An intelligent energy consumption detection method and management system based on Internet of Things technology

Through IoT technology and neural network model, an intelligent energy consumption detection system is built, which solves the problems of sensor data accuracy error and environmental interference, and realizes accurate monitoring and management of energy consumption, providing a basis for energy-saving transformation.

CN118149901BActive Publication Date: 2024-11-05CHINA COAL INFORMATION TECH (BEIJING) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410335909.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-03-22
Publication Date
2024-11-05
Estimated Expiration
2044-03-22

AI Technical Summary

Technical Problem

The existing energy consumption monitoring system has large accuracy errors and interference due to environmental factors, resulting in inaccurate detection data and lack of effective management.

Method used

Using intelligent energy consumption detection method based on IoT technology, the temperature, humidity, flow, power and other data are collected through ZigBee and WiFi wireless transmission technology, and the energy consumption monitoring and transmission Internet of Things network is built, and the cloud server platform is used for big data management and analysis. Combining SSA-optimized TCN time convolution network model, ESN neural network model and BiGRU neural network model, an intelligent energy consumption detection subsystem is built.

Benefits of technology

It improves the accuracy and stability of energy consumption measurement, reduces sensor accuracy errors and interference, and realizes accurate monitoring and management of energy consumption, providing a basis for energy-saving transformation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118149901B_ABST
    Figure CN118149901B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of energy consumption intelligent monitoring method and automation equipment technology based on Internet of Things technology, and discloses an energy consumption intelligent detection method and management system based on Internet of Things technology, constructs an energy consumption information collection module and an energy consumption intelligent detection subsystem, collects energy consumption data through the energy consumption information collection module, and the energy consumption intelligent detection subsystem takes multiple energy consumption sensor outputs, multiple temperature sensor outputs, multiple humidity sensor outputs and multiple energy consumption historical parameter values ​​as inputs, and outputs the energy consumption value of the detected object. Compared with the prior art, the present invention uses the triangular fuzzy number form output by the energy consumption information module to represent the outputs of multiple energy consumption sensors, thereby improving the objectivity and credibility of the sensor values ​​for measuring energy consumption, introducing ZigBee and WiFi wireless transmission technologies to establish a data transmission network, and using a cloud server platform to aggregate data and analyze energy efficiency, so as to clearly control the energy consumption of the monitored object and provide a basis for energy-saving transformation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of energy consumption intelligent monitoring method and automation equipment technology of Internet of Things technology, and specifically to an energy consumption intelligent detection method based on Internet of Things technology and a management system thereof. Background Art

[0002] As the process of industrialization and modernization steadily advances, the problems of rational use of resources and energy that follow also constantly affect the development of the economy and society. In the context of energy conservation and emission reduction, it is of great significance to effectively supervise the energy efficiency of industrial and agricultural production processes, public facilities and buildings, which account for a large proportion of energy consumption. Energy-consuming units need to improve energy utilization efficiency and control the total amount of energy consumption. Effective energy consumption monitoring is an important and indispensable part of achieving energy conservation. It is an urgent need to equip energy consumption supervision systems to control energy consumption.

[0003] Existing energy consumption monitoring systems mostly rely solely on sensor data. However, sensor data will be affected by environmental factors during the energy consumption monitoring process. The uncertainty and randomness of energy consumption sensor accuracy errors, interference, and abnormal measurement temperature values ​​during energy consumption measurement lead to inaccurate detection data and a lack of management for energy consumption monitoring.

[0004] In view of the low status of energy consumption monitoring and management, the present invention designs an intelligent energy consumption monitoring method and management system based on Internet of Things technology. The system introduces ZigBee and WiFi wireless transmission technologies, collects big data on consumption such as temperature, humidity, flow, and power of the monitored objects, establishes an energy consumption monitoring and transmission Internet of Things network, manages big energy consumption data with the help of a cloud server platform, analyzes and processes energy efficiency, controls energy consumption, and provides a basis for energy-saving transformation. Summary of the invention

[0005] Purpose of the invention: In view of the current status of energy consumption monitoring, the present invention invents an intelligent energy consumption detection method and management system based on the Internet of Things technology, collects temperature, humidity, flow, electricity and wind speed consumption data, constructs an intelligent energy consumption detection subsystem, introduces ZigBee and WiFi wireless transmission technologies to establish a data transmission network, aggregates data with the help of a cloud server platform, analyzes energy efficiency, thereby clearly controlling the energy consumption of the monitored object and providing a basis for energy-saving transformation.

[0006] Technical solution: The present invention discloses an intelligent energy consumption detection method based on Internet of Things technology, comprising the following steps:

[0007] Step 1: Construct an energy consumption information collection module, which includes a fuzzy wavelet neural network model of SSA, an energy consumption information SOM clusterer, a TCN temporal convolutional network model of SSA - an ARIMA model, a BiGRU neural network model of SSA, an ESN neural network model of triangular fuzzy numbers - a fuzzy wavelet neural network model 1 of SSA and a TDL beat delay device;

[0008] Step 2: construct an energy consumption intelligent detection subsystem, which includes multiple energy consumption information collection modules, an ESN neural network model-ARIMA model, a TCN temporal convolutional network model-ARIMA model of SSA, a BiGRU neural network model-ARIMA model of SSA, a BiGRU neural network model-ESN neural network model of SSA, a TCN temporal convolutional network model-fuzzy wavelet neural network model of SSA of triangular fuzzy numbers, a BiGRU neural network model-fuzzy wavelet neural network model of SSA of triangular fuzzy numbers, and an ESN neural network model-fuzzy wavelet neural network model of SSA of triangular fuzzy numbers 2;

[0009] Step 3: Multiple energy consumption sensor outputs, multiple temperature sensor outputs, multiple humidity sensor outputs and multiple energy consumption historical parameter values ​​are respectively used as inputs of the corresponding energy consumption information acquisition module. The energy consumption information acquisition module takes the output data of multiple parameter sensors as input. The triangular fuzzy values ​​of the outputs of multiple parameter sensors are composed of a, b and c output by the fuzzy wavelet neural network model 1 of the triangular fuzzy number ESN neural network model-SSA of the energy consumption information acquisition module, which is [a, b, c]; the triangular fuzzy numbers (X, Y, Z) composed of X, Y and Z output by the fuzzy wavelet neural network model 2 of the triangular fuzzy number ESN neural network model-SSA of the energy consumption intelligent detection subsystem are used as the energy consumption value of the detected object.

