Method and device for determining area with unreasonable energy consumption, storage medium and electronic device

Abstract

The invention discloses a method and device for determining an unreasonable energy consumption area, a storage medium and an electronic device, and relates to the field of computers.The method for determining the unreasonable energy consumption area comprises the steps that real-time energy consumption data of m sub-areas of a target area are collected regularly according to a time window T, and m pieces of first real-time energy consumption data are obtained; respectively predicting the energy consumption data of the m sub-regions in the next time window to obtain m pieces of first predicted energy consumption data; performing energy consumption unreasonable analysis on the m sub-regions, and determining t suspected energy consumption unreasonable regions in the m sub-regions; collecting energy consumption influence factor data of the t energy consumption suspected unreasonable areas in the current time window, and classifying the t energy consumption influence factor data to obtain t classification results; and under the condition that p classification results in the t classification results indicate that the areas are abnormal, p energy consumption suspected unreasonable areas corresponding to the p classification results are determined as energy consumption unreasonable areas.

Description

Technical Field

[0001] The present application relates to the field of computers, and more specifically, to a method and device for determining an unreasonable energy consumption area, a storage medium, and an electronic device. Background Art

[0002] With the rise of smart cities, the emergence of new generation information technologies such as the Internet of Things, cloud computing, and big data has brought new opportunities for the development of industrial parks. Many industrial parks with good foundations have sought transformation and development with the help of new generation information technologies to enhance their sustainable competitive advantages in an information environment, which has led to the concept of smart industrial parks.

[0003] The energy consumption management of smart parks focuses on reducing energy waste, improving energy utilization efficiency and intelligent operation management. To this end, real-time monitoring is often required to accurately collect energy consumption data of various equipment in the park, and through dynamic analysis of these data, energy consumption peaks and anomalies can be detected, so as to finely and intelligently control and manage the energy consumption of the park.

[0004] Energy consumption data has the characteristics of time, periodicity, and streaming data. In smart parks, energy consumption data in different areas also have their own characteristics. Energy consumption data is usually obtained using various sensors and measuring devices, such as using smart meters to measure electricity consumption, and is sent to the onboard recording system in a specific data format for real-time storage. Energy consumption anomalies are divided into equipment anomalies, system anomalies, operation anomalies, and environmental anomalies. Energy consumption warning methods include three categories, namely, methods based on statistics and mathematical models, metaheuristic methods, and methods based on machine learning and deep learning. Among them, methods based on statistics and mathematical models often have good interpretability, but for high-dimensional data, there will be problems such as the inability to capture complex nonlinear relationships between features; metaheuristic algorithms are more general, but may fall into local optimal solutions and are suitable for medium-sized data.

[0005] There are multiple areas in a smart park, including office areas, production areas, logistics areas, life service areas, information centers, control centers, energy management areas, etc. Since different areas have different energy consumption characteristics, a single energy consumption prediction and abnormal classification method is often not suitable for energy consumption warning problems in all areas. Since smart parks implement normalized energy consumption monitoring and energy consumption abnormality judgment, when using an energy consumption abnormality judgment method with too high complexity, it will undoubtedly increase the overall energy consumption burden and cause unnecessary energy waste.

[0006] With the development of deep learning, deep learning models such as recurrent neural networks (RNN) and convolutional neural networks (CNN) have attracted widespread attention from scholars in various fields. On the one hand, RNN is suitable for processing and analyzing time series data. Energy consumption data has the characteristics of temporal continuity and volatility, which is suitable for feature extraction using RNN. GRU (Gated Recurrent Unit), as a variant of RNN, has the characteristics of fewer parameters, faster training speed, and less memory usage compared with the widely used long short-term memory network (LSTM), and is suitable as a model for normalized energy consumption prediction. On the other hand, CNN is suitable for processing multi-dimensional data. Energy consumption is affected by various factors, which is suitable for feature extraction using CNN. DenseNet (Densely Connected Convolutional Networks) is a CNN architecture that improves the flow of information and gradients by establishing direct connections between each layer, can achieve feature reuse, and significantly reduce the number of parameters. It is suitable for classifying energy consumption anomalies in resource-limited environments and improving the accuracy of energy consumption warnings.

[0007] Regarding the problem of the lack of an effective method for warning abnormal energy consumption in smart parks in related technologies, no effective solution has been proposed yet.

[0008] Therefore, it is necessary to improve the related technology to overcome the above-mentioned defects in the related technology. Summary of the invention

[0009] The embodiments of the present application provide a method and device for determining an area of ​​unreasonable energy consumption, a storage medium, and an electronic device, so as to at least solve the problem in the related art that there is a lack of an effective method for warning abnormal energy consumption in a smart park.

[0010] According to one aspect of an embodiment of the present application, a method for determining an area with unreasonable energy consumption is provided, comprising: regularly collecting real-time energy consumption data of m sub-areas of a target area according to a time window T, and obtaining m first real-time energy consumption data, wherein m is a positive integer, and the first real-time energy consumption data is the real-time energy consumption data within the current time window; predicting the energy consumption data of the m sub-areas in the next time window by using m energy consumption prediction models and the m first real-time energy consumption data, and obtaining m first predicted energy consumption data, wherein the m energy consumption prediction models correspond one-to-one to the m sub-areas; predicting the energy consumption of the m sub-areas according to the m second real-time energy consumption data and the m first predicted energy consumption data; The unreasonable energy consumption analysis is performed on the domain to determine t suspected unreasonable energy consumption areas among the m sub-areas, wherein t is a positive integer and t is less than or equal to m; the energy consumption influencing factor data of the t suspected unreasonable energy consumption areas in the current time window are collected, and the t energy consumption influencing factor data are respectively classified through t energy consumption anomaly classification models to obtain t classification results, wherein the t energy consumption anomaly classification models correspond to the t suspected unreasonable energy consumption areas one by one; when p classification results among the t classification results indicate regional abnormalities, the p suspected unreasonable energy consumption areas corresponding to the p classification results are determined as unreasonable energy consumption areas, wherein p is a positive integer.

