An online energy monitoring system and method

Abstract

The present application relates to the technical field of energy monitoring, and more particularly to an online energy monitoring system and method, wherein a multi-source data acquisition module realizes all-around and multi-level energy network monitoring through integrating various sensors, ensuring the comprehensiveness and accuracy of data; an adaptive acquisition control module establishes a risk assessment model using historical data, dynamically adjusts the acquisition frequency and parameter range of the sensors, and optimizes the data acquisition efficiency; a data preprocessing module improves data quality through filtering, denoising, alignment and standardization processing, and provides a reliable basis for subsequent analysis; an energy analysis module supports system optimization scheduling and energy-saving decision-making, quickly detects abnormal points, and reduces the cost of manual intervention; and an intelligent response module formulates a response strategy according to the analysis results, realizes self-adjustment and optimization of the system, and improves energy utilization efficiency and equipment operation reliability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy monitoring, and in particular to an online energy monitoring system and method. BACKGROUND

[0002] In the field of energy monitoring, traditional systems rely on fixed-frequency data acquisition and static threshold judgment, making it difficult to cope with complex and variable energy network environments. The existing technology has the following limitations:

[0003] Data acquisition is rigid: sensors usually work at a fixed frequency and cannot be dynamically adjusted according to real-time risks, resulting in insufficient data granularity in high-risk scenarios, while producing redundant data in low-risk situations, wasting storage and computing resources.

[0004] Risk perception lag: traditional risk assessment methods that rely on manually set thresholds lack the ability to dynamically model based on historical data, making it difficult to capture parameter trends, resulting in delayed warnings. This lag can lead to equipment failure or sudden energy efficiency, affecting system stability and efficiency.

[0005] Data processing is extensive: abnormal value filtering mostly relies on fixed threshold methods, which are sensitive to non-Gaussian distributed noise and prone to misjudgment. In addition, commonly used signal denoising techniques such as mean filtering can result in feature loss, affecting the accuracy of subsequent analysis.

[0006] Anomaly detection relies on prior knowledge: supervised anomaly detection requires a large amount of labeled data, while in industrial scenarios, abnormal samples are scarce, and algorithms such as SVM have high computational complexity, making it difficult to meet real-time requirements.

[0007] Shallow device health monitoring: traditional vibration analysis only focuses on time-domain statistics (such as RMS values), ignoring the mapping relationship between frequency-domain features and fault patterns, resulting in a high rate of early fault detection, increasing maintenance costs and downtime risks. SUMMARY

[0008] The present application provides an online energy monitoring system and method, which uses multiple sensors to capture data in real time and adopts a dynamic adjustment strategy to optimize acquisition efficiency, combines BP neural network, isolated forest algorithm, and vibration spectrum analysis technology to evaluate energy efficiency, identify abnormalities, and predict faults, and finally automatically develops response strategies based on analysis results to improve energy utilization efficiency and device operation reliability.

[0009] To achieve the purpose of the present application, the technical solution adopted is: an online energy monitoring system, comprising a multi-source data acquisition module, an adaptive acquisition control module, a data preprocessing module, an energy analysis module, and an intelligent response module;

[0010] The multi-source data acquisition module includes an electric energy sensor group, an electrical characteristic sensor, and a device state sensor, the electric energy sensor group is used for monitoring basic energy data, the electrical characteristic sensor is used for monitoring power grid quality state, and the device state sensor is used for monitoring device operation health degree;

[0011] The adaptive acquisition control module establishes a risk assessment model of a monitoring parameter by using historical data based on data provided by the multi-source data acquisition module, and dynamically adjusts the acquisition frequency and parameter range of the electric energy sensor group, the electrical characteristic sensor, and the device state sensor according to a change trend of the monitoring parameter;

[0012] The data preprocessing module receives original data from the data acquisition module, and performs filtering, denoising, and standardization processing;

[0013] The energy analysis module receives data preprocessed by the data preprocessing module, adopts a BP neural network to evaluate energy efficiency, an isolated forest algorithm to identify an abnormal mode, and a vibration spectrum analysis to predict device failure, and the energy analysis module transmits analysis results to the intelligent response module, and the intelligent response module formulates a corresponding response strategy according to the analysis results;

[0014] The adaptive acquisition control module includes establishment of a risk assessment model of a monitoring parameter, and specifically includes the following steps:

[0015] 1) Parameter risk level division;

[0016]

[0017] Wherein: Q k (p) is a parameter risk level division result; F -1 (p) is an inverse function of an experience distribution function, and represents a value corresponding to a probability p of a cumulative distribution function; and x is a current parameter value;

[0018] 2) Risk weight dynamic calculation;

[0019] ; wherein: H i =-∑p k log2p k ;

[0020] Wherein: w i is a risk weight of an i-th parameter; H i is an information entropy of the i-th parameter; and p k is an occurrence frequency of the i-th parameter in a k-th risk interval;

[0021] 3) The risk assessment model of the monitoring parameter is:

[0022] ;

[0023] wherein: R t Rt represents the comprehensive risk value at the current moment, r i (t) is the risk level of the i-th parameter at the current moment; λ is the trend sensitivity coefficient, and dR / dt is the change rate of the risk value.