[0010] Furthermore, the architecture of the energy consumption information collection module is as follows:

[0011] The outputs of multiple parameter sensors are used as the corresponding inputs of multiple SSA fuzzy wavelet neural network models respectively, the output values ​​of multiple SSA fuzzy wavelet neural network models are used as the input of the energy consumption information SOM clusterer, and the output values ​​of the three types of SSA fuzzy wavelet neural network models output by the energy consumption information SOM clusterer are used as SSA's TCN temporal convolutional network model-ARIMA model 1 and SSA's BiGRU neural network model 1, SSA's TCN temporal convolutional network model-ARIMA model 2 and SSA's BiGRU neural network model 2, and SSA's TCN temporal convolutional network model-ARI The corresponding inputs of MA model 3 and SSA’s BiGRU neural network model 3, the outputs of SSA’s TCN temporal convolutional network model-ARIMA model 1 are respectively used as the corresponding inputs of SSA’s BiGRU neural network model 1 and the ESN neural network model of triangular fuzzy numbers-SSA’s fuzzy wavelet neural network model 1, the outputs of SSA’s TCN temporal convolutional network model-ARIMA model 2 are respectively used as the corresponding inputs of SSA’s BiGRU neural network model 2 and the ESN neural network model of triangular fuzzy numbers-SSA’s fuzzy wavelet neural network model 1, the TCN temporal convolutional network model of SSA -The output of ARIMA model 3 is used as the corresponding input of BiGRU neural network model 3 of SSA and ESN neural network model of triangular fuzzy number -The output of BiGRU neural network model 1 of SSA, the output of BiGRU neural network model 1 of SSA, the output of BiGRU neural network model 2 of SSA and the output of BiGRU neural network model 3 of SSA are used as the input of TDL beat delay 1, TDL beat delay 2 and TDL beat delay 3 respectively, the output of TDL beat delay 1, TDL beat delay 2 and TDL beat delay 3 are used as the ESN neural network of SSA of triangular fuzzy Model - the corresponding input of fuzzy wavelet neural network model 1, the output of ESN neural network model of triangular fuzzy number - SSA fuzzy wavelet neural network model 1 is used as the input of TDL beat delay device 4, the output of TDL beat delay device 4 is used as the corresponding input of SSA BiGRU neural network model 1, SSA BiGRU neural network model 2 and SSA BiGRU neural network model 3, the a, b and c output of ESN neural network model of triangular fuzzy number - SSA fuzzy wavelet neural network model 1 constitute the triangular fuzzy value output of multiple parameter sensors is [a, b, c], the energy consumption information acquisition module is shown in Figure 1 shown.

[0012] Furthermore, the connection structure of the energy consumption intelligent detection subsystem is as follows:

[0013] The outputs of multiple energy consumption sensors, multiple temperature sensors, and multiple humidity sensors are used as inputs of energy consumption information collection module 1, energy consumption information collection module 2, and energy consumption information collection module 3, respectively. The historical parameter values ​​of multiple energy consumption are used as inputs of energy consumption information collection module 4 and the corresponding inputs of ESN neural network model-ARIMA model, and the output of energy consumption information collection module 4 is used as the corresponding input of ESN neural network model-ARIMA model. The outputs of energy consumption information collection module 1, energy consumption information collection module 2, energy consumption information collection module 3, and ESN neural network model-ARIMA model are used as the BiGRU neural network model of SSA-ESN neural network model. The corresponding inputs of the network model, the SSA TCN temporal convolutional network model-ARIMA model, and the SSA BiGRU neural network model-ARIMA model, the outputs of the SSA BiGRU neural network model-ESN neural network model, the SSA TCN temporal convolutional network model-ARIMA model, the SSA BiGRU neural network model-ARIMA model, and the ESN neural network model-ARIMA model are respectively used as the SSA TCN temporal convolutional network model-fuzzy wavelet neural network model of triangular fuzzy numbers and the SSA BiGRU neural network model-fuzzy wavelet neural network model of triangular fuzzy numbers. The input, the output of the TCN temporal convolutional network model of SSA for triangular fuzzy numbers - fuzzy wavelet neural network model and the output of the BiGRU neural network model of SSA for triangular fuzzy numbers - fuzzy wavelet neural network model are used as the corresponding inputs of the BiGRU neural network model of SSA - ESN neural network model, the TCN temporal convolutional network model of SSA - ARIMA model, the BiGRU neural network model of SSA - ARIMA model and the ESN neural network model of SSA for triangular fuzzy numbers - fuzzy wavelet neural network model 2, and the output of the ESN neural network model - ARIMA model is used as the ESN neural network model - ESN neural network model for triangular fuzzy numbers. The corresponding input of SSA's fuzzy wavelet neural network model 2 and the output of triangular fuzzy number ESN neural network model-SSA's fuzzy wavelet neural network model 2 are used as the corresponding input of SSA's TCN temporal convolutional network model-fuzzy wavelet neural network model, SSA's BiGRU neural network model-fuzzy wavelet neural network model and ESN neural network model-ARIMA model respectively. The triangular fuzzy number (X, Y, Z) composed of X, Y and Z output by the fuzzy wavelet neural network model of the ESN neural network model-SSA of triangular fuzzy numbers is used as the energy consumption value of the detected object. The constructed energy consumption intelligent detection subsystem is shown in FIG. Figure 2 shown.