[0011] In an exemplary embodiment, the energy consumption data of the m sub-areas in the next time window are predicted respectively by using m energy consumption prediction models and the m first real-time energy consumption data. Before obtaining the m first predicted energy consumption data, the method also includes: acquiring the historical energy consumption data of the m sub-areas; performing data preprocessing on the m historical energy consumption data to obtain m feature data; and training the m energy consumption prediction models respectively according to the m feature data.

[0012] In an exemplary embodiment, data preprocessing is performed on m historical energy consumption data to obtain m characteristic data, including: obtaining a timestamp sequence corresponding to n moments, and performing weighted merging processing on each timestamp to obtain an integrated time series, wherein n is a positive integer, the timestamp sequence includes multiple timestamps, the integrated time series includes multiple integration times, and the multiple timestamps correspond to the multiple integration times one by one; dividing the m historical energy consumption data according to the time window T, and extracting energy consumption features of the divided m historical energy consumption data according to the multiple integration times to obtain m energy consumption feature matrices, wherein the energy consumption features include: maximum value, minimum value, moving mean, and moving variance; and performing feature normalization on the m energy consumption feature matrices to obtain the m characteristic data.

[0013] In an exemplary embodiment, before the t energy consumption influencing factor data are respectively classified by t energy consumption anomaly classification models to obtain t classification results, the method also includes: collecting the energy consumption influencing factor data of the m sub-areas; performing data preprocessing on the m energy consumption influencing factor data to obtain m energy consumption factor characteristic data; and training m energy consumption anomaly classification models according to the m energy consumption factor characteristic data, wherein the m energy consumption anomaly classification models include the t energy consumption anomaly classification models.

[0014] In an exemplary embodiment, unreasonable energy consumption analysis is performed on the m sub-regions based on the m second real-time energy consumption data and the m first predicted energy consumption data, and t areas of suspected unreasonable energy consumption in the m sub-regions are determined, including: determining the first-level alarm threshold of the m sub-regions to obtain m first-level alarm thresholds; calculating the absolute value of the difference between the second real-time energy consumption data and the first predicted energy consumption data corresponding to each sub-region to obtain m absolute values; determining the t areas of suspected unreasonable energy consumption based on the m first-level alarm thresholds and the m absolute values, wherein the absolute value of the area of ​​suspected unreasonable energy consumption is greater than the first-level alarm threshold.

[0015] In an exemplary embodiment, after t energy consumption influencing factor data are classified respectively by t energy consumption anomaly classification models to obtain t classification results, the method further includes: when r classification results among the t classification results indicate that the area is normal, r energy consumption suspected unreasonable areas corresponding to the r classification results are determined as normal areas, wherein r is a positive integer; and r first-level alarm thresholds corresponding to the r energy consumption suspected unreasonable areas are corrected.

[0016] According to another aspect of the embodiment of the present application, a device for determining an area with unreasonable energy consumption is also provided, including: a first acquisition module, used to periodically collect real-time energy consumption data of m sub-areas of a target area according to a time window T, and obtain m first real-time energy consumption data, wherein m is a positive integer, and the first real-time energy consumption data is the real-time energy consumption data within the current time window; a prediction module, used to predict the energy consumption data of the m sub-areas in the next time window by using m energy consumption prediction models and the m first real-time energy consumption data, and obtain m first predicted energy consumption data, wherein the m energy consumption prediction models correspond one-to-one to the m sub-areas; an analysis module, used to predict the energy consumption data of the m sub-areas according to the m second real-time energy consumption data and the m first predicted energy consumption data. The m sub-regions are analyzed for unreasonable energy consumption to determine t suspected unreasonable energy consumption regions among the m sub-regions, wherein t is a positive integer and t is less than or equal to m; a second acquisition module is used to collect energy consumption influencing factor data of the t suspected unreasonable energy consumption regions in the current time window, and classify the t energy consumption influencing factor data respectively through t energy consumption anomaly classification models to obtain t classification results, wherein the t energy consumption anomaly classification models correspond one-to-one to the t suspected unreasonable energy consumption regions; a determination module is used to determine the p suspected unreasonable energy consumption regions corresponding to the p classification results as unreasonable energy consumption regions when p classification results among the t classification results indicate that the region is abnormal, wherein p is a positive integer.

[0017] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the method for determining the unreasonable energy consumption area when running.

[0018] According to another aspect of an embodiment of the present application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the method for determining an unreasonable energy consumption area through the computer program.

[0019] According to another aspect of the embodiments of the present application, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps of the method described in each embodiment of the present application are implemented.

