New energy power generation equipment monitoring method and system based on multi-source heterogeneous data

Abstract

The invention discloses a new energy power generation equipment monitoring method and system based on multi-source heterogeneous data, and the method comprises the steps: collecting state signal data in real time through a sensor group arranged on a wind turbine generator, carrying out the statistical analysis after data preprocessing, and forming a standardized multi-dimensional data matrix under a unified time reference; carrying out credibility analysis on each data sequence group by calculating the characteristic entropy Hs (t) of each data sequence group, calculating a time sequence resonance index when a data source is in a stable state, carrying out traversal summation, evaluating the energy collaboration among signals, and constructing a comprehensive health index to carry out health state evaluation when the wind turbine generator has energy response imbalance. When the wind turbine generator set is in the abnormal state, generating a fault alarm and transmitting the fault alarm to operation and maintenance personnel for processing; according to the invention, through fusion analysis of multi-source heterogeneous data, a unified multi-dimensional data matrix is constructed, and omnibearing and multi-dimensional accurate monitoring of the operation state of the new energy power generation equipment is realized.

Description

Technical Field

[0001] This invention belongs to the field of new energy technology, and in particular relates to a monitoring method and system for new energy power generation equipment based on multi-source heterogeneous data. Background Technology

[0002] With the acceleration of the global energy transition and the deepening of the "dual carbon" target, the proportion of new energy power generation in the power structure continues to increase, and clean energy sources such as wind power and photovoltaic power generation have become important components in building a new power system. The large-scale and clustered deployment of new energy power generation equipment places higher demands on operational reliability and maintenance efficiency, making real-time monitoring and intelligent diagnosis of equipment status an increasingly important focus of the industry. Against this backdrop, utilizing multi-source data fusion and analysis technologies to achieve full lifecycle health management of equipment has become an important direction for technological development in the new energy field.

[0003] In existing technologies, monitoring of new energy power generation equipment such as wind turbine generators typically employs analysis methods based on data from a single type of sensor. For example, vibration sensors monitor the mechanical state of the transmission system, or temperature sensors track the temperature rise of key components to assess the operating status of specific subsystems. Some systems further introduce time-series data analysis methods to perform threshold judgments or trend analysis on collected electrical parameters such as voltage, current, and power, thereby identifying obvious anomalies. These systems often rely on independent data acquisition modules and preset rule-based judgment mechanisms to monitor the local operating status of the equipment.

[0004] However, existing monitoring methods have significant limitations, primarily in their failure to effectively integrate the inherent correlations and collaborative changes among multi-source heterogeneous data. Because mechanical, electrical, and environmental data differ in acquisition frequency, data dimensions, and physical meaning, existing systems often only analyze each data source independently or simply overlay them, ignoring the temporal resonance and dynamic coupling relationships between cross-domain data. This fragmented data processing makes it difficult for the system to comprehensively capture the gradual changes in equipment status and early signs of faults, limiting the accuracy and timeliness of status assessment and fault warning. Summary of the Invention

[0005] Purpose of the invention: The purpose of this invention is to provide a monitoring method and system for new energy power generation equipment based on multi-source heterogeneous data. The aim is to achieve comprehensive and accurate monitoring of the operating status of new energy power generation equipment and timely early warning of early faults by integrating and analyzing multi-source heterogeneous data of new energy power generation equipment.

[0006] Technical solution: The new energy power generation equipment monitoring method of the present invention includes the following steps:

[0007] S1. Status signal data is collected in real time by a sensor array deployed on the wind turbine. After preprocessing the status signal data, aggregation and calibration processes are performed sequentially to form a single-cycle data block X(t). k ); for the single-cycle data block X(t) k Statistical analysis was performed to obtain the standardized signal Z. i,o (t k The standardized signal Z is then processed in chronological order. i,o (t k Assembled into a multidimensional data matrix JZ;

[0008] S2. Group the multidimensional data matrix JZ by category to obtain several data sequence groups; calculate the probability distribution p of each parameter in each data sequence group. o (Tw), and based on the probability distribution p o (Tw) Calculate the feature entropy H for each data sequence group. s (t); based on the characteristic entropy H s (t) Perform a reliability analysis on each data sequence group to determine whether the data source is in a stable state. If the data source is in a stable state, proceed to the next step.

[0009] S3, Based on the standardized signal Z i,o (t k Perform signal coordination analysis on different parameter signals of the wind turbine; sum the results of the signal coordination analysis, conduct average energy coordination analysis, and determine whether there is energy response imbalance in the wind turbine. If there is energy response imbalance in the wind turbine, proceed to the next step.

[0010] S4. Fit the reliability analysis results with the average energy synergy analysis results; perform a health status analysis on the overall operating status of the wind turbine based on the fitting results, and generate a health status assessment report for the wind turbine based on the health status analysis results.

[0011] Preferably, the status signal data acquired in step S1 includes:

[0012] The status signal data of the wind turbine is collected in real time by sensor groups deployed in the nacelle, tower and environmental measuring points of the wind turbine.

[0013] The sensor group includes strain gauges, accelerometers, speed encoders, current sensors, power meters, stator temperature sensors, anemometers, barometers, and humidity sensors.

[0014] By installing strain gauges, acceleration sensors, and speed encoders inside the wind turbine nacelle, the tower strain ε, vibration acceleration a, and main shaft speed ω are collected in real time.