[0014] The present invention also discloses an intelligent energy consumption management system based on the Internet of Things technology, including an energy consumption data perception layer, a data transmission layer and a data application layer. The energy consumption data perception layer is composed of a large number of energy consumption perception terminals. The energy consumption perception terminals include energy consumption sensors, temperature sensors, and humidity sensors. The energy consumption information coordinator that collects temperature, humidity and energy consumption big data is aggregated to the data transmission layer. The energy consumption information coordinator receives the big data collected by the energy consumption perception terminal and uploads it to the data application layer. The data application layer is composed of an energy consumption OneNET cloud server, an energy consumption monitoring terminal and an energy consumption monitoring mobile phone APP. The energy consumption monitoring terminal and the energy consumption monitoring mobile phone APP send instructions to the energy consumption perception terminal through the energy consumption OneNET cloud server and the energy consumption information coordinator to control the external devices of the energy consumption perception terminal in real time. The energy consumption monitoring terminal and the energy consumption perception terminal, the energy consumption information coordinator and the energy consumption OneNET cloud server execute the steps of the above-mentioned intelligent energy consumption detection method based on the Internet of Things technology. The intelligent energy consumption management system based on the Internet of Things technology disclosed by the present invention is shown in FIG. Figure 3 shown.

[0015] Compared with the prior art, the present invention has the following advantages:

[0016] 1. The SSA TCN time convolution network model of the present invention is a SSA-optimized TCN time convolution network model. The parameters of the TCN time convolution network model are optimized by utilizing the characteristics of SSA such as faster convergence speed, better robustness and accuracy. The TCN time convolution network model can be made to adaptively select model parameters for specific energy consumption prediction problems, thereby reducing the workload of manual parameter debugging and improving the energy consumption prediction accuracy of the TCN time convolution network model.

[0017] 2. The ESN neural network model of the present invention designs the network hidden layer into a sparse network composed of many neurons, and achieves the function of memorizing energy consumption data by adjusting the characteristics of the internal weights of the network. Its internal dynamic reserve pool contains a large number of sparsely connected neurons, contains the operating status of the system, and has the function of short-term memory of energy consumption data. The stability of the recursive network inside the reserve pool is guaranteed by presetting the spectral radius of the internal connection weight matrix of the ESN neural network model, thereby improving the stability and accuracy of the predicted energy consumption parameters.

[0018] 3. The current state of the reserve pool of the ESN neural network model of the present invention inherits the energy consumption data state of the previous moment, and has a short-term memory characteristic for the energy consumption historical data input to the ESN neural network model. The research results show that the ESN neural network model with historical memory has a better energy consumption prediction effect. The ESN neural network has the characteristics of high precision, high accuracy, high timeliness and stability, and can be used as a means to quickly and effectively predict the energy consumption parameters input to the ESN neural network model; as a new type of dynamic recursive neural network, the ESN neural network model uses a linear regression method to establish an energy consumption prediction model, avoiding the problems of slow convergence speed and easy to fall into local minima of traditional neural networks, simplifying the complexity of the training process, and achieving the purpose of accurately predicting energy consumption parameters.

[0019] 4. The TCN temporal convolutional network model constructs a deep neural network through convolutional layers to achieve the memory function of the historical information of the input energy consumption sequence. The TCN temporal convolutional network model includes dilated convolution and residual connection. The dilated convolution is divided into two convolution modes: causal dilated convolution and non-causal dilated convolution. The causal dilated convolution can prevent the leakage of future energy consumption time series information and realize real-time energy consumption information prediction; while the use of non-causal dilated convolution can enable the TCN temporal convolutional network model to use all the input energy consumption information from the past to the future for prediction; the dilated convolution performs convolution on each layer of the TCN temporal convolutional network model The kernel is zero-padded so that the perception of the convolution kernel increases exponentially with the increase in the number of convolution layers. Therefore, more local energy consumption information can be obtained, thereby saving longer energy consumption history information. In addition, the TCN time convolution network model can also take energy consumption information sequences of any length as input and generate output sequences of the same length; the TCN time convolution network model contains a residual network structure, and its residual block allows the TCN time convolution network model to learn the residual errors of the feature mapping of each layer, avoiding the gradient vanishing problem of the deep network, so that the TCN time convolution network model still maintains stable performance in the deep structure.

[0020] 5. The present invention aims at the uncertainty and randomness of energy consumption sensor accuracy error, interference and abnormal measurement temperature value in the process of energy consumption sensor measuring energy consumption. The patent of the present invention represents the output of multiple energy consumption sensors in the form of triangular fuzzy numbers output by the energy consumption information module, which effectively handles the fuzziness and uncertainty of the measurement parameters of the energy consumption sensors and improves the objectivity and credibility of the measurement sensor values ​​of energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is the energy consumption information collection module of the present invention;

[0022] Figure 2 It is the energy consumption intelligent detection subsystem of the present invention;

[0023] Figure 3 This invention provides an intelligent energy consumption management system based on Internet of Things technology;

[0024] Figure 4 The energy consumption sensing terminal of the present invention;

[0025] Figure 5 The energy consumption information coordinator of the present invention;

[0026] Figure 6 This is the software function of the energy consumption monitoring terminal of the present invention. DETAILED DESCRIPTION

[0027] In order to better explain the present invention, the following Figure 1-6 The technical solution of the present invention is described in detail. The following embodiments are explanations of the invention, and the present invention is not limited to the following embodiments.

[0028] The present invention discloses an intelligent energy consumption detection method based on Internet of Things technology, comprising the following steps:

[0029] Step 1, construct an energy consumption information collection module, which includes SSA's fuzzy wavelet neural network model, energy consumption information SOM clusterer, SSA's TCN time convolution network model-ARIMA model, SSA's BiGRU neural network model, triangular fuzzy number ESN neural network model-SSA's fuzzy wavelet neural network model and TDL beat delay device.