[0020] Through the present application, the real-time energy consumption data of m sub-areas of the target area are first collected according to a fixed time window T to obtain the first real-time energy consumption data in the m current time windows, where m is a positive integer; the energy consumption data of the next time window is predicted for the first real-time energy consumption data through the energy consumption prediction model of each sub-area to obtain m first predicted energy consumption data; then, the unreasonable energy consumption of the m sub-areas is analyzed based on the m second real-time energy consumption data and the m first predicted energy consumption data to determine t unreasonable energy consumption areas; the energy consumption influencing factor data of the t suspected unreasonable energy consumption areas in the current time window are collected, and they are classified through t energy consumption anomaly classification models , and obtain t classification results; when there are p classification results indicating regional abnormalities, the corresponding p areas of suspected unreasonable energy consumption are determined as unreasonable energy consumption areas; using the above scheme, energy consumption prediction models and energy consumption anomaly classification models are trained for different areas respectively, and the next time window is predicted according to the energy consumption data of different areas in the current time window, and then the difference between the actual energy consumption data and the predicted results is combined to determine whether there is an abnormality. If there is an abnormality, a secondary judgment is further made according to the energy consumption anomaly classification model to accurately determine whether the area has an energy consumption abnormality problem; thereby solving the problem of the lack of an effective smart park energy consumption abnormality early warning method in the relevant technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0022] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0023] Figure 1 It is a hardware structure block diagram of a computer terminal of a method for determining an unreasonable energy consumption area in an embodiment of the present application;

[0024] Figure 2 is a flow chart of a method for determining an unreasonable energy consumption area according to an embodiment of the present application;

[0025] Figure 3 It is a schematic diagram of the application structure of an optional smart park energy consumption early warning method according to an embodiment of the present application;

[0026] Figure 4 It is a flow chart of an optional smart park energy consumption early warning method according to an embodiment of the present application;

[0027] Figure 5It is a structural block diagram of a device for determining an unreasonable energy consumption area according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] The method embodiments provided in the embodiments of the present application can be executed in a computer terminal or a similar computing device. Taking running on a computer terminal as an example, Figure 1 1 is a hardware structure block diagram of a computer terminal for determining an unreasonable energy consumption area according to an embodiment of the present application. Figure 1 As shown, the computer terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor (Central Processing Unit, MCU) or a programmable logic device (Field Programmable Gate Array, FPGA)) and a memory 104 for storing data, wherein the above-mentioned computer terminal may also include a transmission device 106 and an input and output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above-mentioned computer terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0031] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method for determining the unreasonable energy consumption area in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0032] A wireless network provided by a communication provider of a computer terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet wirelessly.

[0033] In this embodiment, a method for determining an unreasonable energy consumption area is provided. Figure 2 is a flow chart of a method for determining an unreasonable energy consumption area according to an embodiment of the present application. Figure 2 As shown, the process includes the following steps S202-S210:

[0034] Step S202: regularly collecting real-time energy consumption data of m sub-areas of the target area according to the time window T, and obtaining m first real-time energy consumption data, wherein m is a positive integer, and the first real-time energy consumption data is the real-time energy consumption data within the current time window;

[0035] Step S204: predicting the energy consumption data of the m sub-regions in the next time window respectively by using the m energy consumption prediction models and the m first real-time energy consumption data to obtain m first predicted energy consumption data, wherein the m energy consumption prediction models correspond to the m sub-regions one by one;

[0036] Step S206: performing unreasonable energy consumption analysis on the m sub-regions according to the m second real-time energy consumption data and the m first predicted energy consumption data, and determining t suspected unreasonable energy consumption regions in the m sub-regions, where t is a positive integer and t is less than or equal to m;

[0037] Step S208: collecting the energy consumption influencing factor data of the t energy consumption suspected unreasonable areas in the current time window, and classifying the t energy consumption influencing factor data respectively through t energy consumption abnormality classification models to obtain t classification results, wherein the t energy consumption abnormality classification models correspond one-to-one to the t energy consumption suspected unreasonable areas;

[0038] Step S210: When p of the t classification results indicate that the area is abnormal, the p suspected unreasonable energy consumption areas corresponding to the p classification results are determined as unreasonable energy consumption areas, where p is a positive integer.

[0039] Through the above steps, the real-time energy consumption data of m sub-areas of the target area are first collected according to a fixed time window T to obtain m first real-time energy consumption data in the current time window, where m is a positive integer; the energy consumption data of the next time window is predicted for the first real-time energy consumption data through the energy consumption prediction model of each sub-area to obtain m first predicted energy consumption data; then, the unreasonable energy consumption of the m sub-areas is analyzed based on the m second real-time energy consumption data and the m first predicted energy consumption data to determine t unreasonable energy consumption areas; the energy consumption influencing factor data of the t suspected unreasonable energy consumption areas in the current time window are collected, and they are classified through t energy consumption anomaly classification models. class, and obtain t classification results; when there are p classification results indicating regional abnormalities, the corresponding p areas of suspected unreasonable energy consumption are determined as unreasonable energy consumption areas; using the above scheme, energy consumption prediction models and energy consumption anomaly classification models are trained for different areas respectively, and the next time window is predicted according to the energy consumption data of different areas in the current time window, and then the difference between the actual energy consumption data and the predicted results is combined to determine whether there is an abnormality. If there is an abnormality, a secondary judgment is further made according to the energy consumption anomaly classification model to accurately determine whether the area has an energy consumption abnormality problem; thereby solving the problem of the lack of an effective smart park energy consumption abnormality early warning method in the relevant technology.

[0040] In an exemplary embodiment, the energy consumption data of the m sub-areas in the next time window are predicted respectively by using m energy consumption prediction models and the m first real-time energy consumption data. Before obtaining the m first predicted energy consumption data, the method also includes: acquiring the historical energy consumption data of the m sub-areas; performing data preprocessing on the m historical energy consumption data to obtain m feature data; and training the m energy consumption prediction models respectively according to the m feature data.

