Method and system for realizing intelligent operation and maintenance function of digital energy management platform, and medium

Abstract

The invention provides an intelligent operation and maintenance function realization method of a digital energy management platform, and aims to realize intelligent maintenance and efficient operation and maintenance of energy management equipment through evaluation of the health state of the equipment and life prediction. The method comprises the steps of data acquisition and fusion, data storage and management, data analysis and modeling and intelligent operation and maintenance strategies, adopts various sensor data as input, extracts key features of equipment through feature enhancement and deep learning technologies, and dynamically evaluates feature data in combination with an autoregression model, a convolutional neural network and the like. And the health adjustment function is used for comprehensively evaluating the operation state of the equipment, so that the residual life is accurately predicted, and active maintenance, fault early warning and performance optimization of the equipment are realized. The method has the beneficial effects of high accuracy, high intelligence, improved operation and maintenance efficiency and the like, and is suitable for equipment health monitoring and intelligent operation and maintenance scenes in digital energy management.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic equipment maintenance, and mainly to a method, system and medium for realizing intelligent operation and maintenance functions of a digital energy management platform. Background Art

[0002] With the rapid development of society and the continuous increase in energy demand, efficient use and intelligent management of energy have become particularly important. Traditional energy management systems mainly manage the operating conditions and energy usage of various types of equipment through manual operation and basic monitoring methods. However, this management model has significant limitations, such as energy waste, frequent manual intervention, and low data processing efficiency. Especially in the complex environment of large-scale industrial equipment, building energy management and smart cities, traditional management methods cannot meet the requirements of modern energy systems for real-time, accuracy and efficiency.

[0003] In practical applications, energy management systems need to have intelligent operation and maintenance functions such as automated analysis of equipment status, fault prediction, and operation optimization to reduce equipment failure rates and maintenance costs and improve the overall efficiency of the system. However, existing energy management platforms usually lack intelligent operation and maintenance strategies and cannot effectively perform predictive maintenance and exception handling on equipment. Therefore, how to design a digital energy management platform that combines advanced intelligent operation and maintenance functions to achieve optimized energy use and efficient operation and maintenance of equipment is a topic that needs to be urgently addressed in the current technical field. Summary of the invention

[0004] The present invention aims to at least solve the technical problems existing in the prior art, and in particular innovatively proposes a

[0005] In order to achieve the above-mentioned object of the present invention, the present invention provides a method for realizing the intelligent operation and maintenance function of a digital energy management platform, comprising:

[0006] S1, data collection and fusion, collects energy-related data from sensors and combines them into a matrix;

[0007] S2, data storage and management, storing data in the database;

[0008] S3, data analysis and modeling, using predictive models to model energy usage trends and service life;

[0009] S4, intelligent operation and maintenance strategy, uses model output to perform health assessment and life prediction of equipment for predictive maintenance.

[0010] In a preferred embodiment of the present invention, the step 1 comprises:

[0011] Collect equipment data, including:

[0012]

[0013] Among them, D i (t) represents the data collected by the i-th device, V i (t) is the voltage of the ith device, I i (t) is the current of the ith device, P i (t) is the power of the ith device, f i (t) is the frequency of the ith device, i is the index of the device;

[0014] Combine the acquired data into a matrix:

[0015] D(t)=[D1(t) D2(t) … D N (t)];

[0016] Where D(t) is the data matrix, t represents time, and N represents the number of devices.

[0017] In a preferred embodiment of the present invention, the step 2 comprises:

[0018] The data of each device is stored in the database, which can be expressed as:

[0019] S database ={(t,D i (t))∣i=1,2,…,N,t∈T};

[0020] Among them, S database Represents the data set of all devices stored in the database, D i (t) represents the data collected by the i-th device, where i is the index of the device;

[0021] Based on the time prediction model, we predict possible high-frequency query data and use the predicted caching strategy to reduce query latency:

[0022]

[0023] f LSTM (*) is the predicted output based on the LTSM model, D i (tr),…,D i (t-1) is the historical time series of device i, r is the maximum historical time step, and the model infers the commonly used device data based on the past query history and converts the inferred commonly used query data Put it into cache to reduce query latency.

