Building energy management system and method based on cloud service
Through a building energy management system based on cloud services, using sensors and big data analysis technology, a variety of models are established to predict and optimize energy consumption, which solves the problem that traditional systems cannot dynamically adjust and process multi-dimensional data, and achieves the accuracy of energy consumption Management and energy efficiency improvement.
Patent Information
- Application Number
- CN202410993135.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-08
AI Technical Summary
Traditional building energy management systems cannot dynamically adjust the equipment operation mode and cannot effectively utilize real-time data, resulting in poor energy waste and energy consumption optimization results, and a single model is difficult to deal with multi-dimensional data and complex nonlinear relationships.
Adopt a building energy management system based on cloud services, through data acquisition, transmission, data analysis and evaluation feedback modules, use sensors to collect data in real time, use big data analysis and machine learning algorithms to establish a variety of models for energy consumption prediction and optimization, and dynamically adjust it. Equipment operation parameters.
Accurate prediction and dynamic adjustment of energy consumption are achieved, the limitations of a single model are overcome, energy efficiency is significantly improved, energy consumption is reduced, and user comfort and equipment stability are ensured.
Smart Images

Figure CN118915862A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of cloud services, and in particular to a building energy management system and method based on cloud services. Background Art
[0002] At present, elevators and air conditioning systems in buildings often operate according to preset fixed modes. This mode does not fully take into account real-time environmental changes and user needs, and the air conditioning system remains in operation in unused rooms and cannot automatically adjust operating parameters, all of which leads to energy waste. In addition, due to the lack of effective use of real-time data, it is difficult for existing management systems to dynamically adjust the operating mode of equipment and achieve optimization of energy consumption.
[0003] Traditional building energy management systems often rely on simple schedules and fixed parameter settings and lack the ability to respond to real-time data. For example, during the hot summer months, if the air-conditioning system fails to adjust according to the real-time ambient temperature and indoor traffic flow, it may lead to excessive energy consumption and reduced user comfort. This situation not only increases the operating costs of the building, but may also cause excessive wear of the equipment and shorten its service life.
[0004] At the same time, traditional optimization methods usually rely on a single model and cannot comprehensively consider multiple influencing factors. This method has significant shortcomings because it cannot simultaneously process multi-dimensional data and complex nonlinear relationships. For example, a single regression model may only be able to make simple predictions based on historical energy consumption data, but cannot take into account the comprehensive impact of multiple parameters such as real-time crowd density, ambient temperature, and humidity. The limitations of this single model lead to unsatisfactory optimization results and cannot achieve true energy saving and consumption reduction. At the same time, due to the lack of dynamic adjustment capabilities, a single model is difficult to adapt to rapidly changing environments and usage requirements, and cannot provide real-time optimization strategies, thereby limiting the overall performance and efficiency of the building energy management system. Summary of the invention
[0005] In view of the deficiencies in the prior art, the present invention provides a building energy management system and method based on cloud services, which solves the problems mentioned in the background technology.
[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A building energy management system based on cloud services, including a data acquisition module, a transmission module, a data analysis module, a model building module and an evaluation feedback module;
[0007] The data acquisition module collects human flow information, environmental information and equipment energy consumption information through several sensors integrated in the power equipment to form a collection data group;
[0008] The transmission module marks the collected data group, compresses and packages it, and transmits the packaged collected data group to the cloud server via the Internet;
[0009] The data analysis module pre-processes the uploaded collection data group, including data cleaning, data filtering, data deduplication and data standardization, to form a processed cloud storage data group, which is synchronously stored in the cloud collection history database;
[0010] The model building module uses big data analysis technology and machine learning algorithms to build time series models, regression models, random forest models and neural network models based on the data in the cloud-collected historical database, and simultaneously trains and analyzes to obtain time-predicted energy consumption. Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption
[0011] The evaluation feedback module predicts the energy consumption of the time obtained by training Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption Perform fitting to obtain the model optimization energy consumption fitting index E opt (t), synchronously match with the preset energy consumption optimization evaluation threshold P, obtain the power equipment energy optimization evaluation plan, and perform iterative training according to the content of the power equipment energy optimization evaluation plan.
[0012] Preferably, the data acquisition module includes an extraction unit;
[0013] The extraction unit collects information through several sensors integrated in the power equipment, and simultaneously extracts the flow of people information with fixed period content, including the number and density of people, extracts the internal and external environment information of the building, including the indoor and outdoor temperature and the indoor and outdoor light intensity, and extracts the energy consumption information of the working equipment, including the equipment start and stop cycle and the energy consumption, to form a collection data group;
[0014] The sensors include infrared sensors, cameras, temperature sensors, humidity sensors, light sensors, current sensors and voltage sensors;
[0015] The collected data group includes human flow information, environmental information, equipment energy consumption information, electricity price and equipment information.
