical signals, and use a multimeter and oscilloscope to measure static parameters and dynamic waveforms to generate multi-condition data for algorithm verification.

[0008] Step 2: An improved dynamic local outlier factor algorithm is proposed to analyze outliers in the time series data stream and detect anomalies. The algorithm calculates the outlier factor based on local density differences and enhances its adaptability to multi-condition data through dynamic pattern analysis. The outliers are gradually updated and iteratively calculated in the time series data to achieve rapid identification of outliers.

[0009] Step 3: Establish typical load abnormality patterns to simulate electrical stress changes under different working conditions. Load abnormality patterns include reciprocating step switching, reciprocating ramp switching, and pulse switching, which are simulated by programmable electronic loads. Each pattern reflects the characteristics of load changes under common operating conditions and is used to verify the robustness and accuracy of the detection algorithm.

[0010] Step 4: Evaluate the performance of the detection algorithm, including false alarm rate, missed alarm rate, and detection time. Through offline analysis of multiple sets of experimental data, the false alarm rate and missed alarm rate are calculated to evaluate the detection accuracy of the algorithm, and the total time required to complete anomaly detection is recorded to verify real-time performance. The experimental results provide a quantitative basis for optimizing the algorithm.

[0011] Furthermore, the method for constructing the experimental platform described in step 1 specifically includes:

[0012] S1-1. Design the hardware architecture of the experimental platform, including DC power supply, DC / DC converter, programmable electronic load and equivalent cable circuit, to simulate the operating environment of the actual DC power supply system. The platform architecture supports multiple load change modes to collect typical abnormal working condition data.

[0013] S1-2. High-precision sensors are arranged at the input end of the DC power supply and the output end of the load to collect voltage, current and ripple signals to ensure that the operating characteristics of the system under dynamic and static conditions can be captured.

[0014] S1-3. Use a high-precision multimeter to measure the static electrical parameters of the power supply and load nodes, including DC voltage and current. Use a high-bandwidth oscilloscope to record the dynamic voltage ripple waveform to capture abnormal signals such as switch state jumps and load transients.

[0015] Furthermore, the method for proposing an improved algorithm in step 2 is characterized in that S2 specifically includes:

[0016] S2-1, Data segmentation processing: The sampled time series data is divided into initial segments and update segments. The initial segment is used to build a stable reference data set, and its length is longer than the update segment; the update segment appends data in real time and combines with the initial segment to form a dynamic data stream, providing continuous input for outlier detection.

[0017] S2-2. Calculation of local outlier factors: Based on the local outlier theory, the key calculation steps are defined, including the following:

[0018] First, the k-distance and k-distance neighborhood of the sample points are calculated to identify the local neighborhood range of each point:

[0019] For any positive integer k, the k-distance represents the distance from point p to the nearest point o within k. In the data set D, points o and o′ should satisfy the following constraints: there are k points o′ that satisfy d(p,o′)≤d(p,o), and at least k-1 points that satisfy d(p,o′)<d(p,o). By definition, point o is not unique.

[0020] When calculating the k-distance neighborhood of point p, for point p, the k-distance neighborhood represents the set of points whose distance from point p does not exceed k:

[0021] N k-distance(p) (p)={q∈D\{p}∣d(p,q)≤k-distance(p)}

[0022] Among them, N k-distance(p) (p) can be simplified to N k (p).

[0023] Secondly, calculate the reachable distance between points p and o. For a fixed k, the reachable distance between points p and o is defined as the maximum value of the k distance and the actual distance between points p and o:

[0024] reach-dist k (p,o)=max{k-distance(o),d(p,o)}

[0025] When p1 is within the circular area formed by the distances between points o and k: reach-dist k (p1,o) = k-distance(o), on the contrary, when p1 is outside the circular area: reach-dist k (p2,o)=d(p2,o), the definition of the reachable distance significantly reduces the statistical fluctuation of the distance d(p,o) through smoothing.

[0026] Then, based on the neighborhood density, the local reachable density of each point is calculated, and the degree of abnormality is characterized by measuring the density relationship between the point and its neighboring points:

[0027] The local reachability density of a point p is defined as the inverse of the average reachability distance within its MinPts neighborhood:

[0028]

[0029] When |N MinPts When (p)|=0, lrd MinPts (p) will tend to infinity.