[0041] After obtaining the historical energy consumption data of each area, the energy consumption data is sliced ​​according to the specified time slice, and the feature data such as minimum value, maximum value, moving average, moving variance and integration time are extracted based on the energy consumption fragments, and the m energy consumption prediction models are trained using GRU.

[0042] Optionally, data preprocessing is performed on the m historical energy consumption data to obtain m characteristic data, including: obtaining a timestamp sequence corresponding to n moments, and performing weighted merging processing on each timestamp to obtain an integrated time series, wherein n is a positive integer, the timestamp sequence includes multiple timestamps, the integrated time series includes multiple integration times, and the multiple timestamps correspond to the multiple integration times one by one; dividing the m historical energy consumption data according to the time window T, and extracting energy consumption features of the divided m historical energy consumption data according to the multiple integration times to obtain m energy consumption feature matrices, wherein the energy consumption features include: maximum value, minimum value, moving mean, and moving variance; and performing feature normalization on the m energy consumption feature matrices to obtain the m characteristic data.

[0043] First, obtain the electric energy consumption data sequence of each area at n moments areaElectricityConsumption = {areaElectricityConsumption 1 ,areaElectricityConsumption 2 ,…,

[0044] areaElectricityConsumption i ,…,areaElectricityConsumption m-1 ,

[0045] areaElectricityConsumption m},in

[0046] The electric energy consumption data sequence of the i-th area areaElectricityConsumption i =

[0047] {electricityConsumption i1 ,electricityConsumption i2 ,…,

[0048] electricityConsumption ij ,…,electricityConsumption in-1 ,

[0049] electricityConsumption in}, where electricityConsumption ijis the electric energy consumption of the i-th region at the j-th moment. Get the timestamp sequence corresponding to the electric energy consumption data sequence of each region at n moments timestamp=

[0050] {timestamp 1 ,timestamp 2 ,…,timestamp i ,…,timestamp n-1 ,timestamp n}, where the i-th timestamp is timestamp i ={month i ,day i ,hour i}, where month i is the month and day of the ith moment i is the day of the ith moment, hour i is the time of the ith moment.

[0051] Then traverse the areaElectricityConsumption data sequence of each area, and for the area ith areaElectricityConsumption data sequence i , divided according to the time window T and stored in the i-th area energy consumption fragment matrix areaEC by row i , we get the energy consumption fragment matrix set of each area areaEC = {areaEC 1 ,areaEC 2 ,…,areaEC i ,…,areaEC m-1 ,areaEC m}.

[0052] Traverse the timestamp sequence timestamp of n moments, and for the i-th timestamp timestamp i , calculate the i-th integration time moment according to the following formula (weighted merging): i , the integrated time series moment = {moment 1 ,moment 2 ,…,moment i ,…,moment n-1 ,moment n}:moment i =month i *10000+day i *100+hour i .

[0053] Traverse the set of energy consumption fragment matrices of each region areaEC, and for the i-th region energy consumption fragment matrix areaEC i Each row (jth row) and the element electricityConsumption of that row ik , get the maximum value of the row ij 、Minimum value min ij , count the number of negative and empty values ​​in the row errorCnt ij , calculate the moving average avg of the row according to the following formula (moving average, moving variance) ij , moving variance var ij Combined with the time series moment, the energy consumption feature matrix areaFeature of the i-th area is obtained i . Store the energy consumption feature matrices of all regions into the energy consumption feature set (i.e. the m energy consumption feature matrices mentioned above). ij ={electricityConsumption ik |electricityConsumption ik ≥electricityConsumption is ,k≠s,1≤j≤n-T+1,j≤k,s≤j+T-1},

[0054] min ij ={electricityConsumption ik |electricityConsumption ik ≤electricityConsumption is ,k≠s,1≤j≤n-T+1,j≤k,s≤j+T-1},

[0055]

[0056]

[0057]

[0058] where max ij is the energy consumption fragment matrix of the i-th area areaEC i The maximum value of the jth row, min ij is the energy consumption fragment matrix of the i-th area areaEC i The minimum value of the jth row, avg ij is the energy consumption fragment matrix of the i-th area areaEC i The mean of the jth row, var ijis the energy consumption fragment matrix of the i-th area areaEC i The variance of the jth row, electricityConsumption ik is the electric energy consumption of the i-th region at the k-th moment, and T is the time step.

[0059] Traverse the energy consumption feature set Feature, and for the energy consumption feature matrix areaFeature of the i-th area i , normalize each column of features according to the following formula (Min-Max) to obtain the standardized feature matrix normFeature of the i-th region i , store the standardized feature matrices of all regions into the standardized feature matrix set normFeature = {normFeature 1 ,normFeature 2 ,…,normFeature i ,…,normFeature m-1 ,normFeature m}, maxMX i ={max ik |max ik ≥max is ,k≠s,1≤k,s≤n-T+1},maxMN i ={max ik |max ik ≤max is ,k≠s,1≤k,s≤n-T+1}

[0060] minMX i ={min ik |min ik ≥min is ,k≠s,

[0061] 1≤k,s≤n-T+1},minMN i ={min ik |min ik ≤min is ,k≠s,1≤k,s≤n-T+1}

[0062] avgMX i ={avg ik |avg ik ≥avg is ,k≠s,1≤k,s≤n-T+1},

[0063] avgMN i ={avgik |avg ik ≤avg is ,k≠s,1≤k,s≤n-T+1} varMX i ={var ik |var ik ≥var is ,k≠s,1≤k,s≤n-T+1},varMN i ={var ik |var ik ≤var is ,k≠s,1≤k,s≤n-T+1}, momentMX i ={moment ik |moment ik ≥moment is ,k≠s,1≤k,s≤n-T+1}, momentMN i ={moment ik |moment ik ≤moment is ,k≠s,1≤k,s≤n-T+1},

[0064]

[0065] Get these m feature data.