[0024] In a preferred embodiment of the present invention, step 3 comprises:

[0025] The autoregressive integrated moving average model is used to build the model. The initial predicted feature matrix is ​​D′(t). The formula for making a preliminary prediction of future feature data is:

[0026] Φ(B)(1-B) d D′(t)=Θ(B)∈(t);

[0027] Where B is the lag operator, Φ(B) is the autoregressive polynomial, Θ(B) is the moving average polynomial, d is the difference order, and ∈(t) is the white noise term;

[0028] Among them, D′(t)=[D′1(t) D′2(t) … D′ N (t)];

[0029]

[0030] D″(t) is the feature data matrix after enhanced prediction, γ0 is the intercept term, which represents the baseline value without feature input;

[0031] is the feature enhancement term, α u (t) is the attention weight, which dynamically assigns weights to each feature to determine the importance of the feature on energy consumption at the current time point. The number of attention-enhanced features is U, and D′ enhanced,u (t) is the u-th attention enhancement feature, which is calculated from the u-th eigenvector in the initial predicted feature matrix D′(t), and the mean and variance in the sliding window are extracted as the enhancement feature;

[0032] in,

[0033] Extract the data of the first k time steps for mean calculation, and calculate the variance within the sliding window to reflect the degree of fluctuation of the feature, D′ u (τ) is the feature vector in the initial predicted feature matrix D′(τ), u is the vector index, the value at time τ;

[0034] is the convolution feature term, W conv,j is the weight of the convolution feature, D′ j (t) is the feature vector in the initial predicted feature matrix D′(t), j is the index of the vector in the matrix, the number of convolution features is J, and γ(t) is the error term;

[0035] Among them, the number N = U = J;

[0036] Combining the advantages of linear regression and deep learning, the model's ability to understand complex feature relationships is improved by introducing enhanced features, attention mechanisms, and convolutional features.

[0037] In a preferred embodiment of the present invention, step 4 comprises:

[0038] The enhanced predicted feature matrix D″(t) is added to the comprehensive health status evaluation function for predictive maintenance:

[0039] RUL(t)=R0e -λt f h (D″ m (t));

[0040] Where RUL(t) is the remaining useful life, R0 is the initial life, and e -λt is the exponential decay term, λ is the decay rate of the device;

[0041] f h (D″ m (t)) is a comprehensive health status assessment function, which uses a weighted linear combination of the data features of the equipment to assess the health status of the equipment and then infer the remaining life of the equipment;

[0042]

[0043] β0 is a constant term used to adjust the benchmark of the health index and provide a basic reference value for equipment life in the absence of other health influencing factors;

[0044] represents the weighted impact of data features from different devices on device life, β m is the weight of the mth device, S(D″ m (t)) is the projection function, which transforms the feature data D″ in vector form m (t) Projecting into a standardized value in scalar form so that the data of each device can be weighted and summarized under the same standard;

[0045] D″ m (t) is the feature data after enhanced prediction on the mth device, M is the number of devices, where M=N.

[0046] The present invention also discloses a computer system comprising:

[0047] a memory for storing processor-executable instructions;

[0048] Among them, the processor is configured to implement the method for realizing the intelligent operation and maintenance function of a digital energy management platform when executing the executable instructions.

[0049] The present invention also discloses a computer-readable storage medium comprising:

[0050] a memory having a computer program stored thereon;

[0051] A processor is used to execute the program in the memory to implement the method for realizing the intelligent operation and maintenance function of a digital energy management platform.

[0052] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0053] First, the present invention realizes real-time monitoring and analysis of the equipment operation status by adopting intelligent data collection and processing technology. By using IoT sensors and big data technology, the platform can obtain a variety of energy-related data, perform efficient data fusion and processing, and thus provide the system with more comprehensive energy consumption information. This real-time data processing capability greatly improves the accuracy and efficiency of energy management, allowing managers to make accurate decisions based on the status of the equipment and reduce energy waste.

[0054] Secondly, the present invention uses machine learning algorithms to model the historical data of the equipment, realizing predictive analysis and fault diagnosis of the equipment's operating status. Through technologies such as remaining useful life (RUL) prediction and equipment health index (HI) evaluation, the system can issue warnings before the equipment is about to fail and formulate maintenance plans in advance. This predictive maintenance not only reduces the risk of equipment failure, but also reduces the economic losses caused by unexpected equipment downtime, thereby significantly reducing operation and maintenance costs.