[0030] Finally, the local outlier factor is defined, and the degree of outlier is measured by the ratio of the average local reachable density of points in the neighborhood to the local reachable density of the current point. The local outlier factor (LOF) of point p is defined as the ratio of the average local reachable density of points in the MinPts-neighborhood of point p to the local reachable density of point p:

[0031]

[0032] LOF MinPts (p) represents the ratio of the average local reachability density of point p in the MinPts neighborhood to the local reachability density of point p. Through the basic steps of LOF, the local outlier factor LOF of the point can be calculated MinPts (p), thereby detecting outliers in the sampled data set.

[0033] S2-3. Dynamic mode algorithm framework design: The traditional static local outlier factor algorithm is improved into a dynamic mode algorithm framework to adapt to the characteristics of time series data. By introducing an iterative update mechanism, the newly collected update segment data is processed in real time, and the continuity of the algorithm for the entire time series data is ensured while dynamically calculating the local outlier values ​​of the newly added data points. This design effectively solves the problems of low computational efficiency and difficulty in continuous monitoring in the traditional outlier factor algorithm in real-time data processing, thereby improving the timeliness and applicability of anomaly detection.

[0034] S2-4. Result visualization and abnormal alarm: Use monitoring charts to dynamically display the outlier factor results and compare the outlier value with the set threshold. Once the outlier value exceeds the preset threshold, an abnormal alarm is triggered and the detection results are fed back in real time. At the same time, adaptive learning is performed on the data segments that do not trigger the alarm to continuously optimize the model parameters.

[0035] Furthermore, the method for establishing the abnormal mode in step 3 is specifically as follows:

[0036] S3-1. Construct three typical abnormal modes: According to the actual operating characteristics and potential fault conditions in the DC power supply and distribution system, three typical abnormal working modes are designed, including reciprocating step switching, reciprocating ramp switching and pulse switching. The reciprocating step switching mode simulates the large-scale start and stop process of the load, which is common in the sudden start or stop of high-power equipment; the reciprocating ramp switching mode reflects the rise and fall of the load power; the pulse switching mode simulates the sudden short-term high-power load operation, such as the short-term impact load of the switching circuit.

[0037] S3-2. Use programmable electronic load to simulate the electrical stress changes of each abnormal mode: In the experimental platform, the electrical load change curves corresponding to the above three abnormal modes are generated respectively through the precise setting of the programmable electronic load. In the reciprocating step switching mode, the electronic load is set to alternately present high-power and low-power states at different time intervals; in the reciprocating ramp switching mode, the power of the programmed control load presents a linear increasing or decreasing trend; in the pulse switching mode, the load is set to generate instantaneous high-power pulses in a short time to simulate sudden stress. The above settings ensure the authenticity and diversity of the abnormal modes.

[0038] S3-3. Record input voltage fluctuations and output power changes under abnormal conditions: In each abnormal mode, the experimental platform is used to collect corresponding operating data, including voltage fluctuations at the input end and power change characteristics at the output end. In the experiment, dynamic electrical parameters are recorded in real time by high-precision multimeters and oscilloscopes. The collected data is annotated and marked as abnormal data and normal data, forming a comparison sample set to provide high-quality experimental data support for subsequent training and testing of algorithm models. These annotated samples can be used to analyze the electrical stress characteristics under different abnormal modes, thereby further verifying the effectiveness and robustness of the detection algorithm.

[0039] Furthermore, the method for evaluating the performance of the detection algorithm described in step 4 and outputting the evaluation result is characterized in that the specific process is as follows:

[0040] S4-1. Define false alarm rate, missed alarm rate and detection time as evaluation criteria: To evaluate the effectiveness of the improved dynamic mode local outlier factor algorithm, three core evaluation indicators are set, namely false alarm rate, missed alarm rate and detection time. The false alarm rate is used to quantify the proportion of abnormal points that the system mistakenly determines as abnormal points when there are no abnormalities, reflecting the false alarm level of the algorithm; the missed alarm rate measures the proportion of abnormal points that the system fails to correctly identify when there are abnormalities, reflecting the missed alarm performance of the algorithm; the detection time indicates the time required to complete the entire abnormality detection process, which is directly related to the real-time performance and efficiency of the algorithm.