[0015] By installing current sensors, power meters, and stator temperature sensors in the generator and electrical cabinet area, the current I, output power P, and stator temperature Ts are collected in real time.

[0016] By installing an anemometer, barometer, and humidity sensor at the wind vane on the top of the wind turbine tower, wind speed V, air density ρa, and air humidity H are collected in real time.

[0017] Preferably, step S1 further includes:

[0018] The state signal data is preprocessed using edge analysis nodes. The data preprocessing includes:

[0019] Noise filtering: Apply a low-pass filtering algorithm to continuously fluctuating state signal data to remove high-frequency noise components;

[0020] Time reference alignment: Periodic time synchronization is performed between the network time protocol and the data acquisition timestamps of the sensor group, so that the status signal data of different channels can be synchronized and aligned at the same time step;

[0021] After data preprocessing, the state signal data is periodically packaged. Based on a unified sampling period ΔT, the state signal data in all partitions are aggregated and calibrated according to the sensor's acquisition timestamp sequence to form a single-period data block X(t). k ),

[0022] Where X(t) k )={X i (t k Let X represent the set of multi-source heterogeneous state signal data of the wind turbine generator in each sampling period, where i = 1, 2, ..., n. k () represents a set of heterogeneous signal vectors with a unified timestamp, where each dimension represents a sensor type, i is the channel index (i∈[1,n]), n is the total number of sampling parameters, and t k For the k-th sampling period, X i (t k ) indicates that the i-th sensor channel is in t k Parameters collected at all times;

[0023] For each parameter sequence X i (t k Perform statistical analysis to obtain the sequence X of each parameter. i (t k The historical mean μ i With standard deviation σ i The standardized signal Z is obtained by performing standardization processing. i,o (t k ):

[0024]

[0025] Among them, Z i,o (t k ) represents the o-th parameter in the i-th sensing channel at t k The standardized signal at time o is the parameter index;

[0026] The standardized parameters are assembled into a unified multidimensional data matrix JZ according to the time series order, JZ=[ε, a, ω, I, P, Ts, V, ρa, H] T , where T represents the matrix transpose.

[0027] Preferably, step S2 includes:

[0028] Within a set sliding time window Tw, the multidimensional data matrix JZ is grouped by category to construct data sequence groups for mechanical signals, electrical signals, and environmental signals, respectively.

[0029] Normalize and statistically analyze the instantaneous amplitudes of each parameter in each data sequence group to obtain the probability distribution p of each parameter occurring within the current time window Tw. o (Tw) represents the probability distribution of the o-th parameter within the current time window Tw, where o is the parameter index;

[0030] Based on the probability distribution p o (Tw), the characteristic entropy H of each data sequence group is calculated using information entropy theory. s (t), the calculation formula is:

[0031]

[0032] Where n represents the number of parameters in the current data sequence group, and ln represents the natural logarithm function.

[0033] Preferably, step S2 further includes:

[0034] The average entropy value H of the wind turbine under historical stable operating conditions ref Based on the characteristic entropy H of each data sequence group, s (t) is fitted, and the reliability index Ltr of each data sequence group is calculated:

[0035]

[0036] When the credibility index Ltr < 1, it is determined that the data source of the current data sequence group has an abnormal disturbance trend, a feedback signal is sent to the data acquisition layer, and a resampling process is performed on the data acquisition channel that triggered the feedback signal.

[0037] When the confidence index Ltr≥1, it is determined that the data source of the current data sequence group is in a stable state, triggering signal collaborative analysis.

[0038] Preferably, step S3 includes:

[0039] When the data source is in a stable state, based on the standardized signal Z i,o (t k Signal coordination analysis is performed on the energy synchronization degree of signals with different parameters in the time domain.

[0040] The temporal resonance index Sgz between heterogeneous signals is calculated to characterize the dynamic similarity and energy resonance intensity between them in the time domain. The calculation formula is as follows:

[0041]

[0042] Among them, Sgz o,q Z represents the degree of energy synchronization between the normalized signals of the 0th and qth parameters in the time domain, where τ represents the time variable, Δτ represents the compensation delay, and Z represents the time difference. o (τ) and Z q (τ) represents the standardized signals of the 0th and qth parameters at time τ, respectively, and dτ represents the time integral variable.

[0043] Preferably, step S3 further includes:

[0044] The overall average resonance intensity Sgz is calculated by iterating and summing the temporal resonance indices Sgz among all heterogeneous signals. avg This represents the average energy coherence among all signal pairs:

[0045]

[0046] Where M represents the number of valid data sources participating in heterogeneous resonance analysis within the current calculation cycle;

[0047] The overall average resonance intensity Sgz avg Compare with a preset synergy threshold; when Sgz avg When the value is below the aforementioned synergy threshold, the wind turbine is determined to have energy response imbalance and drive control abnormality, triggering a health status analysis; when Sgz avg When the overall operation of the wind turbine is not lower than the aforementioned coordination threshold, it is determined that the wind turbine is operating in a coordinated manner and maintaining a normal monitoring status.

[0048] Preferably, S4 includes:

[0049] When determining that a wind turbine has an energy response imbalance, the confidence index Ltr and the overall average resonance intensity Sgz are compared. avgA comprehensive health index, Ecj, is constructed by fitting the data. This index comprehensively considers the differences in signal resonance and signal reliability, reflecting the overall coordination level of multi-domain data of the wind turbine and indicating the health stability of the wind turbine in the current cycle.