[0016] Preferably, the transmission module includes a communication unit;
[0017] The communication unit marks the collected data group, including adding a timestamp, device ID, device location and data type, and then compresses and packages it using a compression algorithm, and simultaneously performs block packaging according to the device ID, generates a compressed package to form a compressed packaged collected data group, and transmits the packaged collected data group to the cloud server via the Internet;
[0018] Among them, compression algorithms include GZIP compression algorithm and LZ77 compression algorithm; packaging formats include JSON format and Protobuf format.
[0019] Preferably, the data analysis module includes a preprocessing unit and a storage unit;
[0020] The preprocessing unit preprocesses the uploaded collected data group, including data cleaning cloud service, data filtering, data deduplication and data standardization, to form a processed cloud storage data group;
[0021] Among them, data cleaning includes the use of interpolation, mean filling and statistical analysis methods to handle missing values and detect outliers; data filtering includes the use of filtering algorithms to filter noise; data deduplication includes the use of hash algorithms to detect and process deduplication; data standardization includes normalization and standardization to unify data formats;
[0022] The storage unit synchronously stores the processed cloud storage data group in the cloud acquisition history database, and performs data archiving processing and data type partitioning storage.
[0023] Preferably, the model building module includes a modeling unit and a training unit;
[0024] The modeling unit uses big data analysis technology and machine learning algorithms to establish a training analysis model for the data in the cloud collection history database, including using a time series analysis method to establish a time series model, using a linear regression method to establish a regression model, using a random forest algorithm to establish a random forest model, and using a neural network algorithm to establish a neural network model;
[0025] The training unit collects data stored in the historical database through the cloud and performs fixed-period interception to obtain the training feature vector X, inputs the training feature vector X into the established time series model, regression model, random forest model and neural network model machine for training, and obtains the time prediction energy consumption through the training of the time series model Obtain the regression energy consumption E through regression model training u (t), obtain the random predicted energy consumption E through the training of random forest model rf (t), obtain the network predicted energy consumption through the training of the neural network model
[0026] Among them, the training feature vector X includes {X p , X e , X c and X d}, where X p represents the subset of crowd flow information, X e represents the environmental information subset, X c Equipment energy consumption information subset, X d Electricity price and equipment information, a subset includes stored related collection information.
[0027] Preferably, the time-predicted energy consumption Obtained through the following calculation formula:
[0028]
[0029] In the formula, Time predicted energy consumption, specifically represents the predicted energy consumption at time t, represents the constant term, specifically the intercept of the model, p represents the natural regression order, represents the natural regression coefficient, specifically reflecting the impact of energy consumption at time m on current energy consumption, E(t+i) represents the time energy consumption at time ti, q represents the moving average order, θ j Represents the moving average coefficient, specifically reflecting the impact of the white noise at the previous j moments on the current energy consumption, ∈(tj) represents the white noise at time tj, and ∈(t) represents the white noise at the current time point t.
[0030] The regression uses energy consumption E u (t) is obtained by the following calculation formula:
[0031] E u (t) = α*G p (t)+β*G u (t)+χ*L u (t)+δ*R e (t);
[0032] In the formula, E u (t) Regression energy consumption, specifically represents the energy consumption of the device at time point t, G p (t) represents the density gradient of the flow of people at time t, G u (t) represents the number of equipment use cycles at time t, L u (t) represents the indoor light change rate at time t, R e (t) represents the environmental change rate at time point t, specifically including humidity rate, temperature change rate and air flow rate change rate.
[0033] Preferably, the random prediction energy consumption E rf(t) is obtained by the following calculation formula:
[0034]
[0035] In the formula, E rf (t) represents the energy consumption predicted by the random forest at time t, N represents the number of decision trees, T k (X) represents the output of the k-th decision tree, and X represents the training feature vector;
[0036] The network predicts energy consumption Obtained through the following calculation formula:
[0037]
[0038] In the formula, represents the network prediction energy consumption, specifically the neural network prediction energy consumption at time t, σ represents the activation function, specifically including ReLU, Sigmoid and Tanh nonlinear functions, to introduce nonlinear features, W (L) represents the weight matrix of the Lth layer, specifically the connection weights between the neurons in the Lth layer and the L-1th layer, b (L) It represents the bias vector of the Lth layer, specifically the bias term of the Lth layer neuron, X represents the training feature vector, and L represents the number of layers of the neural network.