[0024] As an optimization scheme of the present application, the electric energy sensor group includes an electric energy sensor and a power meter, which capture the power consumption and power information in real time; the electrical characteristic sensor includes a voltage sensor, a current sensor, and a power factor sensor, which monitor the stability and health condition of the power grid; the device state sensor includes a temperature sensor and a vibration sensor, which evaluate the working state of the device by monitoring the working temperature and vibration frequency of the key device.

[0025] As an optimization scheme of the present application, the core target of dynamically adjusting the acquisition frequency is to optimize the data acquisition efficiency according to Rt and the change trend dR / dt, which specifically includes:

[0026] A) generating the comprehensive risk value at the current moment based on the risk evaluation model;

[0027] B) capturing the change trend of the risk by using the sliding window mechanism and the time decay factor;

[0028] C) dynamically adjusting the acquisition frequency according to the comprehensive risk value and the change trend;

[0029] The dynamic boundary calculation adjusts the upper and lower limit ranges of the parameters according to the risk value, further optimizing the acquisition efficiency;

[0030] As an optimization scheme of the present application, the acquisition frequency adjustment strategy is formulated according to the comprehensive risk value Rt and the change trend dR / dt:

[0031] 1) the high-risk region is R t >R high or dR / dt>dR threshold :

[0032] The acquisition frequency is increased to monitor the high-risk parameters in real time; the adjustment formula is: f new =f base ×k high； f new is the new acquisition frequency, f base is the basic acquisition frequency, k high is the high-risk adjustment coefficient, R high is the high-risk threshold, dR threshold is the threshold of the risk change rate;

[0033] 2) the medium-risk region is R low ≤R t ≤R highand dR / dt≤dR threshold , maintain the basic acquisition frequency, f new = f base ;

[0034] 3) low-risk area R t <R low and dR / dt≤dR threshold , reduce the acquisition frequency to save resources, adjust the formula: f new = f base ×k low , k low is a low-risk adjustment coefficient, R low is a low-risk threshold.

[0035] As an optimization scheme of the present application, the data preprocessing module processes include:

[0036] Data alignment solves the time deviation problem of multi-source data through time synchronization and resampling;

[0037] Outlier processing uses an improved Hampel filter to remove outliers and fills in missing values through an ARIMA model;

[0038] Signal denoising uses wavelet threshold denoising technology to remove high-frequency noise and retain the main features of the signal;

[0039] Feature standardization converts data to a uniform scale through dynamic Z-score standardization.

[0040] As an optimization scheme of the present application, the energy analysis module receives data preprocessed from the data preprocessing module, and the specific steps for evaluating energy efficiency using a BP neural network are:

[0041] 1) Prepare the training data set; the data preprocessed by the data preprocessing module constitutes the training data set;

[0042] 2) Input layer: n-dimensional characteristic electric energy, power factor and temperature;

[0043] Hidden layer: uses hyperbolic tangent activation function;

[0044] Output layer: Sigmoid function outputs energy efficiency score;

[0045] Train the BP neural network model: adjust the weights through the back propagation algorithm to minimize the error.

[0046] As an optimization scheme of the present application, the energy analysis module transmits the analysis result to the intelligent response module, and the intelligent response module formulates a corresponding response strategy according to the analysis result, specifically including:

[0047] S71, receive and analyze the analysis result;

[0048] Input: Energy efficiency score, anomaly score, and device health status from the energy analysis module;

[0049] Energy efficiency score: output by BP neural network, ranging from (0, 1];

[0050] Anomaly score: calculated by Isolation Forest algorithm, reflecting whether the data point is abnormal;

[0051] Device health status: based on vibration spectrum analysis, assesses whether the device is at risk of failure;

[0052] S72, formulating response strategies; according to the results of analysis, the intelligent response module needs to formulate specific response strategies; including energy efficiency score response strategy, anomaly score response strategy and device health status response strategy.

[0053] As an optimization scheme of the present application, formulating response strategies; according to the results of analysis, the intelligent response module needs to formulate specific response strategies;

[0054] Energy efficiency score response strategy; condition: when energy efficiency score S efficiency <0.5, indicating low energy efficiency; strategies include reducing the running time or load of high-energy-consuming devices, increasing the workload of low-energy-consuming devices; adjusting the device running schedule to take advantage of the peak-valley electricity price difference to reduce costs;

[0055] Anomaly score response strategy; condition: when anomaly score S anomaly ≥0.6, indicating clear anomaly; when 0.5≤S anomaly <0.6, indicating suspected anomaly; clear anomaly triggers an alarm: immediately notify relevant personnel for manual verification, and take appropriate measures according to the anomaly type; suspected anomaly records logs: records the abnormal situation in the system log for subsequent analysis;

[0056] 3) Device health status response strategy; when the device health status shows potential risk of failure, perform advance maintenance, arrange preventive maintenance of the device according to the results of vibration spectrum analysis to avoid sudden failure.