[0066] In an optional embodiment, the GRU model can be constructed in the following manner, using mean square error as the loss function and Adam as the optimizer: the first hidden layer contains 32 neurons, and the activation function is sigmoid; the second hidden layer contains 32 neurons, and the activation function is Tanh; the output layer contains 1 neuron and has no activation function.

[0067] Then traverse the standardized feature matrix set normFeature of all regions, and for the standardized feature matrix normFeature of the i-th region i The dataset is randomly divided into 70% training set, 10% test set and 20% validation set.

[0068] Use the GRU model for training and set the parameters as follows. After training, the energy consumption prediction model for each area is obtained: areaGRU{areaGRU 1 ,areaGRU 2 ,…,areaGRU i ,…,areaGRU m-1 ,areaGRU m}, where areaGRU i For the energy consumption prediction model of the i-th area: max_epochs=1000, batch_size=64, sequence_length=T, update_learning_rate=0.01, features=5, objective_loss_function=squared_error, verbose=1.

[0069] After completing the model training, you can perform energy consumption warning by following the steps below:

[0070] Step 1: Collect and obtain the real-time energy consumption data sequence of each area at T moments nowAreaElectricityConsumption = {nowAreaElectricityConsumption 1 ,nowAreaElectricityConsumption 2 ,…,nowAreaElectricityConsumption i ,…,nowAreaElectricityConsumption m-1 ,nowAreaElectricityConsumption m} and the current timestamp nowTimestamp, where the real-time electric energy consumption data sequence nowAreaElectricityConsumption of the i-th area i ={nowElectricityConsumption i1 ,nowElectricityConsumption i2 ,…,nowElectricityConsumption ij ,…,nowElectricityConsumption iT-1 ,nowElectricityConsumption iT}, where nowElectricityConsumption ij The real-time power consumption of the i-th region at the j-th moment. Process the current timestamp to obtain the current integration time nowMoment. Traverse nowNormFeature = {nowNormFeature 1 ,nowNormFeature 2 ,…,nowNormFeature i,…,nowNormFeature m-1 ,nowNormFeature m}.

[0071] Step 2: Traverse the standardized real-time feature matrix set nowNormFeature of all regions, and convert the real-time energy consumption feature nowNormFeature of the i-th region into i Input to the energy consumption prediction model areaGRU of the i-th area after training i The model predicts and the output result is the real-time energy consumption prediction value energyPredict of the i-th region i . Get the real-time energy consumption prediction value set of all regions energyPredict = {energyPredict 1 ,energyPredict 2 ,…,energyPredict i ,…,energyPredict m-1 ,energyPredict m}.

[0072] Step 3: traverse the real-time energy consumption prediction value set energyPredict of all regions and the first-level alarm threshold thr of all regions = {thr 1 ,thr 2 ,…,thr i ,…,thr m-1 ,thr m}, the real-time energy consumption prediction value energyPredict for the i-th area i and the first level alarm threshold thr of the i-th area i , when |energyPredict i -nowElectricityConsumptioni T |>thr i , set the first level warning flag warning for the ith area i The value is true, and the set of all regional level 1 alarms warning = {warning 1 , warning 2 , ..., warning i , ..., warning m-1 , warning m}.

[0073]

[0074] Optionally, before classifying the t energy consumption influencing factor data respectively through t energy consumption anomaly classification models to obtain t classification results, the method also includes: collecting the energy consumption influencing factor data of the m sub-areas; performing data preprocessing on the m energy consumption influencing factor data to obtain m energy consumption factor characteristic data; and training m energy consumption anomaly classification models according to the m energy consumption factor characteristic data, wherein the m energy consumption anomaly classification models include the t energy consumption anomaly classification models.

[0075] After obtaining the historical energy consumption influencing factor data of each region, the data is standardized and the energy consumption anomaly classification model is trained using DenseNet. For areas with abnormal energy consumption, the trained energy consumption anomaly classification model is used to determine the energy consumption anomaly category of the area. If the category is normal, the first-level energy consumption threshold of the area is adaptively adjusted, otherwise a second-level alarm is issued.

[0076] Specifically, the process of training the energy consumption anomaly classification model is as follows:

[0077] Collect energy consumption influencing factor data of all regions at n moments energyRelevant = {energyRelevant 1 ,energyRelevant 2 ,…,energyRelevant i ,…,energyRelevant m-1 ,energyRelevant m}, where the energy consumption influencing factor data of the i-th region is energyRelevant i ={factor i1 ,factor i2 ,…,factor ij ,…,factor il-1 ,factor il}, where factor ij is the jth energy consumption factor sequence of the ith region. factor ij ={factor ij1 ,factor ij2 ,…,factor ijk ,…,factor ijn-1 ,factor ijn}, where factor ijk is the energy consumption factor value of the jth energy consumption factor in the i-th region at the k-th moment.