[0039] Preferably, the evaluation feedback module includes a fitting unit and a feedback unit;
[0040] The fitting unit predicts the energy consumption of the time obtained by training Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption Perform fitting to obtain the model optimization energy consumption fitting index E opt (t);
[0041] The feedback unit synchronously matches the preset energy consumption optimization evaluation threshold value P, obtains the power equipment energy optimization evaluation scheme, and iteratively optimizes the training parameters of the model according to the content of the power equipment energy optimization evaluation scheme. When the content of the power equipment energy optimization evaluation scheme is the result of the power equipment energy optimization evaluation unqualified scheme, the operating parameters of the power equipment are dynamically adjusted, and the model optimization energy consumption fitting index E is calculated again. opt (t), the iteration ends when the result of the power equipment energy optimization evaluation plan is a qualified result of the power equipment energy optimization evaluation plan;
[0042] The power equipment energy optimization evaluation plan is obtained through the following matching method:
[0043] Model optimization energy consumption fitting index Eopt (t)≤energy consumption optimization evaluation threshold P, obtain the qualified solution result of power equipment energy optimization evaluation, and currently do not adjust the power equipment energy optimization strategy;
[0044] Model optimization energy consumption fitting index E opt (t)>energy consumption optimization evaluation threshold P, obtain the unqualified solution results of the power equipment energy optimization evaluation, and adjust the power equipment energy optimization strategy, including the power equipment operating power and operating parameters.
[0045] Preferably, the model optimizes the energy consumption fitting index E opt (t) is obtained by the following calculation formula:
[0046]
[0047] Where E opt (t) represents the model optimization energy consumption fitting index, specifically the equipment energy consumption optimization result at time t, represents the time prediction energy consumption, E u (t) represents the regression energy consumption, E rf (t) represents the random predicted energy consumption, represents the network predicted energy consumption, ω1, ω2, ω3 and ω4 represent the time predicted energy consumption respectively. Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption The weight value of .
[0048] A building energy management method based on cloud services comprises the following steps:
[0049] Step 1: The data acquisition module collects human flow information, environmental information and equipment energy consumption information through several sensors integrated in the power equipment to form a collection data group;
[0050] Step 2: The transmission module marks the collected data group, compresses and packages it, and transmits the packaged collected data group to the cloud server via the Internet;
[0051] Step 3: The data analysis module pre-processes the uploaded collection data group, including data cleaning cloud service, data filtering, data deduplication and data standardization, to form a processed cloud storage data group, which is synchronously stored in the cloud collection history database;
[0052] Step 4: The model building module uses big data analysis technology and machine learning algorithms to build time series models, regression models, random forest models and neural network models for the data in the cloud-collected historical database, and simultaneously trains and analyzes to obtain time-predicted energy consumption. Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption
[0053] Step 5: Evaluate the feedback module to predict the energy consumption of training time Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption Perform fitting to obtain the model optimization energy consumption fitting index E opt (t), synchronously match with the preset energy consumption optimization evaluation threshold P, obtain the power equipment energy optimization evaluation plan, and perform iterative training according to the content of the power equipment energy optimization evaluation plan.
[0054] The present invention provides a cloud-based building energy management system and method, which has the following features:
[0055] Beneficial effects:
[0056] (1) When the system is running, the integrated sensors collect information about human flow, environment, and equipment energy consumption in real time to ensure the comprehensiveness and accuracy of the data. The data is compressed and packaged and then transmitted to the cloud server to ensure the efficiency and security of data transmission. The uploaded data is preprocessed, and the time series model, regression model, random forest model, and neural network model are established and trained using big data analysis technology and machine learning algorithms to obtain high-precision energy consumption predictions. The evaluation feedback module fits the prediction results of each model to generate the model optimization energy consumption fitting index E. opt (t), and matches it with the preset energy consumption optimization evaluation threshold P, and finally generates an energy optimization evaluation plan for power equipment, and performs iterative training. Through this multi-model, multi-level optimization method, the limitations of the traditional single model in processing multi-dimensional data and nonlinear relationships are overcome, and the dynamic adjustment of equipment operation mode is realized, which significantly improves energy efficiency and reduces energy consumption.
[0057] (2) Time series model training to obtain time prediction energy consumption Obtain the regression energy consumption E through regression model training u (t), obtain the random predicted energy consumption E through the training of random forest model rf (t), obtain the network predicted energy consumption through the training of the neural network model This multi-model, multi-level optimization approach can make full use of historical data and real-time data to achieve accurate prediction and dynamic adjustment of the operating mode of electric equipment. It has unique advantages in addressing the shortcomings of traditional methods that cannot comprehensively consider multi-dimensional data and complex nonlinear relationships. It can not only provide more scientific optimization strategies, but also ensure energy efficiency while maintaining user comfort and equipment stability.