[0055] In addition, the present invention uses an energy scheduling optimization algorithm to intelligently allocate and control the use of energy. The system can optimize the energy use strategy based on the actual operation of the equipment, dynamic electricity prices, environmental conditions and other factors, reduce the impact of peak loads on the power grid, and achieve energy saving and consumption reduction effects. Compared with the traditional manual adjustment method, the energy scheduling method provided by the present invention is more flexible and more automated, and can effectively improve energy efficiency.

[0056] Finally, the present invention introduces device status index and data compression storage technology, which significantly improves the storage and query efficiency of large-scale data. Through differential compression storage and device indexing, the platform greatly saves storage space while ensuring data integrity and improves data access speed, making the system more efficient when processing massive device data.

[0057] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0059] Figure 1It is a system schematic diagram of the present invention. DETAILED DESCRIPTION

[0060] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0061] like Figure 1 As shown, the present invention proposes a method for realizing the intelligent operation and maintenance function of a digital energy management platform, including:

[0062] 1. Data collection and fusion

[0063] Energy data type:

[0064] Collect multiple data types related to energy consumption, including voltage, current, power, and frequency, and combine these data into vectors.

[0065]

[0066] Among them, D i (t) represents the data collected by the i-th device, V i (t) is the voltage of the ith device, I i (t) is the current of the ith device, P i (t) is the power of the ith device, f i (t) is the frequency of the ith device, i is the index of the device;

[0067] Combine the acquired data into a matrix:

[0068] D(t)=[D1(t)D2(t)…D N (t)];

[0069] Where D(t) is the data matrix, t represents time, and N represents the number of devices.

[0070] Energy equipment monitoring:

[0071] Monitor the operating status of energy equipment, including power transformers, photovoltaic inverters, air conditioning systems, and lighting systems. Collect real-time operating parameters of these devices to fully understand the operating status of the energy system.

[0072] 2. Data cleaning and preprocessing

[0073] Data quality assurance:

[0074] For power quality data (harmonics and voltage fluctuations), data cleaning is required to remove abnormal data points to ensure that the calculation model is accurate.

[0075] Feature extraction:

[0076] Extract features from energy consumption data, including peak and trough values, day and night differences, and load characteristics in different time periods. Extracting these features will help in the subsequent analysis of energy consumption patterns.

[0077] 3. Data storage and management

[0078] Energy data classification storage:

[0079] Categorize and store energy according to energy type and equipment, including storing electricity, cooling energy, and thermal energy in different data tables.

[0080] 3.1 Data Storage Format

[0081] The data of each device is stored in the database, which can be expressed as:

[0082] S database ={(t,D i (t))∣i=1,2,…,N,t∈T};

[0083] Among them, S database Represents the data set of all devices stored in the database, D i (t) represents the data collected by the i-th device, where i is the index of the device;

[0084] 3.2 Data Management and Query

[0085] Based on the time prediction model, we predict possible high-frequency query data and use the predicted caching strategy to reduce query latency:

[0086]

[0087] f LSTM (*) is the predicted output based on the LTSM model, D i (tr),...,D i (t-1) is the historical time series of device i, r is the maximum historical time step, and the model infers the commonly used device data based on the past query history.

[0088] Frequently used query data Put it into cache to reduce query latency.

[0089] Real-time and historical data management:

[0090] The real-time power consumption data is stored as a short-term database for real-time analysis; at the same time, the historical energy consumption data is stored as a long-term database for periodic analysis and trend identification.

[0091] 4. Data Analysis and Modeling

[0092] Energy usage trend analysis:

[0093] Analyze time series data of energy consumption, including daily, weekly, and monthly trends of electricity consumption, identify peak power consumption periods, and determine the main load sources of energy consumption.