[0050]

[0051] Where i is the channel index, N represents the total number of data acquisition channels, and ε is a minimum value used to prevent the denominator from being zero.

[0052] Preferably, S4 further includes:

[0053] The mean value of the comprehensive health index Ecj when the wind turbine is operating normally in history is calculated by statistical method. Based on the mean value, a health diagnosis threshold Jz is preset, and the health status is assessed by comparing it with the real-time acquired comprehensive health index Ecj.

[0054] When the comprehensive health index Ecj < health diagnosis threshold Jz, it indicates that the wind turbine is in an abnormal state. At this time, fault information is generated and transmitted to the operation and maintenance personnel for fault investigation.

[0055] When the comprehensive health index Ecj is greater than or equal to the health diagnosis threshold Jz, it indicates that the wind turbine is operating normally. At this time, the wind turbine should be kept running normally and monitored regularly.

[0056] The new energy power generation equipment monitoring system of the present invention includes:

[0057] The multi-source data acquisition module is used to collect status signal data in real time through a sensor array deployed on the wind turbine. After preprocessing the status signal data, it sequentially performs aggregation and calibration processing to form a single-cycle data block X(t). k ); and for the single-cycle data block X(t) k Statistical analysis was performed to obtain the standardized signal Z. i,o (t k The standardized signal Z is arranged in time series order. i,o (t k Assembled into a multidimensional data matrix JZ;

[0058] The credibility analysis module is used to group the multidimensional data matrix JZ by category to obtain several data sequence groups, and calculate the probability distribution p of each parameter in each data sequence group. o (Tw), and based on the probability distribution p o (Tw) Calculate the feature entropy H for each data sequence group. s (t), based on the characteristic entropy H s (t) Perform reliability analysis on each data sequence group to determine whether the data source is in a stable state;

[0059] The energy response identification module is used to identify the energy response based on the standardized signal Z when the data source is in a stable state. i,o (t k The signal coordination analysis of different parameters of the wind turbine is performed, and the results of the signal coordination analysis are summed to carry out the average energy coordination analysis to determine whether there is an energy response imbalance in the wind turbine.

[0060] The health assessment module is used to fit the reliability analysis results with the average energy synergy analysis results when there is an energy response imbalance in the wind turbine. Based on the fitting results, the module performs a health status analysis on the overall operating status of the wind turbine and generates a health status assessment report for the wind turbine based on the health status analysis results.

[0061] Beneficial Effects: Compared with existing technologies, this invention has the following significant advantages: 1. By fusing and analyzing multi-source heterogeneous data, this invention constructs a unified multi-dimensional data matrix, achieving comprehensive and multi-dimensional precise monitoring of the operating status of new energy power generation equipment, laying a solid data foundation for timely detection and early warning of early faults; 2. By calculating the characteristic entropy of data sequence groups and performing credibility analysis, this invention can effectively identify and judge the stability and reliability of the data source itself, ensuring the data quality on which subsequent analysis is based, and guaranteeing the accuracy of monitoring results from the source; 3. Through signal collaborative analysis and time-series resonance index, this invention quantitatively assesses the dynamic energy coordination relationship between different parameters inside the equipment, and can keenly capture abnormal states such as energy response imbalance, achieving early identification and early warning of potential faults; 4. By fusing data credibility and system synergy analysis results, this invention constructs a comprehensive health index to assess the overall status of the equipment, generating an intuitive health status report, providing a scientific basis for operation and maintenance decisions, and effectively improving the operational reliability and maintenance efficiency of the equipment. Attached Figure Description

[0062] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0063] Figure 2 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0064] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0065] like Figure 1 As shown, this embodiment of the invention provides a method for monitoring new energy power generation equipment based on multi-source heterogeneous data, including the following steps:

[0066] S1. Status signal data is collected in real time by a sensor array deployed on the wind turbine. After data preprocessing, the data is aggregated and calibrated to form a single-cycle data block X(t).k Then, statistical analysis is performed to obtain the standardized signal Z. i,o (t k Then, the data is assembled into a multidimensional data matrix JZ according to the time series order.

[0067] S2. Group the multidimensional data matrix JZ by category and obtain the probability distribution p of each parameter. o (Tw) Calculate the feature entropy H for each data sequence group. s (t), and based on the feature entropy H s (t) Perform a reliability analysis on each data sequence group and trigger S3 when the data source is in a stable state;

[0068] S3, based on standardized signal Z i,o (t k Perform signal coordination analysis on signals with different parameters, and perform average energy coordination analysis by summing the results of the signal coordination analysis. Trigger S4 when there is an energy response imbalance in the wind turbine.

[0069] S4. Fit the reliability analysis results with the average energy synergy analysis results to perform a health status analysis of the overall operation of the wind turbine, and generate a health status assessment based on the analysis results.