[0030] The outputs of multiple parameter sensors are used as the corresponding inputs of multiple SSA fuzzy wavelet neural network models respectively, the output values ​​of multiple SSA fuzzy wavelet neural network models are used as the inputs of the energy consumption information SOM clusterer, and the output values ​​of the three types of SSA fuzzy wavelet neural network models output by the energy consumption information SOM clusterer are used as SSA's TCN temporal convolutional network model-ARIMA model 1 and SSA's BiGRU neural network model 1, SSA's TCN temporal convolutional network model-ARIMA model 2 and SSA's BiGRU neural network model 2, and SSA's TCN temporal convolutional network model-A The corresponding inputs of RIMA model 3 and SSA’s BiGRU neural network model 3, the outputs of SSA’s TCN temporal convolutional network model-ARIMA model 1 are respectively used as the corresponding inputs of SSA’s BiGRU neural network model 1 and the ESN neural network model of triangular fuzzy numbers-SSA’s fuzzy wavelet neural network model 1, the outputs of SSA’s TCN temporal convolutional network model-ARIMA model 2 are respectively used as the corresponding inputs of SSA’s BiGRU neural network model 2 and the ESN neural network model of triangular fuzzy numbers-SSA’s fuzzy wavelet neural network model 1, the TCN temporal convolution of SSA The outputs of network model-ARIMA model 3 are used as the corresponding inputs of BiGRU neural network model 3 of SSA and ESN neural network model of triangular fuzzy number-fuzzy wavelet neural network model 1 of SSA, the outputs of BiGRU neural network model 1 of SSA, BiGRU neural network model 2 of SSA and BiGRU neural network model 3 of SSA are used as the inputs of TDL beat delay 1, TDL beat delay 2 and TDL beat delay 3 respectively, and the outputs of TDL beat delay 1, TDL beat delay 2 and TDL beat delay 3 are used as the ESN neural network model of SSA of triangular fuzzy number respectively. The corresponding input of SN neural network model - fuzzy wavelet neural network model 1, the output of triangular fuzzy number ESN neural network model - SSA fuzzy wavelet neural network model 1 is used as the input of TDL beat delay device 4, the output of TDL beat delay device 4 is used as the corresponding input of SSA BiGRU neural network model 1, SSA BiGRU neural network model 2 and SSA BiGRU neural network model 3, the triangular fuzzy number ESN neural network model - SSA fuzzy wavelet neural network model 1 output a, b and c constitute the triangular fuzzy value [a, b, c] output by multiple parameter sensors. Energy consumption information module see Figure 1 shown.

[0031] (1) Design of fuzzy wavelet neural network model for SSA

[0032] The fuzzy wavelet neural network model of SSA is to use SSA to optimize the fuzzy wavelet neural network model, embed the wavelet neural network into the fuzzy model to form a fuzzy wavelet neural network model. The fuzzy wavelet neural network model consists of 5 layers. The first layer is the input layer, which directly leads to the second layer of membership function layer, the third layer is the fuzzy rule layer, and the fourth layer is the wavelet function layer. The first-order partial derivative of the Gaussian function is selected as φ(x)=xexp(-0.5x 2 ) is used as the mother wavelet function, which has good energy consumption information fitting performance. According to the selected mother wavelet function, it is put into the neurons of the second layer as an activation function after scaling and translation transformation. In the fifth layer, the output of the fourth wavelet layer is × the output of the fuzzy rule layer node of the third layer. The parameters of the fuzzy wavelet neural network are automatically updated and adjusted during the network training process, and the network parameters are updated using the back propagation algorithm. In order to improve the convergence speed and generalization ability of the fuzzy wavelet neural network model, a fuzzy wavelet neural network model parameter adjustment method based on dynamic learning rate is adopted. When the error between the ideal output and the actual output of the fuzzy wavelet neural network model becomes larger, the convergence speed is accelerated, that is, the learning rate is increased. When the error becomes smaller, it is necessary to carefully search for the optimal solution and reduce the learning rate. SSA is a sparrow search algorithm with the characteristics of strong global convergence ability and fast convergence speed. In order to avoid the effect of the fuzzy wavelet neural network model adjusted by artificial experience, SSA optimizes the learning rate, number of iterations, number of hidden layer neurons, and connection weight parameters of the fuzzy wavelet neural network model. The fuzzy wavelet neural network model process of SSA is as follows:

[0033] A. Initialization of the fuzzy wavelet neural network model of SSA. Initialize the parameters of SSA, including the position of the sparrow population, the maximum number of iterations, the upper and lower limits of the parameter values, and initialize the fuzzy wavelet neural network model structure, the number of hidden layer nodes, the connection weights and the learning rate as the optimization objectives.

[0034] B. Objective Function Establishment,The objective function of SSA is the root mean square error between the,predicted value of the untrained fuzzy wavelet neural network model and the,actual value.

[0035] C. Optimization, updating the position of the sparrow according to the result of the objective function. When the number of iterations set initially is met, the initial value of the fuzzy wavelet neural network model reaches the optimal value.

[0036] (2) Design of SOM clusterer for energy consumption information

[0037] The energy consumption information SOM clusterer includes an input layer and an output layer. The input layer consists of N node neurons, which are responsible for receiving the input energy consumption information data; the output layer consists of a neuron matrix arranged in rows and columns. During the operation of the network, only one neuron in the competition layer is activated at each moment, and it obtains priority activation in the mutual competition. The activated neuron represents the classification result of the current input energy consumption information data.

[0038] A: Parameter initialization, setting the maximum number of clustering iterations K, and randomizing the weight vectors between neurons in the competition layer.

[0039] B: Neuron competition. The Euclidean distance is used to compare the weight vectors corresponding to all neurons in the competition layer with the energy consumption information input vector of the input layer, and the neuron with the closest distance to the energy consumption information input vector is regarded as the winning neuron in the competition.

[0040] C: Adjust the weight vector of the competition layer. The weight vectors of the neighboring neurons of the winning neuron are adjusted to make them closer to the input vector.