[0058] (3) Optimize the energy consumption fitting index E through the model opt (t) Matches with the preset energy consumption optimization evaluation threshold P to obtain the energy optimization evaluation plan for the power equipment, and iteratively optimizes the training parameters of the model according to the content of the plan. When the evaluation result is unqualified, the system automatically and dynamically adjusts the operating parameters of the power equipment until the evaluation result is qualified. Through this closed-loop feedback mechanism, not only can the prediction accuracy of the model be continuously optimized, but also the equipment operation strategy can be adjusted in real time, thereby achieving optimal management of energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS
[0059] Figure 1 This is a schematic diagram of a building energy management system based on cloud services in the present invention;
[0060] Figure 2 This is a schematic diagram of the steps of a building energy management method based on cloud services of the present invention. DETAILED DESCRIPTION
[0061] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0062] Example 1
[0063] The present invention provides a building energy management system based on cloud services. Figure 1 , including data acquisition module, transmission module, data analysis module, model building module and evaluation feedback module;
[0064] The data acquisition module collects human flow information, environmental information and equipment energy consumption information through several sensors integrated in the power equipment to form a collection data group;
[0065] The transmission module marks the collected data group, compresses and packages it, and transmits the packaged collected data group to the cloud server via the Internet;
[0066] The data analysis module pre-processes the uploaded collection data group, including data cleaning, data filtering, data deduplication and data standardization, to form a processed cloud storage data group, which is synchronously stored in the cloud collection history database;
[0067] The model building module uses big data analysis technology and machine learning algorithms to build time series models, regression models, random forest models and neural network models based on the data in the cloud-collected historical database, and simultaneously trains and analyzes to obtain time-predicted energy consumption. Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption
[0068] The evaluation feedback module predicts the energy consumption of the time obtained by training Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption Perform fitting to obtain the model optimization energy consumption fitting index E opt (t), synchronously match with the preset energy consumption optimization evaluation threshold P, obtain the power equipment energy optimization evaluation plan, and perform iterative training according to the content of the power equipment energy optimization evaluation plan.
[0069] In this embodiment, a comprehensive optimization of the use mode of power equipment in the building is achieved through a multi-level structure including a data acquisition module, a transmission module, a data analysis module, a model building module and an evaluation feedback module. First, the data acquisition module collects human flow information, environmental information and equipment energy consumption information in real time through integrated sensors to ensure the comprehensiveness and accuracy of the data. The transmission module marks, compresses and packages the collected data group and transmits it to the cloud server to ensure the efficiency and security of data transmission. The data analysis module pre-processes the uploaded data, including data cleaning, filtering, deduplication and standardization to ensure data quality. The model building module uses big data analysis technology and machine learning algorithms to establish and train time series models, regression models, random forest models and neural network models to obtain high-precision energy consumption predictions. The evaluation feedback module fits the prediction results of each model to generate a model optimization energy consumption fitting index E. opt (t), and matches it with the preset energy consumption optimization evaluation threshold P, and finally generates an energy optimization evaluation plan for power equipment, and performs iterative training. Through this multi-model, multi-level optimization method, the limitations of the traditional single model in processing multi-dimensional data and nonlinear relationships are overcome, and the dynamic adjustment of equipment operation mode is realized, which significantly improves energy efficiency and reduces energy consumption.
[0070] Example 2
[0071] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the data acquisition module includes an extraction unit;
[0072] The extraction unit collects information through several sensors integrated in the power equipment, and simultaneously extracts the flow of people information with fixed period content, including the number and density of people, extracts the internal and external environment information of the building, including the indoor and outdoor temperature and the indoor and outdoor light intensity, and extracts the energy consumption information of the working equipment, including the equipment start and stop cycle and the energy consumption, to form a collection data group;
[0073] The sensors include infrared sensors, cameras, temperature sensors, humidity sensors, light sensors, current sensors and voltage sensors;
[0074] The collected data group includes human flow information, environmental information, equipment energy consumption information, electricity price and equipment information.
[0075] The transmission module includes a communication unit;
[0076] The communication unit marks the collected data group, including adding a timestamp, device ID, device location and data type, and then compresses and packages it using a compression algorithm, and simultaneously performs block packaging according to the device ID, generates a compressed package to form a compressed packaged collected data group, and transmits the packaged collected data group to the cloud server via the Internet;
[0077] Among them, compression algorithms include GZIP compression algorithm and LZ77 compression algorithm; packaging formats include JSON format and Protobuf format.