[0057] In order to achieve the purpose of the present application, the technical scheme adopted is: a monitoring method of an online energy monitoring system, comprising: multi-source data acquisition, adaptive acquisition control, data preprocessing, energy analysis and intelligent response;

[0058] The multi-source data acquisition includes an electric energy sensor group, an electrical characteristic sensor and a device state sensor, the electric energy sensor group is used for monitoring basic energy data, the electrical characteristic sensor is used for monitoring power grid quality status, and the device state sensor is used for monitoring device running health degree;

[0059] The adaptive acquisition control is based on the data provided by the multi-source data acquisition module, a risk assessment model of the monitoring parameter is established by using historical data, and the acquisition frequency and parameter range of the electric energy sensor group, the electrical characteristic sensor and the equipment state sensor are dynamically adjusted according to the change trend of the monitoring parameter;

[0060] The data preprocessing receives the original data from the data acquisition module, and performs filtering, denoising and standardization processing;

[0061] The energy analysis receives the data preprocessed from the data preprocessing module, adopts a BP neural network to evaluate energy efficiency, an isolated forest algorithm to identify abnormal patterns, and vibration spectrum analysis to predict equipment failure, and the energy analysis transmits the analysis results to the intelligent response module, and the intelligent response module formulates a corresponding response strategy according to the analysis results.

[0062] The present application has positive effects:

[0063] 1) The present application improves data acquisition efficiency and accuracy; the multi-source data acquisition integrates various types of sensors (such as the electric energy sensor group, the electrical characteristic sensor and the equipment state sensor), and realizes multi-level and all-around data acquisition of the energy network. This comprehensive data acquisition method not only provides rich basic information, but also lays a solid foundation for subsequent in-depth analysis. The adaptive acquisition control: based on the risk assessment model established by historical data analysis, the acquisition frequency and parameter range of the sensor can be dynamically adjusted according to the current risk value and change trend. It ensures that key data can be obtained in real time in high-risk situations, and resources are effectively saved in low-risk conditions.

[0064] 2) The present application enhances the intelligent level of the system: sliding window mechanism and trend sensitive mechanism: these mechanisms are used to capture the change trend of risk and early warn potential problems. They not only improve the response speed of the system, but also enhance the prediction ability, so that the system can take preventive measures before the fault occurs. BP neural network evaluates energy efficiency: by nonlinear modeling of multidimensional features, the BP neural network can accurately evaluate the energy efficiency level of the system or device, and provide optimization scheduling suggestions. This helps to identify low-efficiency operation modes, and thus improves the overall energy utilization efficiency. Isolated forest algorithm identifies abnormal patterns: without labeled data, abnormal detection can be completed, the cost of manual intervention is reduced, and abnormal points can be quickly identified to avoid major losses.

[0065] 3) The present application optimizes data quality and reduces operation cost: data preprocessing: including time synchronization, outlier processing, signal denoising and feature standardization, etc., which significantly improves the data quality. High-quality data provides reliable guarantee for subsequent analysis, and reduces the misjudgment caused by data quality problems.

[0066] Dynamic boundary calculation: dynamically adjust the upper and lower limit range of parameters according to risk value, further optimize the data acquisition efficiency, avoid the collection of redundant or invalid data, and reduce the operation cost.

[0067] 4) The application improves the reliability of equipment operation and prolongs the service life. The FFT is applied to convert the vibration signal into a frequency domain signal, accurately diagnose whether the equipment has a specific type of fault (such as imbalance, misalignment, etc.), which makes the maintenance work more targeted, can arrange preventive maintenance in advance, reduces the probability of sudden failure, and prolongs the service life of the equipment. According to the analysis results, the response strategy is automatically formulated, covering energy efficiency score response strategy, abnormal score response strategy and equipment health state response strategy, etc., finally realizing self-adjustment and optimization of the system. BRIEF DESCRIPTION OF DRAWINGS

[0068] The application will be further described in detail below in combination with the drawings and specific embodiments.

[0069] Figure 1 is the principle diagram of the application. DETAILED DESCRIPTION

[0070] The terms "first", "second", "third", "fourth" and the like in the description and claims of the application and above drawings (if any) are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily limit to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0071] For the sake of understanding, the specific flow of the embodiments of the application is described below. Please refer to Figure 1 An online energy monitoring system in the embodiments of the application includes a multi-source data acquisition module, an adaptive acquisition control module, a data preprocessing module, an energy analysis module and an intelligent response module.

[0072] The multi-source data acquisition module includes an electric energy sensor group, an electrical characteristic sensor and a device state sensor. The electric energy sensor group is used to monitor basic energy data, the electrical characteristic sensor is used to monitor the quality state of the power grid, and the device state sensor is used to monitor the operation health degree of the equipment.

[0073] The electric energy sensor group includes electric energy sensors and power meters, which capture real-time power consumption, power, and other information, helping to understand power consumption and trends. The electrical characteristic sensor includes voltage sensors, current sensors, and power factor sensors, which monitor the stability and health of the power grid, such as detecting voltage fluctuations, current imbalances, and power factor changes, which are crucial for ensuring the safe operation of the power grid. Temperature sensors, vibration sensors, and other device state sensors monitor the working temperature and vibration frequency of key equipment (transformers, circuit breakers) to assess the working state of the equipment and detect potential faults in advance, such as overheating or abnormal vibration, which may be an early sign of mechanical component wear. The multi-source data acquisition module integrates multiple types of sensors to achieve comprehensive and multi-level monitoring of the energy network, providing solid data support for the online energy monitoring system.