[0070] In this embodiment, S1 collects the status signal data of the wind turbine in real time through sensor groups deployed at various key parts of the wind turbine, and forms a single-cycle data block X(t) after data preprocessing. k The data is then standardized to integrate the signal data from various sensors into a multi-dimensional data matrix JZ. This improves the accuracy and consistency of the data acquisition process, ensuring the synergistic effect between different signal sources during subsequent analysis. Compared with traditional single-sensor data monitoring methods, this invention achieves synchronous processing of diverse signals. S2 groups the multi-dimensional data matrix JZ by category, normalizes and statistically analyzes the instantaneous amplitude of each signal group, and then obtains the probability distribution p of each parameter using statistical methods. o (Tw), based on probability distribution p o (Tw) and calculate the characteristic entropy H of each data sequence group using information entropy theory. s(t) Perform credibility analysis on each data sequence group. S3 When the data source is in a stable state, automatically trigger signal collaborative analysis. This method can identify abnormal fluctuations in the signal and determine whether there are abnormal disturbances in the data source through credibility analysis, ensuring early detection of potential problems during wind turbine operation and reflecting the stability and reliability of wind turbine operation. S4 Fit the credibility analysis results with the average energy collaborative analysis results to perform a health status analysis on the overall operation of the wind turbine, triggering health status analysis in a timely manner and generating evaluation results. This step not only enhances the adaptability of the wind turbine in complex operating environments but also improves the scheduling efficiency and energy utilization rate of the wind turbine through energy collaborative analysis. Compared with existing monitoring methods that rely solely on single mechanical or electrical parameters, this solution improves the comprehensive performance and fault early warning capability of the wind turbine through multi-dimensional collaborative analysis. This solution improves the monitoring method of wind turbines, enhances the efficiency of data acquisition and processing, and optimizes the health assessment mechanism through intelligent fusion and precise analysis of multi-source heterogeneous data.

[0071] Specifically, S1 includes:

[0072] S11. Real-time status signal data of the wind turbine is collected by sensor groups deployed in the nacelle, tower and environmental measuring points of the wind turbine.

[0073] The sensor group includes strain gauges, accelerometers, speed encoders, current sensors, power meters, stator temperature sensors, anemometers, barometers, and humidity sensors.

[0074] By installing strain gauges, acceleration sensors, and speed encoders inside the wind turbine nacelle, the tower strain ε, vibration acceleration a, and main shaft speed ω are collected in real time.

[0075] By installing current sensors, power meters, and stator temperature sensors in the generator and electrical cabinet area, the current I, output power P, and stator temperature Ts are collected in real time.

[0076] By installing an anemometer, barometer, and humidity sensor at the wind vane on the top of the wind turbine tower, wind speed V, air density ρa, and air humidity H are collected in real time.

[0077] S12. Perform data preprocessing on the status signal data through edge analysis nodes;

[0078] The data preprocessing includes noise filtering and time base alignment;

[0079] The noise filtering is used to remove high-frequency noise components from continuously fluctuating state signal data using a low-pass filtering algorithm; the time reference alignment is used to periodically synchronize the state signal data of different channels with the acquisition timestamps of the sensor group through the Network Time Protocol (NTP) so that the state signal data of different channels can be synchronized at the same time step.

[0080] After data preprocessing, the state signal data is periodically packaged. Based on a unified sampling period ΔT, the state signal data in all partitions are aggregated and calibrated according to the sensor acquisition timestamps, and combined into a single-period data block X(t) according to the data source order. k ), X(t) k )={X i (t k Let X = |i=1,2,...,n}, where X represents the set of multi-source heterogeneous state signal data of the wind turbine in each sampling period. Essentially, it is a multi-dimensional state signal data vector synchronously collected and preprocessed by different types of sensors. k () represents a set of heterogeneous signal vectors with a unified timestamp, where each dimension represents a sensor type, i is the channel index (i∈[1,n]), n is the total number of sampling parameters, and t k For the k-th sampling period, X i (t k ) indicates that the i-th sensor channel is in t k Parameters collected at all times;

[0081] S13. For each parameter sequence X i (t k Perform statistical analysis to obtain the sequence X of each parameter. i (t k The historical mean μ i With standard deviation σ i Then, standardization is performed to obtain the parameter sequence X. i (t k The standardized signal Z) i,o (t k ):

[0082]

[0083] Among them, Z i,o (t k ) represents the o-th parameter in the i-th sensing channel at t k The standardized signal at time t is used, with 'o' as the parameter index. After standardization, the standardized parameters are assembled into a unified multidimensional data matrix JZ in time series order, where JZ = [ε, a, ω, I, P, Ts, V, ρa, H]. T, where T represents the matrix transpose symbol, used to mathematically organize multiple parameters into a vertically arranged column vector.

[0084] In this embodiment, S11 involves deploying multiple sensor groups at the nacelle, tower, and environmental measuring points of the wind turbine to collect different types of status signal data in real time, including tower strain ε, vibration acceleration a, main shaft speed ω, current I, output power P, stator temperature Ts, wind speed V, air density ρa, and air humidity H. S12 performs noise filtering and time base alignment on the status signal data using edge analysis nodes to ensure that the signals from different sensors are synchronously aligned at a unified time step. The pre-processed data is then aggregated into a single-cycle data block X(t). k S13 for each parameter sequence X i (t k ) Perform statistical analysis to obtain the standardized signal Z for each parameter. i,o (t k This data is then further assembled into a multidimensional data matrix JZ. This step enables the efficient fusion of data from different types of sensors, improving data reliability and comparability, and providing accurate data support for subsequent analysis and health status assessment. This not only improves the accuracy and real-time performance of data processing but also enhances the comprehensiveness and accuracy of wind turbine condition monitoring, as well as the operating efficiency and maintenance management level of wind turbines.