[0078] The data analysis module includes a preprocessing unit and a storage unit;
[0079] The preprocessing unit preprocesses the uploaded collected data group, including data cleaning cloud service, data filtering, data deduplication and data standardization, to form a processed cloud storage data group;
[0080] Among them, data cleaning includes the use of interpolation, mean filling and statistical analysis methods to handle missing values and detect outliers; data filtering includes the use of filtering algorithms to filter noise; data deduplication includes the use of hash algorithms to detect and process deduplication; data standardization includes normalization and standardization to unify data formats;
[0081] The storage unit synchronously stores the processed cloud storage data group in the cloud acquisition history database, and performs data archiving processing and data type partitioning storage.
[0082] Example 3
[0083] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the model building module includes a modeling unit and a training unit;
[0084] The modeling unit uses big data analysis technology and machine learning algorithms to establish a training analysis model for the data in the cloud collection history database, including using a time series analysis method to establish a time series model, using a linear regression method to establish a regression model, using a random forest algorithm to establish a random forest model, and using a neural network algorithm to establish a neural network model;
[0085] The training unit collects data stored in the historical database through the cloud and performs fixed-period interception to obtain the training feature vector X, inputs the training feature vector X into the established time series model, regression model, random forest model and neural network model machine for training, and obtains the time prediction energy consumption through the training of the time series model Obtain the regression energy consumption E through regression model training u (t), obtain the random predicted energy consumption E through the training of random forest model rf (t), obtain the network predicted energy consumption through the training of the neural network model
[0086] Among them, the training feature vector X includes {X p , X e , X c and X d}, where X p represents the subset of crowd flow information, X e represents the environmental information subset, X c Equipment energy consumption information subset, X d A subset of electricity price and equipment information, which includes stored related collection information;
[0087] Among them, the flow information subset X p Including crowd density G p , density gradient of human flow G' p and flow direction D p ;
[0088] Environmental Information Subset X e Including indoor ambient temperature T sn , Outdoor ambient temperature T sw , Indoor humidity H sn , Outdoor humidity H sw , air velocity change rate L bh , Indoor light intensity L sn , Outdoor light intensity L sw and the rate of change of illumination L u ;
[0089] Equipment energy consumption information subset X c Including real-time energy consumption E r , Equipment use cycle times G u 、Equipment life cycle C u and equipment standby energy consumption E s ;
[0090] Electricity price and equipment information subset X d Including electricity price fluctuation index V e 、Device Type T e and equipment operating status S e .
[0091] The time predicted energy consumption Obtained through the following calculation formula:
[0092]
[0093] In the formula, Time predicted energy consumption, specifically represents the predicted energy consumption at time t, represents the constant term, specifically the intercept of the model, p represents the natural regression order, represents the natural regression coefficient, specifically reflecting the impact of energy consumption at time m on current energy consumption, E(t+i) represents the time energy consumption at time ti, q represents the moving average order, θ j Represents the moving average coefficient, specifically reflecting the impact of the white noise at the previous j moments on the current energy consumption, ∈(tj) represents the white noise at time tj, and ∈(t) represents the white noise at the current time point t.
[0094] The regression uses energy consumption E u (t) is obtained by the following calculation formula:
[0095] E u (t) = α*G p (t)+β*G u (t)+χ*L u (t)+δ*R e (t);
[0096] In the formula, E u (t) Regression energy consumption, specifically represents the energy consumption of the device at time point t, G p (t) represents the density gradient of the flow of people at time t, G u (t) represents the number of equipment use cycles at time t, L u (t) represents the indoor light change rate at time t, R e (t) represents the environmental change rate at time point t, specifically including humidity rate, temperature change rate and air flow rate change rate.
[0097] The random prediction energy consumption E rf (t) is obtained by the following calculation formula:
[0098]
[0099] In the formula, E rf (t) represents the energy consumption predicted by the random forest at time t, N represents the number of decision trees, T k (X) represents the output of the k-th decision tree, and X represents the training feature vector;
[0100] The network predicts energy consumption Obtained through the following calculation formula:
[0101]
[0102] In the formula, represents the network prediction energy consumption, specifically the neural network prediction energy consumption at time t, σ represents the activation function, specifically including ReLU, Sigmoid and Tanh nonlinear functions, to introduce nonlinear features, W (L) represents the weight matrix of the Lth layer, specifically the connection weights between the neurons in the Lth layer and the L-1th layer, b (L) It represents the bias vector of the Lth layer, specifically the bias term of the Lth layer neuron, X represents the training feature vector, and L represents the number of layers of the neural network.