[0094] 4.1 Energy usage trend analysis

[0095] The autoregressive integrated moving average model is used to build the model. The initial predicted feature matrix is ​​D′(t). The formula for making a preliminary prediction of future feature data is:

[0096] Φ(B)(1-B) d D′(t)=Θ(B)∈(t);

[0097] Where B is the lag operator, Φ(B) is the autoregressive polynomial, Θ(B) is the moving average polynomial, d is the difference order, and ∈(t) is the white noise term;

[0098] 4.2 Prediction Model

[0099] D′(t)=[D′1(t) D′2(t) … D′ N (t)];

[0100] The initial predicted feature matrix D′(t) is input into the linear regression model to establish the relationship between feature data and influencing factors:

[0101]

[0102] D″(t) is the feature data matrix after enhanced prediction, γ0 is the intercept term, which represents the baseline value without feature input;

[0103] is the feature enhancement term, α u (t) is the attention weight, which dynamically assigns weights to each feature to determine the importance of the feature on energy consumption at the current time point. The number of attention-enhanced features is U, and D′ enhanced,u (t) is the u-th attention enhancement feature, which is calculated from the u-th eigenvector in the initial predicted feature matrix D′(t), and the mean and variance in the sliding window are extracted as the enhancement feature;

[0104] in,

[0105] Extract the data of the first k time steps for mean calculation, and calculate the variance within the sliding window to reflect the degree of fluctuation of the feature, D′ u (τ) is the feature vector in the initial predicted feature matrix D′(τ), u is the vector index, the value at time τ;

[0106] is the convolution feature term, W conv,j is the weight of the convolution feature, D′ j (t) is the feature vector in the initial predicted feature matrix D′(t), j is the index of the vector in the matrix, the number of convolution features is J, and γ(t) is the error term;

[0107] Among them, the quantity N=U=J.

[0108] 5. Intelligent operation and maintenance strategy design

[0109] Predictive maintenance of energy equipment:

[0110] Predictive maintenance is performed based on the health status of energy equipment. For example, for power transformers, the health status of the transformer is assessed by monitoring the current and load conditions. When a parameter deviates from the normal range, advance maintenance is performed to prevent energy interruptions caused by faults.

[0111] The enhanced predicted feature matrix D″(t) is added to the comprehensive health status evaluation function for predictive maintenance:

[0112] RUL(t)=R0e -λt f h (D″ m (t));

[0113] Where RUL(t) is the remaining useful life, R0 is the initial life, and e -λt is the exponential decay term, λ is the decay rate of the device;

[0114] f h (D″ m (t)) is a comprehensive health status assessment function, which uses a weighted linear combination of the data features of the equipment to assess the health status of the equipment and then infer the remaining life of the equipment;

[0115]

[0116] β0 is a constant term used to adjust the benchmark of the health index and provide a basic reference value for equipment life in the absence of other health influencing factors;

[0117] represents the weighted impact of data features from different devices on device life, β mis the weight of the mth device, S(D″ m (t)) is the projection function, which transforms the feature data D″ in vector form m (t) Projecting into a standardized value in scalar form so that the data of each device can be weighted and summarized under the same standard;

[0118] D″ m (t) is the feature data after enhanced prediction on the mth device, M is the number of devices, where M=N.

[0119] Through this evaluation method, we can dynamically evaluate the remaining service life of the equipment, combine the health status and the impact of time, and formulate a more scientific equipment maintenance strategy to reduce the risk of failure and maintenance costs.

[0120] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A method for realizing intelligent operation and maintenance functions of a digital energy management platform, characterized in that: include: S1, data collection and fusion, collects energy-related data from sensors and combines them into a matrix; S2, data storage and management, stores data in the database and uses caching strategies to reduce query latency; S3, data analysis and modeling, using predictive models to model energy usage trends and service life; S4, intelligent operation and maintenance strategy, uses model output to perform health assessment and life prediction of equipment for predictive maintenance.

2. The method for realizing the intelligent operation and maintenance function of the digital energy management platform according to claim 1 is characterized in that: The step 1 comprises: Collect equipment data, including: Among them, D i (t) represents the data collected by the i-th device, V i (t) is the voltage of the ith device, I i (t) is the current of the ith device, P i (t) is the power of the ith device, f i (t) is the frequency of the i-th device, i is the index of the device, i=1, 2, 3, ..., N; Combine the acquired data into a matrix: D(t)=[D1(t)D2(t)…D N (t)]; Where D(t) is the data matrix, t represents time, and N represents the number of devices.