[0085] Specifically, S2 includes:

[0086] S21. Within the set sliding time window Tw, the multidimensional data matrix JZ is grouped by category to construct data sequence groups for mechanical signals, electrical signals, and environmental signals, respectively. After normalizing and statistically analyzing the instantaneous amplitude of each signal group, the probability distribution p of each parameter is obtained through statistical methods. o (Tw) represents the probability distribution of the 0th parameter within the current time window Tw;

[0087] S22, Based on probability distribution p o (Tw) and calculate the characteristic entropy H of each data sequence group using information entropy theory. s (t):

[0088]

[0089] Where n represents the number of parameters in the current data sequence group, and ln represents the logarithmic function.

[0090] S23. Calculate the average entropy value H of the wind turbine under historical stable operating conditions using statistical methods. ref Based on the characteristic entropy H of each data sequence group, s(t) is fitted, credibility analysis is performed on each data sequence group, and the credibility index Ltr of the data sequence group is calculated:

[0091]

[0092] When the credibility index Ltr < 1, it indicates that there is an abnormal disturbance trend in the data source of the current data sequence group. At this time, a feedback signal is sent to the data acquisition layer, and a resampling process is performed on the data acquisition channel that triggered the feedback signal.

[0093] When the confidence index Ltr≥1, it indicates that the data source of the current data sequence group is in a stable state, and signal collaborative analysis is triggered at this time.

[0094] In this embodiment, S21 constructs data sequence groups for mechanical signals, electrical signals, and environmental signals by grouping the multidimensional data matrix JZ by category and performing sliding time window analysis. After normalizing and statistically analyzing the instantaneous amplitude of each signal group, the probability distribution p of each parameter is obtained through statistical methods. o (Tw), S22 is based on probability distribution p o (Tw) and calculate the characteristic entropy H of each data sequence group using information entropy theory. s (t). This step utilizes information entropy theory to conduct an in-depth analysis of each data sequence group, evaluating its stability and reliability. This formula is based on the classical entropy formula. The derivation yields the following: P(t) represents the probability of each event, reflecting its likelihood of occurrence. The negative sign in the formula ensures that the entropy is positive. In this scheme, considering the variation characteristics of each signal source within different time windows, the probability distribution p of each parameter is first calculated using statistical methods. o (Tw), and then through the probability distribution p o (Tw) Logarithmic calculations are performed to quantify the unpredictability of each signal. Finally, the contributions of all signals are summed to obtain the characteristic entropy H of each data sequence group. s (t). S23 is compared with the average entropy value H under historical stable operating conditions. ref The method involves fitting data and calculating a reliability index (Ltr) to identify abnormal disturbance trends in the data source. If the reliability index Ltr < 1, a feedback signal is triggered and the data acquisition channel is resampled to ensure data accuracy. When the reliability index Ltr ≥ 1, the system enters a stable state and triggers subsequent signal collaborative analysis. This method, through precise anomaly identification and real-time feedback mechanisms, not only improves data reliability but also enhances the health monitoring and fault early warning capabilities of wind turbines. This effectively improves the operational stability and maintenance efficiency of wind turbines, providing strong technical support for the intelligent operation and maintenance of wind turbines.

[0095] Specifically, S3 includes:

[0096] S31. When the data source is in a stable state, based on the standardized signal Z i，o (t k Signal co-synchronization analysis is performed on the energy synchronization degree of signals with different parameters in the time domain, and the temporal resonance index Sgz between heterogeneous signals is calculated to characterize the dynamic similarity and energy resonance intensity between heterogeneous signals in the time domain.

[0097]

[0098] Among them, Sgz o,q Z represents the degree of energy synchronization between the standardized signals of the 0th and qth parameters in the time domain, τ represents the time variable, Δτ represents the compensation delay, used to eliminate the time offset introduced by the sampling time difference and response lag of each data source, and Z o (τ) and Z q (τ) represents the standardized signals of the 0th and qth parameters at time τ, respectively, and dτ represents the time integral variable.

[0099] S32. Based on the temporal resonance index Sgz of all data sources, perform ergodic summation, conduct average energy coherence analysis on all signals, and construct the overall average resonance intensity Sgz. avg This represents the average energy coherence among all signal pairs:

[0100]

[0101] Where M represents the number of valid data sources participating in heterogeneous resonance analysis within the current calculation cycle;

[0102] When the average resonance intensity Sgz avg A value <0.85 indicates an imbalance in energy response and a drive control anomaly in the wind turbine, triggering a health status analysis.

[0103] When the average resonance intensity Sgz avg A value of ≥0.85 indicates that the wind turbine unit is operating in a coordinated manner, and normal monitoring should be maintained at this time.