[0103] In this embodiment, the time prediction energy consumption is obtained by training the time series model. Obtain the regression energy consumption E through regression model training u (t), obtain the random predicted energy consumption E through the training of random forest model rf (t), obtain the network predicted energy consumption through the training of the neural network model This multi-model, multi-level optimization approach can make full use of historical data and real-time data to achieve accurate prediction and dynamic adjustment of the operating mode of electric equipment. It has unique advantages in addressing the shortcomings of traditional methods that cannot comprehensively consider multi-dimensional data and complex nonlinear relationships. It can not only provide more scientific optimization strategies, but also ensure energy efficiency while maintaining user comfort and equipment stability.
[0104] Example 4
[0105] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the evaluation feedback module includes a fitting unit and a feedback unit;
[0106] The fitting unit predicts the energy consumption of the time obtained by training Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption Perform fitting to obtain the model optimization energy consumption fitting index E opt (t);
[0107] The feedback unit synchronously matches the preset energy consumption optimization evaluation threshold value P, obtains the power equipment energy optimization evaluation scheme, and iteratively optimizes the training parameters of the model according to the content of the power equipment energy optimization evaluation scheme. When the content of the power equipment energy optimization evaluation scheme is the result of the power equipment energy optimization evaluation unqualified scheme, the operating parameters of the power equipment are dynamically adjusted, and the model optimization energy consumption fitting index E is calculated again. opt (t), the iteration ends when the result of the power equipment energy optimization evaluation plan is a qualified result of the power equipment energy optimization evaluation plan;
[0108] The power equipment energy optimization evaluation plan is obtained through the following matching method:
[0109] Model optimization energy consumption fitting index E opt (t)≤energy consumption optimization evaluation threshold P, obtain the qualified solution result of power equipment energy optimization evaluation, and currently do not adjust the power equipment energy optimization strategy;
[0110] Model optimization energy consumption fitting index E opt (t)>energy consumption optimization evaluation threshold P, obtain the unqualified solution results of the power equipment energy optimization evaluation, and adjust the power equipment energy optimization strategy, including the power equipment operating power and operating parameters.
[0111] The model optimizes the energy consumption fitting index E opt (t) is obtained by the following calculation formula:
[0112]
[0113] Where E opt (t) represents the model optimization energy consumption fitting index, specifically the equipment energy consumption optimization result at time t, represents the time prediction energy consumption, E u (t) represents the regression energy consumption, E rf (t) represents the random predicted energy consumption, represents the network predicted energy consumption, ω1, ω2, ω3 and ω4 represent the time predicted energy consumption respectively. Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption The weight value of .
[0114] In this embodiment, the energy consumption fitting index E is optimized by the model. opt (t) Matches with the preset energy consumption optimization evaluation threshold P to obtain the energy optimization evaluation plan for the power equipment, and iteratively optimizes the training parameters of the model according to the content of the plan. When the evaluation result is unqualified, the system automatically and dynamically adjusts the operating parameters of the power equipment until the evaluation result is qualified. Through this closed-loop feedback mechanism, not only can the prediction accuracy of the model be continuously optimized, but also the equipment operation strategy can be adjusted in real time, thereby achieving optimal management of energy consumption.
[0115] Example 5
[0116] A building energy management method based on cloud services, please refer to Figure 2 , specifically: including the following steps:
[0117] Step 1: The data acquisition module collects human flow information, environmental information and equipment energy consumption information through several sensors integrated in the power equipment to form a collection data group;
[0118] Step 2: The transmission module marks the collected data group, compresses and packages it, and transmits the packaged collected data group to the cloud server via the Internet;
[0119] Step 3: The data analysis module pre-processes the uploaded collection data group, including data cleaning cloud service, data filtering, data deduplication and data standardization, to form a processed cloud storage data group, which is synchronously stored in the cloud collection history database;
[0120] Step 4: The model building module uses big data analysis technology and machine learning algorithms to build time series models, regression models, random forest models and neural network models for the data in the cloud-collected historical database, and simultaneously trains and analyzes to obtain time-predicted energy consumption. Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption
[0121] Step 5: Evaluate the feedback module to predict the energy consumption of training time Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption Perform fitting to obtain the model optimization energy consumption fitting index E opt (t), synchronously match with the preset energy consumption optimization evaluation threshold P, obtain the power equipment energy optimization evaluation plan, and perform iterative training according to the content of the power equipment energy optimization evaluation plan.
[0122] Although 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 appended claims and their equivalents.