[0041] S4-2. Identify and diagnose abnormal patterns, and verify their superiority through evaluation indicators: Under three typical abnormal modes, use the multi-operating data collected by the experimental platform to identify and diagnose abnormal patterns. For the results of each mode, calculate the false alarm rate, missed alarm rate and detection time. Through experimental verification, the improved algorithm significantly improves the detection efficiency while reducing the false alarm rate and missed alarm rate, verifying its superiority and applicability in time series anomaly detection.

[0042] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0043] (1) The present invention proposes a dynamic mode local outlier factor algorithm, which realizes the dynamic extraction and real-time analysis of abnormal features by sampling the multi-time series data of the main bus of the DC power supply and distribution system, improves the accuracy and real-time performance of anomaly detection, and provides reliable technical support for the health monitoring of the power system.

[0044] (2) By proposing an improved dynamic mode algorithm framework, the present invention optimizes the static outlier factor algorithm into a dynamic mode that can process time series, effectively solving the shortcomings of traditional algorithms in real-time data processing, and at the same time improving the ability to recognize complex load switching patterns.

[0045] (3) The present invention realizes accurate identification of various abnormal working conditions by constructing typical load abnormality patterns (including reciprocating step switching, reciprocating ramp switching and pulse switching) and verifying the performance of the detection algorithm, and has wider applicability and engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of a DC bus interference anomaly detection method based on a dynamic local outlier factor algorithm in one embodiment of the present invention;

[0047] Figure 2 This is a structural diagram of an experimental platform in one embodiment of the present invention;

[0048] Figure 3 This is a diagram showing the detection performance in one embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to further understand the present invention, preferred embodiments of the present invention are described below in conjunction with examples. However, it should be understood that these descriptions are only for further illustrating the features and advantages of the present invention, rather than limiting the claims of the present invention.

[0050] The description in this section is only for typical embodiments, and the present invention is not limited to the scope of the embodiments. The combination of different embodiments, the mutual replacement of some technical features in different embodiments, and the mutual replacement of the same or similar prior art means with some technical features in the embodiments are also within the scope of the description and protection of the present invention.

[0051] S1. Build an experimental platform.

[0052] An experimental platform is constructed, and the upstream of the DC / DC converter includes a DC power supply and an equivalent circuit that replaces the cable function. Considering the possibility of switch state jumps and load transients in actual DC power systems, a programmable electronic load is used downstream of the DC / DC converter to simulate actual electrical equipment. Based on this platform, sample data under various abnormal conditions are collected through a series of simulation experiments. The voltage, current, and ripple of the node can be measured directly, while the power and efficiency are obtained by indirect calculation. In order to ensure the accuracy of static testing, the DC components of voltage and current are measured using a high-precision multimeter. The function of the oscilloscope is to meet the requirements of high-bandwidth ripple and dynamic testing.

[0053] S2. Propose an improved dynamic local outlier factor algorithm.

[0054] Data segmentation processing: The sampled time series data is divided into initial segments and update segments. The initial segment is used to build a stable reference data set, and its length is much longer than the update segment; the update segment appends data in real time and combines with the initial segment to form a dynamic data stream, providing continuous input for outlier detection.

[0055] Local outlier factor calculation: Based on the local outlier theory, the key calculation steps are defined, including the following:

[0056] First, the k-distance and k-distance neighborhood of the sampling point are calculated to identify the local neighborhood range of each point; secondly, the reachable distance between points p and o is calculated. For a fixed k, the reachable distance between points p and o is defined as the maximum value of the k-distance and the actual distance between points p and o; then, based on the neighborhood density, the local reachable density of each point is calculated, and its abnormality is characterized by measuring the density relationship between the point and its neighboring points; finally, the local outlier factor is defined, and the degree of outlier is measured by the ratio of the average local reachable density of the points in the neighborhood to the local reachable density of the current point.

[0057] Dynamic mode algorithm framework design: The traditional static local outlier factor algorithm is improved into a dynamic mode algorithm framework to adapt to the characteristics of time series data. By introducing an iterative update mechanism, the newly collected update segment data is processed in real time, and the continuity of the algorithm for the entire time series data is ensured while dynamically calculating the local outlier values ​​of the newly added data points. This design effectively solves the problems of low computational efficiency and difficulty in continuous monitoring in the traditional outlier factor algorithm in real-time data processing, thereby improving the timeliness and applicability of anomaly detection.