[0074] The adaptive acquisition control module establishes a risk assessment model for monitoring parameters based on the data provided by the multi-source data acquisition module, and dynamically adjusts the acquisition frequency and parameter acquisition range of the electric energy sensor group, electrical characteristic sensor, and device state sensor according to the monitoring parameter change trend.

[0075] The adaptive acquisition control module includes the establishment and dynamic adjustment of the risk assessment model for monitoring parameters. The establishment of the risk assessment model for monitoring parameters is as follows:

[0076] 1) Parameter risk level division (dynamic quantile division method);

[0077]

[0078] where: Q k (p) is the parameter risk level division result; F -1 (p) is the inverse function of the empirical distribution function, representing the value of the cumulative distribution function corresponding to the probability p; x is the current parameter value.

[0079] 2) Dynamic calculation of risk weight (weight distribution based on information entropy);

[0080] ; where: H i = -∑p k log2p k ;

[0081] w i is the risk weight of the i-th parameter; H i is the information entropy of the i-th parameter, reflecting its uncertainty; p k is the frequency of the i-th parameter in the k-th risk interval. ∑(1-H j ): the weighted sum of information entropy of all parameters.

[0082] The risk assessment model of the monitoring parameter is:

[0083] ;

[0084] R t represents the comprehensive risk value at the current time (time point t), w i is the risk weight of the i-th parameter, r i (t) is the risk level of the i-th parameter at the current time. λ is the trend sensitivity coefficient, used to adjust the influence degree of the risk change rate on the final risk value. dR / dt is the change rate of the risk value, that is, the change trend of the risk with time.

[0085] The implementation of dynamic adjustment specifically includes:

[0086] 1) Sliding window mechanism;

[0087]

[0088] Wherein: is the variance at the current time, α is the time decay factor, x t is the observation value at the current time, μ t-1 is the mean value at the previous time. ; wherein, is the weighted moving average value at the current time.

[0089] 2) Dynamic boundary calculation;

[0090] ; the dynamic adjustment coefficient k is determined according to the size of R t :

[0091] ; [L t , H t ] is the upper and lower limit range of the monitoring parameter (i.e. dynamic boundary).

[0092] 3) Trend sensitive mechanism; ; T is the time span of the sliding window.

[0093] The risk assessment model of the monitoring parameter is responsible for generating the current risk value, reflecting the instantaneous state of the system.

[0094] The dynamic boundary calculation and sliding window mechanism are responsible for optimizing the data parameter acquisition range, ensuring that the collected data is representative. The trend sensitive mechanism is responsible for capturing the change trend of the risk, helping the system to early warn potential problems.

[0095] The core goal of dynamically adjusting the acquisition frequency is to optimize data acquisition efficiency based on Rt and the change trend dR / dt, ensure real-time monitoring of key parameters, and reduce resource consumption for low-risk parameters. This is divided into the following steps:

[0096] Generate the comprehensive risk value at the current time based on the risk assessment model.

[0097] B) Capture the change trend of risk using the sliding window mechanism and time decay factor.

[0098] C) Dynamically adjust the acquisition frequency based on the comprehensive risk value and change trend.

[0099] D) Optimize the acquisition range (dynamic boundary calculation): Adjust the upper and lower limit range of parameters based on the risk value to further optimize acquisition efficiency.

[0100] Determine the acquisition frequency adjustment strategy: Based on the comprehensive risk value Rt and the change trend dR / dt, develop an acquisition frequency adjustment strategy:

[0101] 1) High-risk area (R t >R high or dR / dt>dR threshold ):

[0102] Increase the acquisition frequency to monitor high-risk parameters in real time. Adjustment formula: f new =f base ×k high； f new is the new acquisition frequency, f base is the basic acquisition frequency. k high is the high-risk adjustment coefficient (usually takes a value of 2~5). R high is the high-risk threshold. dR threshold is the threshold of risk change rate.

[0103] 2) Medium-risk area (R low ≤R t ≤R high and dR / dt≤dR threshold ): Maintain the basic acquisition frequency. f new =f base。

[0104] 3) Low-risk area (R t <R low and dR / dt≤dR threshold ): Reduce the acquisition frequency to save resources. Adjustment formula: f new =f base ×k low , k low is the low-risk adjustment coefficient (usually takes a value of 0.2~0.5). Rlow Low risk threshold.

[0105] The adaptive acquisition control module can dynamically adjust the acquisition frequency and parameter acquisition range according to the risk value and trend of the monitoring parameters, and the specific effects include: improving data acquisition efficiency: increasing the acquisition frequency in high-risk areas to ensure real-time monitoring of key parameters. Reduce the acquisition frequency in the low-risk area, save resources.

[0106] Enhance the response capability of the system: the sliding window mechanism and the trend-sensitive mechanism can quickly capture the risk change trend and give early warning of potential problems.

[0107] Optimize data quality: dynamic boundary calculation ensures that the collected data is representative, avoiding the collection of redundant or invalid data.