Claims
1. A building energy management system based on cloud services, characterized in that: It includes data acquisition module, transmission module, data analysis module, model building module and evaluation feedback module; The data acquisition module collects human flow information, environmental information and equipment energy consumption information through several sensors integrated in the power equipment to form a collection data group; The transmission module marks the collected data group, compresses and packages it, and transmits the packaged collected data group to the cloud server via the Internet; The data analysis module pre-processes the uploaded collection data group, including data cleaning, data filtering, data deduplication and data standardization, to form a processed cloud storage data group, which is synchronously stored in the cloud collection history database; The model building module uses big data analysis technology and machine learning algorithms to build time series models, regression models, random forest models and neural network models based on the data in the cloud-collected historical database, and simultaneously trains and analyzes to obtain time-predicted energy consumption. Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption The evaluation feedback module predicts the energy consumption of the time obtained by training Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption Perform fitting to obtain the model optimization energy consumption fitting index E opt (t), synchronously match with the preset energy consumption optimization evaluation threshold P, obtain the power equipment energy optimization evaluation plan, and perform iterative training according to the content of the power equipment energy optimization evaluation plan.
2. The cloud-based building energy management system according to claim 1, characterized in that: The data acquisition module includes an extraction unit; The extraction unit collects information through several sensors integrated in the power equipment, and simultaneously extracts the flow of people information with fixed period content, including the number and density of people, extracts the internal and external environment information of the building, including the indoor and outdoor temperature and the indoor and outdoor light intensity, and extracts the energy consumption information of the working equipment, including the equipment start and stop cycle and the energy consumption, to form a collection data group; The sensors include infrared sensors, cameras, temperature sensors, humidity sensors, light sensors, current sensors and voltage sensors; The collected data group includes human flow information, environmental information, equipment energy consumption information, electricity price and equipment information.
3. The cloud-based building energy management system according to claim 1, characterized in that: The transmission module includes a communication unit; The communication unit marks the collected data group, including adding a timestamp, device ID, device location and data type, and then compresses and packages it using a compression algorithm, and simultaneously performs block packaging according to the device ID, generates a compressed package to form a compressed packaged collected data group, and transmits the packaged collected data group to the cloud server via the Internet; Among them, compression algorithms include GZIP compression algorithm and LZ77 compression algorithm; packaging formats include JSON format and Protobuf format.
4. The cloud-based building energy management system according to claim 1, characterized in that: The data analysis module includes a preprocessing unit and a storage unit; The preprocessing unit preprocesses the uploaded collected data group, including data cleaning cloud service, data filtering, data deduplication and data standardization, to form a processed cloud storage data group; Among them, data cleaning includes the use of interpolation, mean filling and statistical analysis methods to handle missing values and detect outliers; data filtering includes the use of filtering algorithms to filter noise; data deduplication includes the use of hash algorithms to detect and process deduplication; data standardization includes normalization and standardization to unify data formats; The storage unit synchronously stores the processed cloud storage data group in the cloud acquisition history database, and performs data archiving processing and data type partitioning storage.
5. The cloud-based building energy management system according to claim 4, characterized in that: The model building module includes a modeling unit and a training unit; The modeling unit uses big data analysis technology and machine learning algorithms to establish a training analysis model for the data in the cloud collection history database, including using a time series analysis method to establish a time series model, using a linear regression method to establish a regression model, using a random forest algorithm to establish a random forest model, and using a neural network algorithm to establish a neural network model; The training unit collects data stored in the historical database through the cloud and performs fixed-period interception to obtain the training feature vector X, inputs the training feature vector X into the established time series model, regression model, random forest model and neural network model machine for training, and obtains the time prediction energy consumption through the training of the time series model Obtain the regression energy consumption E through regression model training u (t), obtain the random predicted energy consumption E through the training of random forest model rf (t), obtain the network predicted energy consumption through the training of the neural network model Among them, the training feature vector X includes {X p , X e , X c and X d }, where X p represents the subset of crowd flow information, X e represents the environmental information subset, X c Equipment energy consumption information subset, X d A subset of electricity price and equipment information, which includes stored related collection information.
6. The cloud-based building energy management system according to claim 5, characterized in that: The time predicted energy consumption Obtained through the following calculation formula: In the formula, Time predicted energy consumption, specifically represents the predicted energy consumption at time t, represents the constant term, specifically the intercept of the model, p represents the natural regression order, represents the natural regression coefficient, specifically reflecting the impact of energy consumption at time m on current energy consumption, E(t+i) represents the time energy consumption at time ti, q represents the moving average order, θ j Represents the moving average coefficient, specifically reflecting the impact of the white noise at the previous j moments on the current energy consumption, ∈(tj) represents the white noise at time tj, and ∈(t) represents the white noise at the current time point t. The regression uses energy consumption E u (t) is obtained by the following calculation formula: E u (t)=α*G p (t)+β*G u (t)+x*L u (t)+δ*R e (t); In the formula, E u (t) Regression energy consumption, specifically represents the energy consumption of the device at time point t, G p (t) represents the density gradient of the flow of people at time t, G u (t) represents the number of equipment use cycles at time t, L u (t) represents the indoor light change rate at time t, R e (t) represents the environmental change rate at time point t, specifically including humidity rate, temperature change rate and air flow rate change rate.