[0058] Result visualization and abnormal alarm: Use monitoring graphs to dynamically display outlier factor results and compare outliers with set thresholds. Once an outlier exceeds the preset threshold, an abnormal alarm is triggered and the detection results are fed back in real time. At the same time, adaptive learning is performed on the data segments that do not trigger the alarm to continuously optimize the model parameters.

[0059] S3. Establish typical load abnormality patterns to simulate electrical stress changes under different working conditions.

[0060] According to the actual operating characteristics and potential fault conditions in the DC power supply and distribution system, three typical abnormal working modes are designed, including reciprocating step switching, reciprocating ramp switching and pulse switching. The reciprocating step switching mode simulates the large-scale start and stop process of the load, which is common in the sudden start or stop of high-power equipment; the reciprocating ramp switching mode reflects the rise and fall of the load power; the pulse switching mode simulates the sudden short-term high-power load operation, such as the short-term impact load of the switching circuit.

[0061] In the experimental platform, the electrical load change curves corresponding to the above three abnormal modes are generated respectively through the precise setting of the programmable electronic load. In each abnormal mode, the experimental platform is used to collect the corresponding operating data, including the voltage fluctuation at the input end and the power change characteristics at the output end. In the experiment, the dynamic electrical parameters are recorded in real time by high-precision multimeters and oscilloscopes. The collected data are annotated and marked as abnormal data and normal data respectively to form a comparison sample set, providing high-quality experimental data support for the training and testing of subsequent algorithm models. These annotated samples can be used to analyze the electrical stress characteristics under different abnormal modes, so as to further verify the effectiveness and robustness of the detection algorithm.

[0062] S4. Construct evaluation indicators and experimentally verify the detection effect of the model.

[0063] In order to evaluate the effectiveness of the improved dynamic mode local outlier factor algorithm, three core evaluation indicators are set, namely false alarm rate, missed alarm rate and detection time. The false alarm rate is used to quantify the proportion of abnormal points that the system mistakenly determines as abnormal points when there are no abnormalities, reflecting the false alarm level of the algorithm; the missed alarm rate measures the proportion of abnormal points that the system fails to correctly identify when there are abnormalities, reflecting the missed alarm performance of the algorithm; the detection time indicates the time required to complete the entire abnormality detection process, which is directly related to the real-time performance and efficiency of the algorithm.

[0064] like Figure 2 As shown in the figure, in order to verify the actual effect of the proposed algorithm, an experimental platform was built. The upstream of the DC / DC converter includes a DC power supply and an equivalent circuit that replaces the cable function. Considering the possibility of switch state jumps and load transients in actual DC power systems, a programmable electronic load is used downstream of the DC / DC converter to simulate actual electrical equipment. Based on this platform, sample data under various abnormal conditions are collected through a series of simulation experiments.

[0065] Table 1

[0066]

[0067] like Figure 3 As shown in Figure 1, the improved algorithm is used to verify three typical abnormal load conditions in DC power systems. For easy identification, this paper marks the abnormal data in red and the normal data in blue.

[0068] Table 1 lists the evaluation indicators of the algorithm under three load abnormal conditions. For the reciprocating step switching mode, the false alarm rate is 0.47%, the missed alarm rate is 0.46%, and the detection time is 80 microseconds, which can meet the needs of real-time anomaly detection. In the reciprocating ramp switching mode, the false alarm rate is only 0.012%, which shows the good false alarm control ability of the algorithm. However, due to the gradual change of load in the mode, the difficulty of anomaly detection increases, resulting in a missed alarm rate of 6.92%. At the same time, the detection time of this mode is about twice that of the reciprocating step mode. In contrast, the pulse switching mode is the most easily detected abnormal mode by the algorithm, because the load fluctuation caused by this mode is significant and the data morphology is highly different, making the abnormal characteristics more obvious. The experimental results show that the algorithm has strong robustness and applicability in anomaly detection of complex time series data.

[0069] The description and application of the present invention here are illustrative, and it is not intended to limit the scope of the present invention to the above-mentioned embodiments. The effects or advantages involved in the specification may not be reflected in the actual experimental examples due to the uncertainty of specific condition parameters or other factors, and the effects or advantages are not used to limit the scope of the invention. The deformation and change of the embodiments disclosed here are possible, and the replacement of the embodiments and the various equivalent parts are well known to those of ordinary skill in the art. It should be clear to those skilled in the art that the present invention can be implemented in other forms, structures, arrangements, proportions, and with other components, materials and parts without departing from the spirit or essential features of the present invention. Other deformations and changes can be made to the embodiments disclosed here without departing from the scope and spirit of the present invention.