[0108] Reduce operating costs: by dynamically adjusting the acquisition frequency and range, reduce unnecessary resource consumption. The adaptive acquisition control module combines risk assessment model, sliding window mechanism and dynamic boundary calculation to realize intelligent adjustment of the acquisition frequency and range of the electric energy sensor group, electrical characteristic sensor and device state sensor. This method not only improves the efficiency and quality of data acquisition, but also enhances the stability and reliability of the system, providing important support for the intelligent management of online energy monitoring systems.

[0109] The data preprocessing module receives the raw data from the data acquisition module, and performs filtering, denoising and standardization processing;

[0110] Data alignment: solve the time deviation problem of multi-source data through time synchronization (IEEE 1588 PTP protocol) and resampling.

[0111] Outlier processing: use the improved Hampel filter to remove outliers, and use the ARIMA model (historical values and predicted values) to fill in missing values.

[0112] The improved Hampel filter process is as follows:

[0113] Dynamic window selection: determine the window size k according to the local data characteristics.

[0114] Weighted statistical quantity calculation: calculate the weighted median M w and weighted MAD in the window.

[0115] Adaptive threshold determination: if ∣x i -M w ∣>λ•MAD w , marked as abnormal.

[0116] Multi-stage verification: secondary inspection is performed on the boundary abnormal value.

[0117] ARIMA filling: Predicts and replaces values ​​at outlier locations using historical data.

[0118] Improved Hampel filter adaptive threshold and multi-stage detection reduce false positives, decrease misjudgment rate, reduce the impact of edge noise, and the dynamic window adapts to data changes, thus improving the accuracy of outlier handling.

[0119] Signal denoising: Wavelet threshold denoising technology is used to remove high-frequency noise while preserving the main characteristics of the signal.

[0120] Feature standardization: Transforms data to a uniform scale through dynamic Z-score standardization.

[0121] The energy analysis module receives preprocessed data from the data preprocessing module, uses a BP neural network to evaluate energy efficiency, an isolated forest algorithm to identify abnormal patterns, and vibration spectrum analysis to predict equipment failures. The energy analysis module then transmits the analysis results to the intelligent response module, which formulates corresponding response strategies based on the analysis results.

[0122] 1. The specific steps for receiving preprocessed data from the data preprocessing module and evaluating energy efficiency using a BP neural network are as follows:

[0123] 1) Prepare the training dataset; the data preprocessed by the data preprocessing module constitutes the training dataset.

[0124] 2) Input layer: n-dimensional features such as electrical energy, power factor, and temperature;

[0125] Hidden layer: uses hyperbolic tangent activation function;

[0126] Output layer: The Sigmoid function outputs an energy efficiency score.

[0127] 3) Training the BP neural network model: Adjust the weights using the backpropagation algorithm to minimize the error.

[0128] The basic calculation formula of a BP neural network involves forward propagation and backward propagation processes, where forward propagation is used to calculate the output and backward propagation is used to update the weights to reduce errors.

[0129] 2. The steps for identifying abnormal patterns using the Isolation Forest algorithm are as follows:

[0130] Data preparation: Collect and standardize normal operation data over a period of time.

[0131] Model building: The model is built based on the isolated forest algorithm, and the number of trees and the subsample size are set.

[0132] Anomaly detection: Input real-time data, and the model determines whether a point is an anomaly based on the length of the decision path.

[0133] Isolation Forest (IF) decides whether a sample is an outlier by calculating how many times it needs to be isolated.

[0134] 1) Isolation Tree construction procedure;

[0135] Random subsampling: ψ samples (default 256) are drawn from the dataset without replacement as the training set for a single tree.

[0136] Recursive split: Randomly select a dimension in the feature space. Randomly select a split point between the min and max of that dimension. Split the data into left and right child based on the split value.

[0137] Termination condition: current node contains only 1 sample; all samples have the same feature value; tree height reaches limit l = [log2 ψ]. ψ is the subsample size.

[0138] 2. Outlier detection procedure;

[0139] Forest construction: generate t isolation trees (t = 100).

[0140] Path computation: for each sample x, compute its path length h(x) in each tree. i

[0141] Score computation: compute the average path length ; h(x) is the sample path length. Substitute into the formula s(x, n) = 2 -E(h(x)) / c(n) ; where c(n) is the normalized path length and s(x, n) is the outlier score.

[0142] Decision rule: s ≥ 0.6: clear outlier (needs manual verification); 0.5 ≤ s < 0.6: suspected outlier (log record); s < 0.5: normal sample.

[0143] Example;

[0144] Isolation Forest construction: set ψ = 256, t = 100, then the tree height limit l = [log2 256] = 8;

[0145] Outlier detection: the average path length E(h(x)) = 3.2 of a sample x in 100 trees;

[0146] Compute the normalized path length: c(10000) ≈ 2ln(10000) + 0.5772 = 20.723;

[0147] Outlier score: s = 2 -3.2 / 20.723 ≈ 0.89;

[0148] Decision: s = 0.89 > 0.6, determine as an outlier, trigger equipment maintenance process.​

[0149] 3. Vibration spectrum analysis to predict equipment failure;

[0150] Collect vibration data: Obtain vibration signals from equipment status sensors.

[0151] Spectrum analysis: Apply Fast Fourier Transform (FFT) to convert time-domain signals to frequency-domain signals.