[0078] Traverse the energy consumption influencing factor data energyRelevant of all regions at n moments, and calculate the energy consumption influencing factor data energyRelevant of the i-th region i , the energy consumption factor sequence of the i-th region j is factor ij Store the energy consumption factor matrix of the i-th region in columns. i , we get the energy consumption factor matrix set of all regions FactorMatrix = {FactorMatrix 1 ,FactorMatrix 2 ,…,FactorMatrix i ,…,FactorMatrix m-1 ,FactorMatrix m}:

[0079]

[0080] Traverse the energy consumption factor matrix set FactorMatrix, and for the energy consumption factor matrix FactorMatrix of the i-th region i , normalize each column feature according to the following formula (Min-Max) to obtain the standardized energy consumption factor feature matrix normFactorFeature of the i-th region i , store the standardized feature matrices of all regions into the standardized energy consumption factor feature matrix set normFactorFeature={normFactorFeature 1 ,normFactorFeature 2 ,…,normFactorFeature i ,…,normFactorFeature m-1 ,normFactorFeature m}:factorMX ij ={factor ijk |factor ijk ≥factor ijs ,k≠s,1≤k,s≤n}, factorMN ij ={factor ijk |factor ijk ≤factor ijs ,k≠s,1≤k,s≤n},

[0081] Construct the DenseNet model as follows, using cross entropy as the loss function and Adam as the optimizer:

[0082] Basic model: DensenNet121 pre-trained model, freezing the convolutional layer of DenseNet;

[0083] Global average pooling layer;

[0084] The first hidden layer contains 32 neurons and the activation function is ReLU.

[0085] The second hidden layer contains 32 neurons and the activation function is LReLU.

[0086] The output layer contains 5 neurons and the activation function is softmax.

[0087] Traverse the standardized energy consumption factor feature matrix set normFactorFeature of all regions, and for the standardized energy consumption factor feature matrix normFactorFeature of the i-th region i The dataset is randomly divided into 70% training set, 10% test set and 20% validation set.

[0088] Obtain all regional sample labeling files {label1.txt, label2.txt, …, labeli.txt, …, label1.txt, labelm.txt} and the standardized energy consumption factor feature matrix set normFactorFeature for all regions at n moments of energy consumption influencing factor data in all regions, and use the DenseNet model for training. Set the parameters as follows. After the training is completed, the energy consumption anomaly classification model areaDenseNet {areaDenseNet 1 ,areaDenseNet 2 ,…,areaDenseNet i ,…,areaDenseNet m-1 ,areaDenseNet m}, where areaDenseNet i For the energy consumption anomaly classification model of the i-th region: max_epochs=1000, batch_size=64, initial_learning_rate=0.1, learningRateScheduler=step decay, loss=categorical_crossentropy, verbose=1.

[0089] Optionally, an unreasonable energy consumption analysis is performed on the m sub-regions based on the m second real-time energy consumption data and the m first predicted energy consumption data, and t areas of suspected unreasonable energy consumption among the m sub-regions are determined, including: determining the first-level alarm threshold of the m sub-regions to obtain m first-level alarm thresholds; calculating the absolute value of the difference between the second real-time energy consumption data and the first predicted energy consumption data corresponding to each sub-region to obtain m absolute values; determining the t areas of suspected unreasonable energy consumption based on the m first-level alarm thresholds and the m absolute values, wherein the absolute value of the area of ​​suspected unreasonable energy consumption is greater than the first-level alarm threshold.

[0090] Traverse the real-time energy consumption prediction value set energyPredict of all regions and the first-level alarm threshold thr of all regions = {thr 1 ,thr 2 ,…,thr i ,…,thr m-1 ,thr m}, the real-time energy consumption prediction value energyPredict for the i-th area i and the first level alarm threshold thr of the i-th area i , when |energyPredict i -nowElectricityConsumption iT |>thr i , set the first level warning flag warning for the ith area i The value is true, and the set of all regional level 1 alarms warning = {warning 1 ,warning 2 ,…,warning i ,…,warning m-1 ,warning m}.

[0091]

[0092] (i.e., t areas where energy consumption is suspected to be unreasonable).

[0093] Optionally, after classifying the t energy consumption influencing factor data respectively through t energy consumption anomaly classification models to obtain t classification results, the method also includes: when r classification results among the t classification results indicate that the area is normal, determining the r energy consumption suspected unreasonable areas corresponding to the r classification results as normal areas, wherein r is a positive integer; and correcting the r first-level alarm thresholds corresponding to the r energy consumption suspected unreasonable areas.

[0094] The process of classifying t energy consumption influencing factor data respectively through t energy consumption anomaly classification models includes:

[0095] Traverse the level 1 alarm set warning, when the level 1 alarm indicator warning of the i-th area i The value of is true (that is, the area is determined to be a first-level alarm area, that is, an area where energy consumption is suspected to be unreasonable), and the real-time energy consumption influencing factor data nowEnergyRelevant of the i-th area is collected. i ={nowFactor i1 ,nowFactor i2 ,…,nowFactor ij ,…,nowFactor il-1 ,nowFactor il}, where nowFactor ij is the jth real-time energy consumption impact factor of the ith region. Traverse the real-time energy consumption impact factor data of the ith region nowEnergyRelevant i , according to the following formula (Min-Max) normalization, the standardized real-time energy consumption influencing factor characteristic of the ith area is nowFactorNorm i = {nowFactorNorm i1 ,nowFactorNorm i2 ,…,nowFactorNorm ij ,…,nowFactorNorm il-1 ,nowFactorNorm il}, nowFactorNorm i Input to the energy consumption anomaly classification model areaDenseNet of the i-th area after training i Perform abnormal classification, and the output result is the abnormal classification result of the real-time energy consumption in the i-th area. i , abnormal i The values ​​include "Device Abnormal", "System Abnormal", "Operation Abnormal", "Environment Abnormal", and "Normal".