[0104] In this embodiment, when the data source is in a stable state, step S31 performs a collaborative analysis of the energy synchronization degree of signals with different parameters based on the standardized signal, and calculates the temporal resonance index Sgz between heterogeneous signals to characterize the dynamic similarity and energy resonance intensity between the signals. In the field of signal processing, when analyzing the temporal relationship between two or more signals, a cross-correlation function is used. This method measures the correlation between signals at different time points, where x(t) and y(t) are two signals, τ is the time delay, and the cross-correlation function Rxy is used to measure the similarity between signals x(t) and y(t) at time delay τ. In this scheme, the temporal resonance index Sgz is derived using signal co-analysis, which measures their dynamic similarity and energy resonance intensity by calculating the degree of resonance between signals. This concept is based on an extension of the cross-correlation function, and through signal normalization and energy synchronization analysis, the interaction between signals can be more accurately measured in the time domain. o (τ) and Z q The inner product of (τ) in the time domain represents their degree of temporal synchronization, and the formula is: Where Δτ is the compensation delay, used to handle time delay differences caused by different signal sources; this is achieved by integrating the energy of each signal. and The signal is normalized; the above is then fitted to construct the temporal resonance index Sgz. S32 constructs the overall average resonance intensity Sgz by iteratively summing the temporal resonance indices Sgz of all signals. avg This enables energy coordination analysis among various data sources of the wind turbine. When the average resonance intensity Sgz... avg When the average resonance intensity Sgz is below 0.85, the system can promptly identify energy response imbalances in wind turbines, thereby triggering a health status analysis; while when the average resonance intensity Sgz is below 0.85, the system can promptly identify energy response imbalances in wind turbines, thereby triggering a health status analysis; avg A value greater than or equal to 0.85 indicates that the wind turbine's signals are operating normally and maintaining a normal monitoring status. This formula first sums the time-series resonance indices between all different signal pairs. To ensure the results are unaffected by the number of signals M, a normalization factor is used. The sum is standardized to obtain the average temporal resonance exponent Sgz. avg This method improves the stability and reliability of wind turbine operation through signal collaborative analysis and energy resonance monitoring. It can identify potential anomalies at an early stage and provide real-time warnings, thereby improving the health management capabilities and operation and maintenance efficiency of wind turbines.

[0105] Specifically, S4 includes:

[0106] S41. When there is an energy response imbalance in the wind turbine, the confidence index Ltr and the average resonance intensity Sgz are compared. avg A fitting process is performed to analyze the overall health status of the wind turbine's operation and construct a comprehensive health index Ecj. This index comprehensively considers the differences in signal resonance and signal reliability, reflecting the overall coordination level of the wind turbine's multi-domain data and indicating the health stability of the wind turbine in the current cycle.

[0107]

[0108] Where i is the channel index, N represents the total number of data acquisition channels, and ε represents a minimum value to prevent the denominator from being zero, with a value of 10. -6 .

[0109] S42. Calculate the mean value of the comprehensive health index Ecj when the wind turbine is operating normally in history using statistical methods, and preset the health diagnosis threshold Jz based on the mean value. Then, evaluate the health status with the real-time acquired comprehensive health index Ecj. The specific evaluation scheme is as follows.

[0110] When the comprehensive health index Ecj < health diagnosis threshold Jz, it indicates that the wind turbine is in an abnormal state. At this time, fault information is generated and transmitted to the operation and maintenance personnel for fault investigation.

[0111] When the comprehensive health index Ecj is greater than or equal to the health diagnosis threshold Jz, it indicates that the wind turbine is operating normally. At this time, the wind turbine should be kept running normally and monitored regularly.

[0112] In this embodiment, S41 combines the confidence index Ltr with the average resonance intensity Sgz. avg A fitting process is performed to comprehensively evaluate the operating status of the wind turbine and generate a comprehensive health index Ecj, reflecting the difference in signal resonance and signal reliability, and measuring the overall coordination and health stability of the wind turbine. The denominator of this formula originates from the deviation measurement model in information theory. In Shannon's information entropy theory, the uncertainty of the system state can be quantified by the distance of the probability distribution from the ideal state. Its basic form is:

[0113]

[0114] Where, p iLet p0 be the current probability and p0 be the reference probability. This scheme transforms this idea into a relative deviation measure of credibility. Specifically, Ltr(i) represents the credibility index of the i-th channel, with the ideal credibility state set to 1. Then, |1−Ltr(i)| represents the deviation of the channel from the ideal credibility state. The overall signal credibility deviation is obtained by summing these values. To achieve unified evaluation across different physical quantity dimensions, this application adopts a fractional form, merging energy synergy and credibility deviation as a ratio. By normalizing the denominator, the relative stability function of the overall health level is obtained, ultimately forming the comprehensive health index Ecj. This form originates from the classic "synergy / deviation" ratio model, a commonly used composite evaluation function structure in stability analysis. S42 calculates the mean of the comprehensive health index Ecj under normal operating conditions through statistical analysis of historical operating data, and sets a health diagnosis threshold Jz based on this mean, comparing it with the real-time acquired health index. When the comprehensive health index Ecj falls below the health diagnosis threshold Jz, a fault warning is immediately issued and a fault investigation process is triggered; when the comprehensive health index Ecj is greater than or equal to the health diagnosis threshold Jz, the unit continues to operate normally. This method not only improves the health management capabilities of wind turbine units but also enables rapid response to abnormal conditions, thereby enhancing the operational stability and maintenance efficiency of wind turbine units.

[0115] like Figure 2 As shown, this embodiment of the invention also provides a monitoring system for new energy power generation equipment based on multi-source heterogeneous data, including a multi-source data acquisition module, a reliability analysis module, an energy response identification module, and a health assessment module;

[0116] The multi-source data acquisition module is used to collect status signal data in real time based on the sensor group deployed on the wind turbine, and after data preprocessing, it aggregates and calibrates the data to form a single-cycle data block X(t). k Then, statistical analysis is performed to obtain the standardized signal Z. i，o (t k Then, the data is assembled into a multidimensional data matrix JZ according to the time series order.