[0152] Fault diagnosis: Analyze whether there is a specific type of fault (such as imbalance, etc.) based on frequency components.

[0153] The energy analysis module integrates BP neural network, isolated forest algorithm and vibration spectrum analysis technology, which can comprehensively evaluate energy efficiency, identify abnormal patterns and predict equipment failure. This multi-dimensional analysis method not only improves the intelligent level of energy management, but also provides scientific basis for stable operation of the system. The following is a specific summary of its effects and benefits:

[0154] Through the nonlinear modeling of multi-dimensional features (such as power, power factor, temperature, etc.) by BP neural network, the energy efficiency level of the system or equipment can be accurately evaluated. BP neural network has self-learning ability and can continuously adjust weights according to the latest data to adapt to system changes. Energy efficiency score can help identify inefficient equipment or operation mode, providing basis for optimization scheduling.

[0155] Isolated forest algorithm constructs isolated trees based on random partitioning, which can quickly calculate the path length of each sample and determine whether it is an outlier. It provides clear decision rules to distinguish normal, suspected abnormal and clear abnormal samples. Without labeled data, it can complete anomaly detection, reducing the cost of manual intervention. By identifying abnormal patterns (such as abnormal power consumption, power grid quality fluctuations), it can issue warnings before problems worsen, avoiding major losses. Real-time monitoring and rapid response help ensure the safe operation of the power grid and equipment. Standardization processing using the statistical properties of path length reduces the impact of noise on anomaly detection. The energy analysis module combines BP neural network, isolated forest algorithm and vibration spectrum analysis to achieve comprehensive monitoring of energy systems from energy efficiency, anomalies to equipment health status. The energy analysis module realizes comprehensive monitoring and intelligent management of energy systems through the organic combination of BP neural network, isolated forest algorithm and vibration spectrum analysis.

[0156] The energy analysis module transmits the analysis results to the intelligent response module, which formulates appropriate response strategies based on the analysis results, specifically:

[0157] Receive and analyze analysis results;

[0158] Input: Energy efficiency score, anomaly score and equipment health status from the energy analysis module.

[0159] Energy efficiency score: output by BP neural network, ranging from (0, 1].

[0160] Anomaly score: calculated by Isolation Forest algorithm, reflecting whether the data point is abnormal.

[0161] Device health status: based on vibration spectrum analysis, assess whether the device has a risk of failure.

[0162] Develop response strategies; according to the results of analysis, the intelligent response module needs to develop specific response strategies.

[0163] Energy efficiency score response strategy; condition: when energy efficiency score S efficiency <0.5, indicating low energy efficiency. Strategies include reducing the operating time or load of high-energy-consuming devices, increasing the workload of low-energy-consuming devices; adjusting the device operating schedule to take advantage of the peak-valley electricity price difference to reduce costs. Enable energy-saving mode or take other energy-saving measures.

[0164] Anomaly score response strategy; condition: when anomaly score S anomaly ≥0.6, indicating clear abnormalities; when 0.5≤S anomaly <0.6, indicating suspected abnormalities. Trigger an alarm when there is a clear anomaly: notify relevant personnel immediately for manual verification, and take appropriate measures according to the type of anomaly, such as replacing faulty equipment or repairing the power grid. Record logs when there is a suspected anomaly: record the abnormal situation in the system log for subsequent analysis. Increase the monitoring frequency of the area or device to continuously observe its behavior changes.

[0165] Device health status response strategy; when the device health status shows a potential risk of failure (e.g., certain frequency components are outside the normal range). Perform advance maintenance: based on the results of vibration spectrum analysis, schedule preventive maintenance for the device to avoid sudden failures, and replace critical components (such as bearings, gears, etc.) if severe wear is detected. Adjust the device's operating parameters based on its current state to reduce the risk of further damage.

[0166] An online energy monitoring method, including multi-source data acquisition, adaptive acquisition control, data preprocessing, energy analysis and intelligent response;

[0167] Multi-source data acquisition captures basic energy data such as electricity consumption and power in real time through the deployment of a group of electrical energy sensors. Use electrical characteristic sensors to monitor the quality of the power grid and detect problems such as voltage fluctuations and current imbalances. Use device status sensors to monitor the operating temperature and vibration frequency of key devices such as transformers and circuit breakers to assess the health of the device.

[0168] Adaptive acquisition control establishes a risk assessment model for monitoring parameters based on historical data: parameter risk level classification: risk level classification of parameters through dynamic quantile method; risk weight dynamic calculation: weight allocation based on information entropy, combined with current risk value and trend. Dynamic adjustment of acquisition frequency and parameter range includes:

[0169] 1) Sliding window mechanism: calculate weighted moving average and variance, adjust time decay factor;

[0170] 2) Dynamic boundary calculation: dynamically adjust the upper and lower limit range of monitoring parameters according to the comprehensive risk value.

[0171] 3) Trend-sensitive mechanism: capture the trend of risk changes, and early warning potential problems.