[0096] When abnormal i When the value is 'device abnormality' or 'system abnormality' or 'operation abnormality' or 'environmental abnormality', set the secondary alarm flag UrgentWarning of the i-th area i The value is true; when abnormal i When the value is 'normal', modify the first level alarm threshold thr of the i-th area according to the following formula i Values: where factorMX ij is the energy consumption factor matrix FactorMatrix of the i-th region i The maximum value of the jth column in factorMN ij Energy consumption factor matrix of the i-th region FactorMatrix i The minimum value of the jth column in .

[0097] Through the above embodiments, this application proposes a smart park energy consumption warning method based on GRU and DenseNet. Compared with the general energy consumption warning algorithm, the main advantage of this method is: it adopts a two-level warning mechanism. At the first level, short-term energy consumption characteristics are extracted based on energy consumption fragments to predict the real-time energy consumption of the region, and GRU is used for energy consumption prediction, which reduces the energy consumption of normalized energy consumption prediction. At the second level, for areas with abnormal energy consumption, DenseNet is used to classify the anomalies based on energy consumption influencing factors, which reduces the resource consumption of energy consumption anomaly classification. The first-level alarm threshold is adaptively adjusted according to the secondary classification results, which improves the accuracy of energy consumption warnings.

[0098] The application structure of the early warning method is as follows Figure 3 As shown in the figure: the smart park includes different areas such as office area, production area, life service area, logistics area, information center, control center, energy management area, etc. On the one hand, data is collected from these areas to obtain historical power consumption and real-time power consumption. The historical power consumption is preprocessed to obtain a data set for the model training process of the energy consumption prediction model. The trained energy consumption prediction model is combined with the real-time power consumption to predict energy consumption and obtain the real-time predicted energy consumption. If the difference between the real-time power consumption and the real-time predicted energy consumption exceeds the first-level alarm threshold, a first-level alarm is issued for the corresponding area. On the other hand, energy consumption related factors are collected from these areas. The energy consumption related factors include parameters such as temperature, humidity, wind speed, voltage, current, and electric power. After data preprocessing, the data is obtained to obtain a data set for energy consumption anomaly classification model training. After the first-level alarm is issued in the area, the real-time energy consumption factor data of the area is collected and classified using the energy consumption anomaly classification model. If the classification result is normal, the first-level alarm threshold is updated. If the classification result is abnormal (including equipment abnormality, system abnormality, operation abnormality, and environmental abnormality), a second-level alarm is issued for the area.

[0099] Specifically, different areas in a smart park have different energy consumption characteristics, so it is necessary to routinely collect, monitor, and judge anomalies in the energy consumption data and energy consumption-related factor data of all areas. While pursuing high accuracy, general energy consumption anomaly judgment methods ignore the problems of excessive computational complexity or excessive resource consumption. Energy consumption data has the characteristics of time, periodicity, and streaming data. Within the same length of time, the energy consumption levels in different months, dates, and time periods are different. Statistics based only on energy consumption fragments cannot capture the relationship between energy consumption and time points. There are many types of energy consumption anomalies. After judging the energy consumption anomaly, further anomaly classification can help determine the anomaly category, thereby improving the pertinence and efficiency of maintenance.

[0100] The early warning method described in the above embodiment adopts a two-level early warning mechanism. At the first level, after obtaining the historical energy consumption data of each region, the energy consumption data is sliced ​​according to the specified time slice, and the minimum value, maximum value, moving average, moving variance and integration time are extracted based on the energy consumption fragments. These features are used to train the energy consumption prediction model using GRU. Then, the real-time electric energy consumption of each region is predicted using the trained energy consumption prediction model for each region. By comparing the difference between the real-time measured electric energy consumption of each region and the real-time electric energy consumption of each region with the first-level alarm threshold of each region, the energy consumption abnormality of each region is judged. At the second level, after obtaining the historical energy consumption influencing factor data of each region, the data is standardized, and the energy consumption abnormality classification model is trained using DenseNet. For areas with abnormal energy consumption, the energy consumption abnormality category of the area is determined using the trained energy consumption abnormality classification model for the area. If the category is normal, the first-level energy consumption threshold of the area is adaptively adjusted, otherwise a second-level alarm is issued.

[0101] This smart park energy consumption warning method based on GRU and DenseNet, in the first level of energy consumption prediction, extracts the minimum value, maximum value, moving average, moving variance and integration time based on the energy consumption fragment to predict the real-time energy consumption of the region. The first four features reflect the short-term fluctuations of energy consumption, and the integration time feature helps to identify the periodic pattern of energy consumption, which improves the comprehensive grasp of the local changes and overall trends of energy consumption by the energy consumption prediction model. In order to reduce the energy consumption of normalized energy consumption prediction, GRU with few parameters, fast training speed and low memory usage is used for energy consumption prediction. In the second level of energy consumption classification, for areas with abnormal energy consumption, the anomalies are classified based on the energy consumption influencing factors. The DenseNet used can reuse features and significantly reduce the number of parameters, which reduces the resource consumption of energy consumption anomaly classification. The first level alarm threshold is adaptively adjusted according to the second level classification results, which improves the accuracy of energy consumption warning.