[0041] D: Update data vector. Import new energy consumption information data samples into the SOM clustering network, and return to step B to continue adjusting the competition layer weight vector until all energy consumption information data samples have completed clustering training.

[0042] E: Algorithm convergence judgment: If the weight vector error between two adjacent iterations is less than the preset threshold or equal to the number of iterations, the clustering process ends; otherwise, let K = K + 1 and return to step B.

[0043] (3) SSA TCN temporal convolutional network model-ARIMA model design

[0044] The SSA TCN temporal convolutional network model-ARIMA model is a series connection of the SSA TCN temporal convolutional network model and the ARIMA model. The TCN temporal convolutional network model is a special architecture of CNN, which only has convolutional layers but no pooling layers. The deep neural network is constructed through convolutional layers to achieve the memory function of the historical information of the input energy consumption sequence. The TCN temporal convolutional network model includes dilated convolution and residual connection. The dilated convolution is divided into two convolution modes: causal dilated convolution and non-causal dilated convolution. The causal dilated convolution can prevent the future time energy consumption sequence from being lost. The leakage of information can realize real-time energy consumption prediction; and the use of non-causal dilated convolution can enable the TCN temporal convolutional network model to use all energy consumption inputs from the past to the future for prediction. The dilated convolution fills the convolution kernel of each layer of the TCN temporal convolutional network model with zeros, so that the perception of the convolution kernel increases exponentially with the increase in the number of convolutional layers, so that more local information on energy consumption can be obtained, thereby saving longer energy consumption history information. In addition, the TCN temporal convolutional network model can also take energy consumption sequences of any length as input and generate output sequences of the same length. The TCN temporal convolutional network model contains a residual network structure, and its residual block allows the TCN temporal convolutional network model to learn the residual errors of the feature mapping of each layer, avoiding the gradient vanishing problem of the deep network, so that the TCN temporal convolutional network model still maintains stable performance in the deep structure. The TCN temporal convolutional network model of SSA is optimized by the sparrow search algorithm. The parameters of the TCN temporal convolutional network model correspond to the activation function A, optimizer O, number of residual blocks L, random loss rate D, linear coefficient α value when the activation function is greater than 0, convolution kernel size K, convolution method P, batch size B, and initial learning rate λ. The specific steps for optimizing the parameters of the TCN temporal convolutional network model are as follows:

[0045] A. Population initialization: The initial number of groups for all group SSA optimization algorithms is 15.

[0046] B. Fitness: The smaller the fitness of all groups during the operation of the SSA optimization algorithm, the better the fitness.

[0047] C. Producer position update: The producer starts searching and leads the population towards the food area after finding the optimal solution.

[0048] D. Competitor position update: Some competitors constantly monitor producers and compete with them for food.

[0049] E. Overall update of population position: When the early warning system finds danger, the population quickly moves towards the potential optimal solution. F. Termination condition: When the algorithm completes 60 iterations, the operation is terminated.

[0050] ARIMA (p, d, q) is called the difference autoregressive moving average model, where AR stands for autoregressive, p stands for autoregressive term; MA stands for moving average, q stands for the number of moving average terms; d stands for the number of times the time series needs to be differentiated to become a stationary series. The ARIMA model is a combination of the AR model and the MA model. It means converting a non-stationary energy consumption time series into a stationary energy consumption time series, and then only performing the differential operation on the dependent variable X. t The lagged value and random error term ε t The model is established by regressing the present value and lagged value of.

[0051] (4) BiGRU neural network model design for SSA

[0052] The BiGRU neural network model is a network composed of two GRU neural network models with the same structure connected in reverse order. The network not only utilizes the forward energy consumption information state (historical moment state) but also fully utilizes the backward energy consumption information state (future moment state). Its advantage is that the energy consumption information received by the network unit is more comprehensive, which effectively improves the model's ability to predict energy consumption. The GRU neural network model is proposed to solve the problem of gradient explosion and vanishing of recurrent neural networks for long energy consumption sequences. The GRU neural network model adds a gating mechanism based on RNN to control the transmission of energy consumption information in the network.

[0053] The BiGRU neural network model of SSA is a sparrow search algorithm to optimize the BiGRU neural network model. 1) Set the value range of the batch size batsize and the learning rate alpha of the BiGRU neural network model training parameters, and set the value range of the number of hidden layers and the number of neurons in the layer num of the structural parameters; 2) Calculate the individual fitness of the sparrow population; 3) Update the position of the discoverer; 4) Update the position of the joiner; 5) Update the position of the sparrows that are aware of the danger, and the sparrows at the periphery of the population move closer to the safe area, and the sparrows at the center of the population walk randomly to get closer to other sparrows; 6) Calculate the fitness value of the new position of the individual sparrow, compare the updated fitness value with the original optimal fitness, and replace the inferior sparrow position with the optimal one to update the global optimal information; 7) If the number of iterations is met, proceed to the next step, otherwise return to step 5). 8) Output the optimal individual sparrow position, that is, the optimal energy consumption parameter group for this BiGRU neural network model training.

[0054] (5) Design of ESN neural network model of triangular fuzzy numbers - fuzzy wavelet neural network model 1 of SSA

[0055] ESN neural network model of triangular fuzzy number - fuzzy wavelet neural network model of SSA 1 is a series connection of ESN neural network model and fuzzy wavelet neural network model. ESN neural network model is a new type of dynamic neural network, which has all the advantages of dynamic neural network. At the same time, due to the introduction of the concept of "reserve pool" in echo state network, the "reserve pool" is to transform the middle connection part of the traditional dynamic neural network into a randomly connected "reserve pool". The whole learning process is actually the process of learning how to connect the "reserve pool". The "reserve pool" is actually a randomly generated large-scale recursive structure, in which the neurons are sparsely connected. SD is usually used to represent the percentage of neurons connected to each other in the total neurons N. The state equation of the ESN neural network model is:

[0056]

[0057] Where W is the state variable of the neural network, W in is the input variable of the neural network; W back is the output state variable connection weight matrix of the neural network; x(n) represents the internal state of the neural network; W out is the connection weight matrix between the core reserve pool of the ESN neural network model, the input of the neural network, and the output of the neural network; is the output deviation of the neural network or can represent noise; f = f[f1,f2,…,f n ] is the n activation functions of the neurons inside the “reserve pool”; f i is the hyperbolic tangent function; f out is the ε output function of the ESN neural network model. The outputs of multiple ESN neural network models are the predicted values ​​of various types of time-delay neural network models. The a, b and c output by the fuzzy wavelet neural network model of SSA constitute a triangular fuzzy number (a, b, c). The design method of the fuzzy wavelet neural network model refers to the relevant design steps of this patent.