[0172] Data preprocessing includes data alignment: IEEE 1588 PTP protocol is used to solve the time deviation problem of multi-source data, and resampling is performed. Outlier processing: use improved Hampel filter to remove outliers, and fill in missing values through ARIMA model. Signal denoising: use wavelet threshold denoising technology to remove high-frequency noise and retain the main characteristics of the signal. Feature standardization: convert data to a unified scale through dynamic Z-score standardization.

[0173] Energy analysis assesses energy efficiency through BP neural network, identifies abnormal patterns through isolation forest algorithm, and predicts equipment failure through vibration spectrum analysis.

[0174] Intelligent response receives and analyzes analysis results, formulates response strategies, including energy efficiency score response strategy, abnormal score response strategy and equipment health state response strategy.

[0175] Multi-source data acquisition module: through the integration of various sensors, realize all-round, multi-level energy network monitoring, ensure the comprehensiveness and accuracy of data.

[0176] Adaptive acquisition control module: use historical data to establish a risk assessment model, dynamically adjust the acquisition frequency and parameter range of the sensor, and optimize the data acquisition efficiency and representativeness.

[0177] Data preprocessing module: through filtering, denoising, alignment and standardization processing, improve data quality, and provide reliable foundation for subsequent analysis.

[0178] Energy analysis module: BP neural network: assess energy efficiency level through nonlinear modeling, support system optimization scheduling and energy saving decision. Isolation forest algorithm: based on random partition to build isolated tree, quickly detect abnormal points, reduce artificial intervention cost. Vibration spectrum analysis: extract frequency components of vibration signals through FFT, accurately diagnose equipment failure.

[0179] Intelligent response module: according to the analysis result, the response strategy is formulated, the self adjustment and optimization of the system are realized, and the energy utilization efficiency and equipment operation reliability are improved.

[0180] The computer device can be a server. The computer device comprises a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in the embodiment. The network interface of the computer device is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to implement the above method.

[0181] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, system and unit can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0182] The integrated unit, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0183] The above-described embodiments are only used to illustrate the technical solutions of the present application, and not to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An online energy monitoring system, characterized in that, It includes a multi-source data acquisition module, an adaptive acquisition and control module, a data preprocessing module, an energy analysis module, and an intelligent response module; The multi-source data acquisition module includes a power sensor group, an electrical characteristic sensor, and an equipment status sensor. The power sensor group is used to monitor basic energy data, the electrical characteristic sensor is used to monitor the power grid quality status, and the equipment status sensor is used to monitor the equipment's operational health. The adaptive acquisition and control module uses historical data to establish a risk assessment model for monitoring parameters based on the data provided by the multi-source data acquisition module, and dynamically adjusts the acquisition frequency and parameter range of the power sensor group, electrical characteristic sensor and equipment status sensor according to the changing trend of the monitoring parameters. The data preprocessing module receives raw data from the multi-source data acquisition module and performs filtering, noise reduction, and standardization processing. The energy analysis module receives preprocessed data from the data preprocessing module, uses a BP neural network to evaluate energy efficiency, an isolated forest algorithm to identify abnormal patterns, and vibration spectrum analysis to predict equipment failures. The energy analysis module then transmits the analysis results to the intelligent response module, which formulates corresponding response strategies based on the analysis results. The adaptive acquisition and control module includes the establishment of a risk assessment model for the monitored parameters, specifically: 1) Parameter risk level classification; Among them: Q k (p) represents the parameter risk level classification result; F -1 (p) is the inverse function of the empirical distribution function, representing the cumulative distribution function corresponding to the probability p; x is the current parameter value; 2) Dynamic calculation of risk weights; ;wherein: H i =-∑p k log2p k ; in: w i H represents the risk weight of the i-th parameter; i For the first i Information entropy of each parameter; p k For the first i The parameter in the first... k The frequency of occurrence of each risk zone; 3) The risk assessment model for the monitoring parameters is as follows: ； Where: R t The overall risk value at the current moment, r i ( t ) represents the risk level of the i-th parameter at the current moment; λ is the trend sensitivity coefficient, and dR / dt is the rate of change of the risk value; The core objective of dynamically adjusting the acquisition frequency is based on R t And the changing trend dR / dt, to optimize data acquisition efficiency, specifically including: The comprehensive risk value for the current moment is generated based on the risk assessment model; Utilize sliding window mechanisms and time decay factors to capture changing trends in risk; The data collection frequency is dynamically adjusted based on the overall risk value and its changing trend. Dynamic boundary calculation adjusts the upper and lower limits of parameters based on risk values, further optimizing data collection efficiency; The implementation of dynamic adjustment specifically includes: 1) Sliding window mechanism; ； in: Let be the variance at the current moment, and α be the time decay factor. x t The observed value at the current moment, μ t-1 This is the average value at the previous moment; ; This is the weighted moving average at the current moment; 2) Dynamic boundary calculation; Dynamic adjustment coefficient k Its value is based on R t Size determines: ;[ L t , H t [] represents the upper and lower limits of the monitored parameters, i.e., the dynamic boundaries.

2. The online energy monitoring system according to claim 1, characterized in that: The power sensor group includes power sensors and power meters to capture power consumption and power information in real time; the electrical characteristic sensors include voltage sensors, current sensors and power factor sensors to monitor the stability and health of the power grid; the equipment status sensors include temperature sensors and vibration sensors, which assess the operating status of equipment by monitoring the operating temperature and vibration frequency of key equipment.