[0117] The credibility analysis module is used to group the multidimensional data matrix JZ by category and obtain the probability distribution p of each parameter. o (Tw) Calculate the feature entropy H for each data sequence group. s (t), and based on the feature entropy H s (t) Perform credibility analysis on each data sequence group and trigger the energy response identification module when the data source is in a stable state;

[0118] The energy response identification module is used to identify the energy response based on the standardized signal Z. i，o (t kThe system performs signal coordination analysis on signals with different parameters, and performs average energy coordination analysis by summing the results of the signal coordination analysis. When there is an energy response imbalance in the wind turbine, the health assessment module is triggered.

[0119] The health assessment module is used to fit the reliability analysis results with the average energy synergy analysis results, perform a health status analysis on the overall operation of the wind turbine, and generate a health status assessment based on the analysis results.

Claims

1. A monitoring method for new energy power generation equipment based on multi-source heterogeneous data, characterized in that, Includes the following steps: S1. Status signal data is collected in real time by a sensor array deployed on the wind turbine. After preprocessing the status signal data, aggregation and calibration processes are performed sequentially to form a single-cycle data block X(t). k ); for the single-cycle data block X(t) k Statistical analysis was performed to obtain the standardized signal Z. i,o (t k The standardized signal Z is then processed in chronological order. i,o (t k Assembled into a multidimensional data matrix JZ; S2. Group the multidimensional data matrix JZ by category to obtain several data sequence groups; calculate the probability distribution p of each parameter in each data sequence group. o (Tw), and based on the probability distribution p o (Tw) Calculate the feature entropy H for each data sequence group. s (t); based on the characteristic entropy H s (t) Perform a reliability analysis on each data sequence group to determine whether the data source is in a stable state. If the data source is in a stable state, proceed to the next step. S3, Based on the standardized signal Z i,o (t k Perform signal coordination analysis on different parameter signals of the wind turbine; sum the results of the signal coordination analysis, conduct average energy coordination analysis, and determine whether there is energy response imbalance in the wind turbine. If there is energy response imbalance in the wind turbine, proceed to the next step. S4. Fit the reliability analysis results with the average energy synergy analysis results; Based on the fitting results, a health status analysis of the overall operation of the wind turbine is performed, and a health status assessment report of the wind turbine is generated based on the health status analysis results.

2. The method according to claim 1, characterized in that, The acquired status signal data mentioned in step S1 includes: The status signal data of the wind turbine is collected in real time by sensor groups deployed in the nacelle, tower and environmental measuring points of the wind turbine. The sensor group includes strain gauges, accelerometers, speed encoders, current sensors, power meters, stator temperature sensors, anemometers, barometers, and humidity sensors. By installing strain gauges, acceleration sensors, and speed encoders inside the wind turbine nacelle, the tower strain ε, vibration acceleration a, and main shaft speed ω are collected in real time. By installing current sensors, power meters, and stator temperature sensors in the generator and electrical cabinet area, the current I, output power P, and stator temperature Ts are collected in real time. By installing an anemometer, barometer, and humidity sensor at the wind vane on the top of the wind turbine tower, wind speed V, air density ρa, and air humidity H are collected in real time.

3. The method according to claim 1, characterized in that, Step S1 further includes: The state signal data is preprocessed using edge analysis nodes. The data preprocessing includes: Noise filtering: Apply a low-pass filtering algorithm to continuously fluctuating state signal data to remove high-frequency noise components; Time reference alignment: Periodic time synchronization is performed between the network time protocol and the data acquisition timestamps of the sensor group, so that the status signal data of different channels can be synchronized and aligned at the same time step; After data preprocessing, the state signal data is periodically packaged. Based on a unified sampling period ΔT, the state signal data in all partitions are aggregated and calibrated according to the sensor's acquisition timestamp sequence to form a single-period data block X(t). k ); Where X(t) k )={X i (t k Let X represent the set of multi-source heterogeneous state signal data of the wind turbine generator in each sampling period, where i = 1, 2, ..., n. k () represents a set of heterogeneous signal vectors with a unified timestamp, where each dimension represents a sensor type, i is the channel index (i∈[1,n]), n is the total number of sampling parameters, and t k For the k-th sampling period, X i (t k ) indicates that the i-th sensor channel is in t k Parameters collected at all times; For each parameter sequence X i (t k Perform statistical analysis to obtain the sequence X of each parameter. i (t k The historical mean μ i With standard deviation σ i The standardized signal Z is obtained by performing standardization processing. i,o (t k ): Among them, Z i,o (t k ) represents the o-th parameter in the i-th sensing channel at t k The standardized signal at time o is the parameter index; The standardized parameters are assembled into a unified multidimensional data matrix JZ according to the time series order, JZ=[ε, a, ω, I, P, Ts, V, ρa, H] T , where T represents the matrix transpose.