7. The cloud-based building energy management system according to claim 5, characterized in that: The random prediction energy consumption E rf (t) is obtained by the following calculation formula: In the formula, E rf (t) represents the energy consumption predicted by the random forest at time t, N represents the number of decision trees, T k (X) represents the output of the k-th decision tree, and X represents the training feature vector; The network predicts energy consumption Obtained through the following calculation formula: In the formula, Represents the network prediction energy consumption, specifically the neural network prediction energy consumption at time t, σ represents the activation function, specifically including ReLU, Sigmoid and Tanh nonlinear functions, to introduce nonlinear features, W (L) represents the weight matrix of the Lth layer, specifically the connection weights between the neurons in the Lth layer and the L-1th layer, b (L) It represents the bias vector of the Lth layer, specifically the bias term of the Lth layer neuron, X represents the training feature vector, and L represents the number of layers of the neural network.
8. The cloud-based building energy management system according to claim 1, characterized in that: The evaluation feedback module includes a fitting unit and a feedback unit; The fitting unit predicts the energy consumption of the time obtained by training Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption Perform fitting to obtain the model optimization energy consumption fitting index E opt (t); The feedback unit synchronously matches the preset energy consumption optimization evaluation threshold value P, obtains the power equipment energy optimization evaluation scheme, and iteratively optimizes the training parameters of the model according to the content of the power equipment energy optimization evaluation scheme. When the content of the power equipment energy optimization evaluation scheme is the result of the power equipment energy optimization evaluation unqualified scheme, the operating parameters of the power equipment are dynamically adjusted, and the model optimization energy consumption fitting index E is calculated again. opt (t), the iteration ends when the result of the power equipment energy optimization evaluation plan is a qualified result of the power equipment energy optimization evaluation plan; The power equipment energy optimization evaluation plan is obtained through the following matching method: Model optimization energy consumption fitting index E opt (t)≤energy consumption optimization evaluation threshold P, obtain the qualified solution result of power equipment energy optimization evaluation, and currently do not adjust the power equipment energy optimization strategy; Model optimization energy consumption fitting index E opt (t)>energy consumption optimization evaluation threshold P, obtain the unqualified solution results of the power equipment energy optimization evaluation, and adjust the power equipment energy optimization strategy, including the power equipment operating power and operating parameters.
9. The cloud-based building energy management system according to claim 8, characterized in that: The model optimizes the energy consumption fitting index E opt (t) is obtained by the following calculation formula: Where E opt (t) represents the model optimization energy consumption fitting index, specifically the equipment energy consumption optimization result at time t, represents the time prediction energy consumption, E u (t) represents the regression energy consumption, E rf (t) represents the random predicted energy consumption, represents the network predicted energy consumption, ω1, ω2, ω3 and ω4 represent the time predicted energy consumption respectively. Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption The weight value of .
10. A cloud-based building energy management method, comprising the cloud-based building energy management system according to any one of claims 1 to 9, characterized in that: The following steps are involved: Step 1: The data acquisition module collects human flow information, environmental information and equipment energy consumption information through several sensors integrated in the power equipment to form a collection data group; Step 2: The transmission module marks the collected data group, compresses and packages it, and transmits the packaged collected data group to the cloud server via the Internet; Step 3: The data analysis module pre-processes the uploaded collection data group, including data cleaning cloud service, data filtering, data deduplication and data standardization, to form a processed cloud storage data group, which is synchronously stored in the cloud collection history database; Step 4: The model building module uses big data analysis technology and machine learning algorithms to build time series models, regression models, random forest models and neural network models for the data in the cloud-collected historical database, and simultaneously trains and analyzes to obtain time-predicted energy consumption. Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption Step 5: Evaluate the feedback module to predict the energy consumption of training time Regression energy consumption E u (t), random prediction of energy consumption E rf (t) and network predicted energy consumption Perform fitting to obtain the model optimization energy consumption fitting index E opt (t), synchronously match with the preset energy consumption optimization evaluation threshold P, obtain the power equipment energy optimization evaluation plan, and perform iterative training according to the content of the power equipment energy optimization evaluation plan.
Citation Information
Patent Citations
Building near-zero energy consumption controller based on renewable energy big data deep learning
CN107730006A
Visual prediction method and system for building energy consumption
CN108830413A
Building energy consumption processing method and device based on RF and ARMA algorithms
CN109035067A
Building energy consumption prediction model construction method, energy consumption prediction method and device
CN112766596A
Building energy consumption prediction method and system
CN116738856A