3. The online energy monitoring system according to claim 2, characterized in that: Based on the comprehensive risk value R t Based on the changing trend dR / dt, formulate a sampling frequency adjustment strategy: a) High-risk area is R t >R high Or dR / dt>dR threshold : Increase the data collection frequency to monitor high-risk parameters in real time; adjust the formula: f new =f base ×k high； f new For the new sampling frequency, f base Based on the sampling frequency, k high R is a high-risk adjustment factor. high As a high-risk threshold, dR threshold The threshold for the rate of change of risk; b) The medium-risk area is R low ≤R t ≤R high And dR / dt≤dR threshold Maintain the basic acquisition frequency, f new =f base ; c) Low-risk areas are R t <R low And dR / dt≤dR threshold Reduce the sampling frequency to save resources, and adjust the formula: f new =f base ×k low k low R is the low-risk adjustment factor. low This is the low-risk threshold.

4. The online energy monitoring system according to claim 3, characterized in that: The data preprocessing module handles the following: Data alignment addresses the time discrepancy issue of multi-source data through time synchronization and resampling. Outlier handling utilizes an improved Hampel filter to remove outliers and fills in missing values ​​using an ARIMA model; Signal denoising employs wavelet threshold denoising technology to remove high-frequency noise while preserving the main characteristics of the signal. Feature standardization transforms data to a uniform scale through dynamic Z-score standardization.

5. The online energy monitoring system according to claim 4, characterized in that: The energy analysis module receives preprocessed data from the data preprocessing module and uses a BP neural network to evaluate energy efficiency. The specific steps are as follows: A) Prepare the training dataset; The data preprocessed by the data preprocessing module constitutes the training dataset; B) Input layer: n-dimensional features: electrical energy, power factor, and temperature; Hidden layer: uses hyperbolic tangent activation function; Output layer: The Sigmoid function outputs an energy efficiency score; C) Training the BP neural network model: Adjust the weights using the backpropagation algorithm to minimize the error.

6. The online energy monitoring system according to claim 5, characterized in that: The energy analysis module transmits the analysis results to the intelligent response module, which then formulates corresponding response strategies based on the analysis results, including: S71. Receive and parse the analysis results; Inputs: Energy efficiency score, anomaly score, and equipment health status from the energy analysis module; Energy efficiency score: Output by a BP neural network, ranging from (0, 1]; Anomaly score: Calculated by the Isolation Forest algorithm, reflecting whether a data point is an anomaly; Equipment health status: Based on vibration spectrum analysis, assess whether the equipment has any risk of failure; S72. Formulate response strategies; Based on the analysis results, the intelligent response module needs to formulate specific response strategies, including energy efficiency score response strategies, anomaly score response strategies, and equipment health status response strategies.

7. The online energy monitoring system according to claim 6, characterized in that: Develop a response strategy; based on the analysis results, the intelligent response module needs to develop a specific response strategy. 1) Energy efficiency rating response strategy; Condition: When the energy efficiency rating is S efficiency A value <0.5 indicates low energy efficiency. Strategies include reducing the operating time or load of high-energy-consuming equipment, increasing the workload of low-energy-consuming equipment, adjusting equipment operating schedules, and utilizing peak-valley electricity price differences to reduce costs. 2) Abnormal score response strategy; Condition: When the abnormal score S anomaly A value ≥0.6 indicates a clear abnormality; When 0.5≤S anomaly A value less than 0.6 indicates a suspected anomaly; when an anomaly is confirmed, an alarm is triggered: relevant personnel are immediately notified to conduct manual verification, and appropriate measures are taken according to the anomaly type; when an anomaly is suspected, a log is recorded: the anomaly is recorded in the system log for subsequent analysis; 3) Equipment health status response strategy; When the health status of the equipment indicates a potential risk of failure, maintenance should be carried out in advance. Based on the vibration spectrum analysis results, preventive maintenance of the equipment should be arranged to avoid sudden failures.

8. The monitoring method of the online energy monitoring system according to claim 7, characterized in that, include: Multi-source data acquisition, adaptive acquisition and control, data preprocessing, energy analysis, and intelligent response; The multi-source data acquisition includes a power sensor group, an electrical characteristic sensor, and an equipment status sensor. The power sensor group is used to monitor basic energy data, the electrical characteristic sensor is used to monitor the power grid quality status, and the equipment status sensor is used to monitor the operational health of the equipment. The adaptive acquisition control is based on data provided by the multi-source data acquisition module, uses historical data to establish a risk assessment model for monitoring parameters, and dynamically adjusts the acquisition frequency and parameter range of the power sensor group, electrical characteristic sensor and equipment status sensor according to the changing trend of monitoring parameters. Data preprocessing receives raw data from the data acquisition module and performs filtering, noise reduction, and standardization. Energy analysis receives preprocessed data from the data preprocessing module, uses a BP neural network to evaluate energy efficiency, an isolated forest algorithm to identify abnormal patterns, and vibration spectrum analysis to predict equipment failures. The energy analysis then transmits the analysis results to the intelligent response module, which formulates corresponding response strategies based on the analysis results.

Images