Claims

1. A DC bus interference anomaly detection method based on a dynamic local outlier factor algorithm, characterized in that: The steps include: S1. Build an experimental platform, including a DC power supply, a DC / DC converter, and a programmable electronic load to simulate actual operating conditions; place sensors at the power input and load output to collect key electrical signals, measure static parameters and dynamic waveforms, and generate multi-operating condition data for algorithm verification; S2. An improved dynamic local outlier factor algorithm is proposed to analyze outliers in time series data streams and detect anomalies. The algorithm calculates outlier factors based on local density differences and enhances adaptability to multi-condition data through dynamic pattern analysis. Outliers are gradually updated and iteratively calculated in time series data to achieve rapid identification of outliers. S3. Establish typical load abnormality modes to simulate the changes in electrical stress under different working conditions; load abnormality modes include reciprocating step switching, reciprocating ramp switching and pulse switching, which are simulated by programmable electronic loads; S4. Evaluate the performance of the detection algorithm and construct an evaluation including false alarm rate, missed alarm rate and detection time. Through offline analysis of multiple sets of experimental data, calculate the false alarm rate and missed alarm rate to evaluate the detection accuracy of the algorithm, and record the total time to complete anomaly detection to verify real-time performance.

2. The DC bus interference anomaly detection method based on dynamic local outlier factor algorithm according to claim 1 is characterized in that: Step S1 specifically includes the following steps: S1-1. Design the hardware architecture of the experimental platform, including DC power supply, DC / DC converter, and programmable electronic load; S1-2. High-precision sensors are arranged at the input end of the DC power supply and the output end of the load to collect voltage, current and ripple signals to ensure that the operating characteristics of the system under dynamic and static conditions can be captured; S1-3. Use a high-precision multimeter to measure the static electrical parameters of the power supply and load nodes, including DC voltage and current; record the dynamic voltage ripple waveform through a high-bandwidth oscilloscope to capture abnormal signals of switch state jumps and load transients.

3. The DC bus interference anomaly detection method based on dynamic local outlier factor algorithm according to claim 1 is characterized in that: Step S2 specifically includes the following steps: S2-1, data segmentation processing: the sampled time series data is divided into initial segments and update segments. The initial segment is used to build a stable reference data set, and its length is longer than the update segment. The update segment appends data in real time and combines with the initial segment to form a dynamic data stream, providing continuous input for outlier detection. S2-2. Calculation of local outlier factors: Based on the local outlier theory, define the key calculation steps; S2-3. Dynamic mode algorithm framework design: Improve the static local outlier factor algorithm to a dynamic mode algorithm framework to adapt to the characteristics of time series data. By introducing an iterative update mechanism, the newly collected update segment data is processed in real time. While dynamically calculating the local outlier values ​​of the newly added data points, the continuity of the algorithm for the entire time series data is ensured. S2-4, Result visualization and abnormal alarm: Use monitoring charts to dynamically display the outlier factor results and compare the outlier values ​​with the set thresholds; When the outlier value exceeds the preset threshold, an abnormal alarm is triggered and the detection results are fed back in real time. At the same time, adaptive learning is performed on the data segments that do not trigger the alarm to optimize the model parameters.