3. The method for realizing the intelligent operation and maintenance function of the digital energy management platform according to claim 1 is characterized in that: The step 2 comprises: The data of each device is stored in the database, which can be expressed as: S database ={(t,D i (t))∣i=1,2,...,N}; Among them, S database Represents the data set of all devices stored in the database, D i (t) represents the data collected by the i-th device, i is the index of the device; N represents the number of devices; Based on the time prediction model, we predict possible high-frequency query data and use the predicted caching strategy to reduce query latency: f LSTM (*) is the predicted output based on the LTSM model, D i (tr),…,D i (t-1) is the historical time series of device i, r is the maximum historical time step, and the model infers the commonly used device data based on the past query history and converts the inferred commonly used query data Put it into cache to reduce query latency.

4. The method for realizing the intelligent operation and maintenance function of the digital energy management platform according to claim 1 is characterized in that: The step 3 comprises: The autoregressive integrated moving average model is used to build the model. The initial predicted feature matrix is ​​D′(t). The formula for making a preliminary prediction of future feature data is: Φ(B)(1-B) d D′(t)=Θ(B)∈(t); Where B is the lag operator, Φ(B) is the autoregressive polynomial, Θ(B) is the moving average polynomial, d is the difference order, and ∈(t) is the white noise term; where D′(t) = [D′1(t) D′2(t) … D′ N (t)]; The initial predicted feature matrix D′(t) is input into the linear regression model to establish the relationship between feature data and influencing factors: D″(t) is the feature data matrix after enhanced prediction, γ0 is the intercept term, which represents the baseline value without feature input; is the feature enhancement term, α u (t) is the attention weight, which dynamically assigns weights to each feature to determine the importance of the feature on energy consumption at the current time point. The number of attention-enhanced features is U, and D′ enhanced,u (t) is the u-th attention enhancement feature, which is calculated from the u-th eigenvector in the initial predicted feature matrix D′(t), and the mean and variance in the sliding window are extracted as the enhancement feature; in, Extract the data of the first k time steps for mean calculation, and calculate the variance within the sliding window to reflect the degree of fluctuation of the feature, D′ u (τ) is the feature vector in the initial predicted feature matrix D′(τ), u is the vector index, the value at time τ; is the convolution feature term, W conv,j is the weight of the convolution feature, D′ j (t) is the feature vector in the initial predicted feature matrix D′(t), j is the index of the vector in the matrix, the number of convolution features is J, and γ(t) is the error term; Among them, the number N = U = J; Combining the advantages of linear regression and deep learning, the model's ability to understand complex feature relationships is improved by introducing enhanced features, attention mechanisms, and convolutional features.

5. The method for realizing the intelligent operation and maintenance function of the digital energy management platform according to claim 1 is characterized in that: The step 4 comprises: The enhanced predicted feature matrix D″(t) is added to the comprehensive health status evaluation function for predictive maintenance: RUL(t)=R0e -λt f h (D″ m (t)); Where RUL(t) is the remaining useful life, R0 is the initial life, and e -λt is the exponential decay term, λ is the decay rate of the device; f h (D″ m (t)) is a comprehensive health status assessment function, which uses a weighted linear combination of the data features of the equipment to assess the health status of the equipment and then infer the remaining life of the equipment; β0 is a constant term used to adjust the benchmark of the health index and provide a basic reference value for equipment life in the absence of other health influencing factors; represents the weighted impact of data features from different devices on device life, β m is the weight of the mth device, S(D″ m (t)) is the projection function, which transforms the feature data D″ in vector form m (t) Projecting into a standardized value in scalar form so that the data of each device can be weighted and summarized under the same standard; D″ m (t) is the feature data after enhanced prediction on the mth device, M is the number of devices, where M=N.

6. A computer system, characterized in that: include: a memory for storing processor-executable instructions; Wherein, the processor is configured to implement a method for realizing intelligent operation and maintenance functions of a digital energy management platform as described in any one of claims 1 to 5 when executing the executable instructions.

7. A computer-readable storage medium, characterized in that: include: a memory having a computer program stored thereon; A processor, used to execute the program in the memory to implement a method for realizing the intelligent operation and maintenance function of a digital energy management platform as described in any one of claims 1 to 5.

Images