4. The method according to claim 1, characterized in that, Step S2 includes: Within a set sliding time window Tw, the multidimensional data matrix JZ is grouped by category to construct data sequence groups for mechanical signals, electrical signals, and environmental signals, respectively. Normalize and statistically analyze the instantaneous amplitudes of each parameter in each data sequence group to obtain the probability distribution p of each parameter occurring within the current time window Tw. o (Tw) represents the probability distribution of the o-th parameter within the current time window Tw, where o is the parameter index; Based on the probability distribution p o (Tw), the characteristic entropy H of each data sequence group is calculated using information entropy theory. s (t), the calculation formula is: Where n represents the number of parameters in the current data sequence group, and ln represents the natural logarithm function.

5. The method according to claim 1, characterized in that, Step S2 further includes: The average entropy value H of the wind turbine under historical stable operating conditions ref Based on the characteristic entropy H of each data sequence group, s (t) is fitted, and the reliability index Ltr of each data sequence group is calculated: When the confidence index Ltr < 1, it is determined that the data source of the current data sequence group has an abnormal disturbance trend, a feedback signal is sent to the data acquisition layer, and a resampling process is performed on the data acquisition channel that triggered the feedback signal. When the confidence index Ltr≥1, it is determined that the data source of the current data sequence group is in a stable state, triggering signal collaborative analysis.

6. The method according to claim 1, characterized in that, Step S3 includes: When the data source is in a stable state, based on the standardized signal Z i,o (t k Signal coordination analysis is performed on the energy synchronization degree of signals with different parameters in the time domain. The temporal resonance index Sgz between heterogeneous signals is calculated to characterize the dynamic similarity and energy resonance intensity between them in the time domain. The calculation formula is as follows: Among them, Sgz o,q Z represents the degree of energy synchronization between the normalized signals of the 0th and qth parameters in the time domain, where τ represents the time variable, Δτ represents the compensation delay, and Z represents the time difference. o (τ) and Z q (τ) represents the standardized signals of the 0th and qth parameters at time τ, respectively, and dτ represents the time integral variable.

7. The method according to claim 1, characterized in that, Step S3 further includes: The overall average resonance intensity Sgz is calculated by iterating and summing the temporal resonance indices Sgz among all heterogeneous signals. avg This represents the average energy coherence among all signal pairs: Where M represents the number of valid data sources participating in heterogeneous resonance analysis within the current calculation cycle; The overall average resonance intensity Sgz avg Compare with a preset synergy threshold; when Sgz avg When the value is below the aforementioned synergy threshold, the wind turbine is determined to have energy response imbalance and drive control abnormality, triggering a health status analysis; when Sgz avg When the overall operation of the wind turbine is not lower than the aforementioned coordination threshold, it is determined that the wind turbine is operating in a coordinated manner and maintaining a normal monitoring status.

8. The method according to claim 1, characterized in that, S4 includes: When determining that a wind turbine has an energy response imbalance, the confidence index Ltr and the overall average resonance intensity Sgz are compared. avg A comprehensive health index, Ecj, is constructed by fitting the data. This index comprehensively considers the differences in signal resonance and signal reliability, reflecting the overall coordination level of multi-domain data of the wind turbine and indicating the health stability of the wind turbine in the current cycle. Where i is the channel index, N represents the total number of data acquisition channels, and ε is a minimum value used to prevent the denominator from being zero.

9. The method according to claim 1, characterized in that, S4 further includes: The mean value of the comprehensive health index Ecj when the wind turbine is operating normally in history is calculated by statistical method. Based on the mean value, a health diagnosis threshold Jz is preset, and the health status is assessed by comparing it with the real-time acquired comprehensive health index Ecj. When the comprehensive health index Ecj < health diagnosis threshold Jz, it indicates that the wind turbine is in an abnormal state. At this time, fault information is generated and transmitted to the operation and maintenance personnel for fault investigation. When the comprehensive health index Ecj is greater than or equal to the health diagnosis threshold Jz, it indicates that the wind turbine is operating normally. At this time, the wind turbine should be kept running normally and monitored regularly.

10. A monitoring system for new energy power generation equipment based on multi-source heterogeneous data, characterized in that, include: The multi-source data acquisition module is used to collect status signal data in real time through a sensor array deployed on the wind turbine. After preprocessing the status signal data, it sequentially performs aggregation and calibration processing to form a single-cycle data block X(t). k ); and for the single-cycle data block X(t) k Statistical analysis was performed to obtain the standardized signal Z. i,o (t k The standardized signal Z is arranged in time series order. i,o (t k Assembled into a multidimensional data matrix JZ; The credibility analysis module is used to group the multidimensional data matrix JZ by category to obtain several data sequence groups, and calculate the probability distribution p of each parameter in each data sequence group. o (Tw), and based on the probability distribution p o (Tw) Calculate the feature entropy H for each data sequence group. s (t), based on the characteristic entropy H s (t) Perform reliability analysis on each data sequence group to determine whether the data source is in a stable state; The energy response identification module is used to identify the energy response based on the standardized signal Z when the data source is in a stable state. i,o (t k The signal coordination analysis of different parameters of the wind turbine is performed, and the results of the signal coordination analysis are summed to carry out the average energy coordination analysis to determine whether there is an energy response imbalance in the wind turbine. The health assessment module is used to fit the reliability analysis results with the average energy synergy analysis results when there is an energy response imbalance in the wind turbine. Based on the fitting results, the module performs a health status analysis on the overall operating status of the wind turbine and generates a health status assessment report for the wind turbine based on the health status analysis results.

Images