4. The DC bus interference anomaly detection method based on dynamic local outlier factor algorithm according to claim 3 is characterized in that: Step S2-2 specifically includes the following steps: Step S2-2-1, calculate the k-distance and k-distance neighborhood of the sampling point to identify the local neighborhood range of each point: For any positive integer k, the k-distance represents the distance from point p to the nearest point o within k. In the data set D, points o and o′ should satisfy the following constraints: there are k points o′ satisfying d(p,o′)≤d(p,o), and at least k-1 points satisfying d(p,o′)<d(p,o); by definition, point o is not unique; When calculating the k-distance neighborhood of point p, for point p, the k-distance neighborhood represents the set of points whose distance from point p does not exceed k: N k-distance(p) (p)={q∈D\{p}∣d(p,q)≤k-distance(p)} Among them, N k-distance(p) (p) can be simplified to N k (p); Step S2-2-2, calculate the reachable distance between points p and o. For a fixed k, the reachable distance between points p and o is defined as the maximum value of the k distance and the actual distance between points p and o: reach-dist k (p,o)=max{k-distance(o),d(p,o)} When p1 is within the circular area formed by the distances between points o and k: reach-dist k (p1,o) = k-distance(o). On the contrary, when p1 is outside the circular area: reach-dist k (p2,o) = d(p2,o), the definition of the reachable distance significantly reduces the statistical fluctuation of the distance d(p,o) through smoothing; Step S2-2-3: Based on the neighborhood density, calculate the local reachable density of each point, and characterize its abnormality by measuring the density relationship between the point and its neighborhood points: The local reachability density of a point p is defined as the inverse of the average reachability distance within its MinPts neighborhood: When |N MinPts (p)|=0, lrd MinPts (p) will tend to infinity; Step S2-2-4, define the local outlier factor, and use the ratio of the average local reachable density of points in the neighborhood to the local reachable density of the current point to measure the degree of outlier; the local outlier factor LOF of point p is defined as the ratio of the MinPts of point p - the average local reachable density of points in the neighborhood to the local reachable density of point p: LOF MinPts (p) represents the ratio of the average local reachability density of point p in the MinPts neighborhood to the local reachability density of point p; the local outlier factor LOF of the point is calculated through the basic steps of LOF MinPts (p), thereby detecting outliers in the sampled data set.

5. The DC bus interference anomaly detection method based on dynamic local outlier factor algorithm according to claim 1, characterized in that: Step S3 specifically includes the following steps: S3-1. Construct three typical abnormal modes: According to the actual operating characteristics and potential fault conditions in the DC power supply and distribution system, three typical abnormal working modes are designed, including reciprocating step switching, reciprocating ramp switching and pulse switching; the reciprocating step switching mode simulates the large-scale start and stop process of the load, which is used for sudden start or stop of high-power equipment; the reciprocating ramp switching mode reflects the rise and fall changes of load power; the pulse switching mode simulates sudden short-term high-power load operation; S3-2. Simulate the electrical stress changes of each abnormal mode through programmable electronic load: In the experimental platform, the electrical load change curves corresponding to the above three abnormal modes are generated respectively through the precise setting of the programmable electronic load; in the reciprocating step switching mode, the electronic load is set to alternately present high-power and low-power states at different time intervals; in the reciprocating ramp switching mode, the power of the programmed control load presents a linear increasing or decreasing trend; in the pulse switching mode, the load is set to generate instantaneous high-power pulses in a short time to simulate sudden stress; S3-3. Record input voltage fluctuations and output power changes under abnormal conditions: In each abnormal mode, the experimental platform is used to collect corresponding operating data, including voltage fluctuations at the input end and power change characteristics at the output end; dynamic electrical parameters are recorded in real time through high-precision multimeters and oscilloscopes; the collected data are labeled and marked as abnormal data and normal data respectively to form a comparison sample set, which provides experimental data support for the training and testing of subsequent algorithm models. These labeled samples can be used to analyze the electrical stress characteristics under different abnormal modes and further verify the effectiveness and robustness of the detection algorithm.

6. The DC bus interference anomaly detection method based on dynamic local outlier factor algorithm according to claim 1, characterized in that: Step S4 specifically includes the following steps: S4-1. Define the false alarm rate, missed alarm rate and detection time as evaluation criteria; the false alarm rate is used to quantify the proportion of abnormal points that the system mistakenly determines as abnormal points when there is no abnormality, reflecting the false alarm level of the algorithm; the missed alarm rate measures the proportion of abnormal points that the system fails to correctly identify when there is an abnormality, reflecting the missed alarm performance of the algorithm; the detection time indicates the time required to complete the entire abnormality detection process, which is directly related to the real-time performance and efficiency of the algorithm; S4-2. Identify and diagnose abnormal patterns, and verify their superiority through evaluation indicators: Under three typical abnormal modes, use the multi-operating condition data collected by the experimental platform to identify and diagnose abnormal patterns; for the results of each mode, calculate the false alarm rate, missed alarm rate and detection time and verify them through experiments.

7. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method of claim 1.

8. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to claim 1 are implemented.

9. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to claim 1 are implemented.

Images