[0058] Step 2, construct an energy consumption intelligent detection subsystem. The energy consumption intelligent detection subsystem includes an energy consumption information acquisition module, an ESN neural network model-ARIMA model, a TCN time convolution network model of SSA-ARIMA model, a BiGRU neural network model of SSA-ARIMA model, a BiGRU neural network model of SSA-ESN neural network model, a TCN time convolution network model of SSA of triangular fuzzy numbers-fuzzy wavelet neural network model, a BiGRU neural network model of SSA of triangular fuzzy numbers-fuzzy wavelet neural network model and an ESN neural network model of triangular fuzzy numbers-fuzzy wavelet neural network model of SSA 2. See the energy consumption intelligent detection subsystem. Figure 2 shown.

[0059] The outputs of multiple energy consumption sensors, multiple temperature sensors, and multiple humidity sensors are used as inputs of energy consumption information collection module 1, energy consumption information collection module 2, and energy consumption information collection module 3, respectively. The historical parameter values ​​of multiple energy consumption are used as inputs of energy consumption information collection module 4 and corresponding inputs of ESN neural network model-ARIMA model, and the output of energy consumption information collection module 4 is used as the corresponding input of ESN neural network model-ARIMA model. The outputs of energy consumption information collection module 1, energy consumption information collection module 2, energy consumption information collection module 3, and ESN neural network model-ARIMA model are used as BiGRU neural network model-of SSA, respectively. The corresponding inputs of the ESN neural network model, the SSA TCN temporal convolutional network model-ARIMA model, and the SSA BiGRU neural network model-ARIMA model, and the outputs of the SSA BiGRU neural network model-ESN neural network model, the SSA TCN temporal convolutional network model-ARIMA model, the SSA BiGRU neural network model-ARIMA model, and the ESN neural network model-ARIMA model are respectively used as the SSA TCN temporal convolutional network model-fuzzy wavelet neural network model of triangular fuzzy numbers and the SSA BiGRU neural network model-fuzzy wavelet neural network model of triangular fuzzy numbers. The corresponding inputs of the network model, the outputs of the TCN temporal convolutional network model of SSA of triangular fuzzy numbers - fuzzy wavelet neural network model and the outputs of the BiGRU neural network model of SSA of triangular fuzzy numbers - fuzzy wavelet neural network model are used as the corresponding inputs of the BiGRU neural network model of SSA - ESN neural network model, the TCN temporal convolutional network model of SSA - ARIMA model, the BiGRU neural network model of SSA - ARIMA model and the ESN neural network model of SSA of triangular fuzzy numbers - fuzzy wavelet neural network model 2, and the outputs of the ESN neural network model - ARIMA model are used as the triangular fuzzy numbers. The corresponding input of ESN neural network model - SSA's fuzzy wavelet neural network model 2 and the output of ESN neural network model - SSA's fuzzy wavelet neural network model 2 of triangular fuzzy numbers are respectively used as the corresponding input of SSA's TCN temporal convolutional network model - fuzzy wavelet neural network model of triangular fuzzy numbers, SSA's BiGRU neural network model - fuzzy wavelet neural network model and ESN neural network model - ARIMA model of triangular fuzzy numbers, and the triangular fuzzy number (X, Y, Z) composed of X, Y and Z output by the fuzzy wavelet neural network model of ESN neural network model - SSA of triangular fuzzy numbers is used as the energy consumption value of the detected object.

[0060] (1) ESN neural network model-ARIMA model design

[0061] The ESN neural network model-ARIMA model is a series connection of the ESN neural network model and the ARIMA model. The design method of the ESN neural network model and the ARIMA model refers to the relevant design process of the energy consumption information module of this patent.

[0062] (2) BiGRU neural network model of SSA-ARIMA model design

[0063] The BiGRU neural network model-ARIMA model of SSA is a series connection of the BiGRU neural network model of SSA and the ARIMA model. The design method of the BiGRU neural network model and the ARIMA model of SSA refers to the relevant design method of the energy consumption information module of this patent.

[0064] (3) Design of BiGRU neural network model of SSA-ESN neural network model

[0065] The BiGRU neural network model of SSA - ESN neural network model is the BiGRU neural network model of SSA and the ESN neural network model connected in series. The design method of the BiGRU neural network model of SSA and the ESN neural network model refers to the relevant design method of the energy consumption information module of this patent.

[0066] (4) Design of TCN temporal convolutional network model of SSA of triangular fuzzy numbers - fuzzy wavelet neural network model

[0067] The SSA TCN time convolution network model of triangular fuzzy numbers - the fuzzy wavelet neural network model is a series connection of the SSA TCN time convolution network model and the fuzzy wavelet neural network model. The j, k and l output by the fuzzy wavelet neural network model constitute a triangular fuzzy number (j, k, l). The design of the SSA TCN time convolution network model and the fuzzy wavelet neural network model refers to the relevant design method of the energy consumption information module of this patent.

[0068] (5) BiGRU neural network model of SSA of triangular fuzzy numbers - design of fuzzy wavelet neural network model

[0069] The SSA BiGRU neural network model of triangular fuzzy numbers - the fuzzy wavelet neural network model is a series connection of the SSA BiGRU neural network model and the fuzzy wavelet neural network model. The d, g and h output by the fuzzy wavelet neural network model constitute a triangular fuzzy number (d, g, h). The design of the SSA BiGRU neural network model and the fuzzy wavelet neural network model refers to the relevant design method of the energy consumption information module of this patent. Other modules in the energy consumption intelligent detection subsystem refer to the relevant design method of the energy consumption information module of this patent.

[0070] The present invention discloses an intelligent energy consumption management system based on the Internet of Things technology, which is composed of an energy consumption data perception layer, a data transmission layer and a data application layer. The energy consumption data perception layer is composed of a large number of energy consumption perception terminals. The energy consumption perception terminals collect temperature, humidity and energy consumption big data and summarize them to the energy consumption information coordinator of the data transmission layer. The energy consumption information coordinator receives the big data of the energy consumption perception terminals and uploads them to the data application layer. The data application layer is composed of an energy consumption OneNET cloud server, an energy consumption monitoring terminal and an energy consumption monitoring mobile phone APP, and has energy consumption data processing, analysis, display, monitoring and energy consumption alarm functions. The energy consumption monitoring terminal and the energy consumption monitoring mobile phone APP can send instructions to the energy consumption perception terminal through the energy consumption OneNET cloud server and the energy consumption information coordinator to control the external devices of the energy consumption perception terminal in real time. The data application layer uploads the data sent by the energy consumption information coordinator to the energy consumption OneNET cloud server and sends instructions to the energy consumption information coordinator. An intelligent energy consumption management system based on the Internet of Things technology is disclosed. Figure 3 shown.

[0071] 1. Design of energy consumption sensing terminal

[0072] The present invention adopts an energy consumption sensing terminal as an energy consumption sensing terminal for big data. The energy consumption sensing terminal and the energy consumption information coordinator realize two-way communication through the CC2530 communication module. The energy consumption sensing terminal includes sensors for collecting environmental parameters (temperature and humidity), current, voltage, flow and flame, etc., as well as corresponding signal conditioning circuits, STM32 microprocessors, CC2530 communication modules and relay output modules. The software of the energy consumption sensing terminal mainly realizes the collection and preprocessing of the CC2530 communication module and the energy consumption sensing terminal parameters. The model of the flow sensor is selected as a Hall flowmeter, which is used to detect the medium flow. The software is designed with C language programming, which has a high degree of compatibility, greatly improves the work efficiency of software design and development, and enhances the reliability, readability and portability of the program code. Energy consumption sensing terminal see Figure 4 shown.

[0073] 2. Design of energy consumption information coordinator

[0074] The energy consumption information coordinator includes an STM32 microprocessor, a CC2530 communication module, an ESP8266WiFi module, and a 2-way D / A conversion circuit. The output of the 2-way D / A conversion circuit is used as the input of the environmental control device and the energy consumption regulation device. A self-organizing communication network is constructed through the CC2530 communication module to realize data interaction between the energy consumption sensing terminal and the energy consumption information coordinator. The energy consumption information coordinator realizes two-way data interaction with the energy consumption OneNET cloud server through the ESP8266WiFi module. The energy consumption sensing terminal information is uploaded to the energy consumption OneNET cloud server through the energy consumption information coordinator to realize the monitoring of energy consumption big data. The energy consumption information coordinator drives ESP8266 to connect to the TCP server of the cloud server according to the pre-written WiFi information, API key and port number. The energy consumption data received by the ZigBee module is stored in the sequence table according to the energy consumption sensing terminal node number, and the data stream is encapsulated according to the MQTT protocol message format, and finally uploaded to the energy consumption OneNET cloud server. The energy consumption information coordinator sees Figure 5 .

[0075] 3. Energy consumption monitoring software design

[0076] The energy consumption monitoring terminal is an industrial control computer. Through the information exchange between the 5G network and the energy consumption OneNET cloud server, the energy consumption monitoring terminal realizes the analysis and processing of energy consumption big data and the energy consumption intelligent detection subsystem. The main functions of the energy consumption monitoring terminal are communication parameter setting, energy consumption data analysis, energy consumption data processing and energy consumption intelligent detection subsystem; the line graph is used to display energy consumption efficiency, power consumption and flow respectively, to realize the energy consumption management of temperature, flow, power, etc. The software functions of the energy consumption monitoring terminal are shown in Figure 6 shown.

[0077] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent energy consumption detection method based on Internet of Things technology, characterized in that: The steps include: Step 1: Construct an energy consumption information collection module, which includes a fuzzy wavelet neural network model of SSA, an energy consumption information SOM clusterer, a TCN temporal convolutional network model of SSA - an ARIMA model, a BiGRU neural network model of SSA, an ESN neural network model of triangular fuzzy numbers - a fuzzy wavelet neural network model 1 of SSA and a TDL beat delay device; The outputs of multiple parameter sensors are used as the corresponding inputs of multiple SSA fuzzy wavelet neural network models respectively, the output values ​​of multiple SSA fuzzy wavelet neural network models are used as the inputs of the energy consumption information SOM clusterer, and the output values ​​of the three types of SSA fuzzy wavelet neural network models output by the energy consumption information SOM clusterer are used as SSA's TCN temporal convolutional network model-ARIMA model 1 and SSA's BiGRU neural network model 1, SSA's TCN temporal convolutional network model-ARIMA model 2 and SSA's BiGRU neural network model 2, and SSA's TCN temporal convolutional network model-A The corresponding inputs of RIMA model 3 and SSA’s BiGRU neural network model 3, the outputs of SSA’s TCN temporal convolutional network model-ARIMA model 1 are respectively used as the corresponding inputs of SSA’s BiGRU neural network model 1 and the ESN neural network model of triangular fuzzy numbers-SSA’s fuzzy wavelet neural network model 1, the outputs of SSA’s TCN temporal convolutional network model-ARIMA model 2 are respectively used as the corresponding inputs of SSA’s BiGRU neural network model 2 and the ESN neural network model of triangular fuzzy numbers-SSA’s fuzzy wavelet neural network model 1, the TCN temporal convolution of SSA The outputs of network model-ARIMA model 3 are used as the corresponding inputs of BiGRU neural network model 3 of SSA and ESN neural network model of triangular fuzzy number-fuzzy wavelet neural network model 1 of SSA, the outputs of BiGRU neural network model 1 of SSA, BiGRU neural network model 2 of SSA and BiGRU neural network model 3 of SSA are used as the inputs of TDL beat delay 1, TDL beat delay 2 and TDL beat delay 3 respectively, and the outputs of TDL beat delay 1, TDL beat delay 2 and TDL beat delay 3 are used as the ESN neural network model of SSA of triangular fuzzy number respectively. The corresponding input of SN neural network model - fuzzy wavelet neural network model 1, the output of ESN neural network model of triangular fuzzy number - fuzzy wavelet neural network model 1 of SSA is used as the input of TDL beat delay device 4, the output of TDL beat delay device 4 is used as the corresponding input of BiGRU neural network model 1 of SSA, BiGRU neural network model 2 of SSA and BiGRU neural network model 3 of SSA, a, b and c outputted by fuzzy wavelet neural network model 1 of ESN neural network model of triangular fuzzy number - SSA constitute the triangular fuzzy value [a, b, c] outputted by multiple parameter sensors; Step 2: construct an energy consumption intelligent detection subsystem, which includes multiple energy consumption information collection modules, an ESN neural network model-ARIMA model, a TCN temporal convolutional network model-ARIMA model of SSA, a BiGRU neural network model-ARIMA model of SSA, a BiGRU neural network model-ESN neural network model of SSA, a TCN temporal convolutional network model-fuzzy wavelet neural network model of SSA of triangular fuzzy numbers, a BiGRU neural network model-fuzzy wavelet neural network model of SSA of triangular fuzzy numbers, and an ESN neural network model-fuzzy wavelet neural network model of SSA of triangular fuzzy numbers 2; Step 3: The outputs of multiple energy consumption sensors, multiple temperature sensors, and multiple humidity sensors are used as inputs of energy consumption information collection module 1, energy consumption information collection module 2, and energy consumption information collection module 3, respectively. Multiple energy consumption historical parameter values ​​are used as inputs of energy consumption information collection module 4 and corresponding inputs of ESN neural network model-ARIMA model, and the output of energy consumption information collection module 4 is used as the corresponding input of ESN neural network model-ARIMA model. The outputs of energy consumption information collection module 1, energy consumption information collection module 2, energy consumption information collection module 3, and ESN neural network model-ARIMA model are used as BiGRU neural networks of SSA, respectively. The corresponding inputs of the model-ESN neural network model, SSA's TCN temporal convolutional network model-ARIMA model, and SSA's BiGRU neural network model-ARIMA model, and the outputs of SSA's BiGRU neural network model-ESN neural network model, SSA's TCN temporal convolutional network model-ARIMA model, SSA's BiGRU neural network model-ARIMA model, and ESN neural network model-ARIMA model are respectively used as SSA's TCN temporal convolutional network model-fuzzy wavelet neural network model of triangular fuzzy numbers and SSA's BiGRU neural network model-fuzzy wavelet of triangular fuzzy numbers. The corresponding inputs of the neural network model, the outputs of the TCN temporal convolutional network model of SSA for triangular fuzzy numbers - fuzzy wavelet neural network model and the outputs of the BiGRU neural network model of SSA for triangular fuzzy numbers - fuzzy wavelet neural network model are used as the corresponding inputs of the BiGRU neural network model of SSA - ESN neural network model, the TCN temporal convolutional network model of SSA - ARIMA model, the BiGRU neural network model of SSA - ARIMA model and the ESN neural network model of SSA for triangular fuzzy numbers - fuzzy wavelet neural network model 2, and the outputs of the ESN neural network model - ARIMA model are used as the triangular fuzzy numbers. The corresponding input of ESN neural network model - SSA fuzzy wavelet neural network model 2 and the output of ESN neural network model - SSA fuzzy wavelet neural network model 2 of triangular fuzzy numbers are respectively used as the corresponding input of TCN temporal convolution network model - fuzzy wavelet neural network model of SSA of triangular fuzzy numbers, BiGRU neural network model - fuzzy wavelet neural network model of SSA of triangular fuzzy numbers and ESN neural network model - ARIMA model, and the triangular fuzzy number (X, Y, Z) composed of X, Y and Z output by the fuzzy wavelet neural network model of ESN neural network model - SSA of triangular fuzzy numbers is used as the energy consumption value of the detected object.

2. An intelligent energy consumption management system based on Internet of Things technology, characterized in that: It includes an energy consumption data perception layer, a data transmission layer and a data application layer. The energy consumption data perception layer is composed of a large number of energy consumption perception terminals. The energy consumption perception terminals include energy consumption sensors, temperature sensors, humidity sensors, and an energy consumption information coordinator that collects temperature, humidity and energy consumption big data and summarizes them into the data transmission layer. The energy consumption information coordinator receives the big data collected by the energy consumption perception terminal and uploads it to the data application layer. The data application layer is composed of an energy consumption OneNET cloud server, an energy consumption monitoring terminal and an energy consumption monitoring mobile phone APP. The energy consumption monitoring terminal and the energy consumption monitoring mobile phone APP send instructions to the energy consumption perception terminal through the energy consumption OneNET cloud server and the energy consumption information coordinator to control the external devices of the energy consumption perception terminal in real time. The data application layer uploads the data sent by the energy consumption information coordinator to the energy consumption OneNET cloud server and sends instructions to the energy consumption information coordinator. The energy consumption monitoring terminal, the energy consumption perception terminal, the energy consumption information coordinator and the energy consumption OneNET cloud server execute the steps of the energy consumption intelligent detection method based on the Internet of Things technology as described in claim 1.