[0102] In an optional embodiment, the following combination Figure 4 The processing flow of the above-mentioned smart park energy consumption early warning method is further explained, such as Figure 4As shown, the following steps are included:

[0103] Step 1. Start;

[0104] Step 2: Obtain the power consumption data of each area;

[0105] Step 3: Obtain the timestamp sequence corresponding to the electric energy consumption data sequence of each area;

[0106] Step 4, traverse the power consumption data of each area;

[0107] Step 5: Divide by time window T and store by row into the energy consumption fragment matrix of region i to obtain a set of energy consumption fragment matrices of each region;

[0108] Step 6: traverse the timestamp sequence;

[0109] Step 7: Calculate the integrated time series;

[0110] Step 8, traverse the energy consumption fragment matrix set of each area;

[0111] Step 9, calculate the characteristics of each region and each fragment to obtain the energy consumption characteristic matrix set of all regions;

[0112] Step 10, traverse the energy consumption feature set;

[0113] Step 11, standardize all energy consumption characteristics of each area to obtain a standardized feature matrix set;

[0114] Step 12: Build the GRU model;

[0115] Step 13, divide the standardized feature matrix set into a training set, a validation set, and a test set;

[0116] Step 14: Use the GRU model to train and obtain the energy consumption prediction model for each area;

[0117] Step 15: Obtain the real-time energy consumption data sequence of each area at T moments;

[0118] Step 16: Get the current timestamp.

[0119] Step 17, traverse the real-time energy consumption data sequence;

[0120] Step 18: Calculate the current integration time;

[0121] Step 19: Calculate and standardize the real-time energy consumption characteristics of each area to obtain a standardized real-time characteristic matrix set of all areas;

[0122] Step 20, traverse the standardized real-time feature matrix set;

[0123] Step 21: Use the trained GRU of each region to predict the real-time energy consumption of each region to obtain a set of real-time energy consumption prediction values;

[0124] Step 22, traverse the real-time energy consumption prediction value sets of all regions;

[0125] Step 23, determine whether the traversal is completed, if so, execute step 26; if not, execute step 24;

[0126] Step 24: determine whether the difference between the real-time energy consumption prediction value and the real-time energy consumption data is greater than the first-level alarm threshold; if so, execute step 25; if not, execute step 23;

[0127] Step 25: Set the value of the first-level warning flag warning of area i to true;

[0128] Step 26: Obtain energy consumption influencing factor data for each region;

[0129] Step 27, traverse the energy consumption influencing factor data of each area;

[0130] Step 28, storing the energy consumption factor sequences of all regions into a matrix by column to obtain an energy consumption factor matrix set;

[0131] Step 29, traverse the energy consumption factor matrix set;

[0132] Step 30, standardize the energy consumption factor matrix sequence of each region to obtain a standardized energy consumption factor characteristic matrix set;

[0133] Step 31: Build a DenseNet model.

[0134] Step 32, dividing the standardized energy consumption factor feature matrix set into a training set, a validation set, and a test set;

[0135] Step 33, obtaining the sample annotation files of each region;

[0136] Step 34: Use the DenseNet model to train an abnormal energy consumption classification model for each area;

[0137] Step 35, traverse the first level alarm set;

[0138] Step 36, determine whether the traversal of the first-level alarm set is complete; if so, execute step 44; if not, execute step 37;

[0139] Step 37, determine whether the warning value of the first-level warning flag of the area is true; if so, execute step 38; if not, execute step 36;

[0140] Step 38: Obtain real-time energy consumption influencing factor data of the i-th area;

[0141] Step 39, standardize and obtain the standardized real-time energy consumption influencing factor characteristics of region i;

[0142] Step 40: Use the trained DenseNet to classify the energy consumption anomaly in the area to obtain the real-time energy consumption anomaly classification result of area i;

[0143] Step 41, determine whether the classification result is normal; if so, execute step 43; if not, execute step 42;

[0144] Step 42, setting the value of the second-level alarm flag of the i-th area to true;

[0145] Step 43: modify the value of the first-level alarm threshold of the i-th area;

[0146] Step 44, end.

[0147] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of each embodiment of the present application.

[0148] In this embodiment, a device for determining an unreasonable energy consumption area is also provided, and the device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0149] Figure 5 : is a structural block diagram of a device for determining an unreasonable energy consumption area according to an embodiment of the present application, the device comprising:

[0150] The first collection module 52 is used to periodically collect real-time energy consumption data of m sub-areas of the target area according to the time window T to obtain m first real-time energy consumption data, wherein m is a positive integer, and the first real-time energy consumption data is the real-time energy consumption data within the current time window;

[0151] A prediction module 54, configured to predict the energy consumption data of the m sub-regions in the next time window respectively by using the m energy consumption prediction models and the m first real-time energy consumption data, to obtain m first predicted energy consumption data, wherein the m energy consumption prediction models correspond one-to-one to the m sub-regions;

[0152] The analysis module 56 is used to perform unreasonable energy consumption analysis on the m sub-regions according to the m second real-time energy consumption data and the m first predicted energy consumption data, and determine t suspected unreasonable energy consumption regions in the m sub-regions, where t is a positive integer and t is less than or equal to m;

[0153] The second acquisition module 58 is used to collect the energy consumption influencing factor data of the t energy consumption suspected unreasonable areas in the current time window, and classify the t energy consumption influencing factor data respectively through t energy consumption abnormality classification models to obtain t classification results, wherein the t energy consumption abnormality classification models correspond to the t energy consumption suspected unreasonable areas one by one;

[0154] The determination module 60 is used to determine the p suspected unreasonable energy consumption areas corresponding to the p classification results as unreasonable energy consumption areas when p classification results among the t classification results indicate that the area is abnormal, wherein p is a positive integer.

[0155] Through the above-mentioned device, the real-time energy consumption data of m sub-areas of the target area are first collected according to a fixed time window T, and the first real-time energy consumption data in the m current time windows are obtained, where m is a positive integer; the energy consumption data of the next time window are predicted for the first real-time energy consumption data through the energy consumption prediction model of each sub-area, and m first predicted energy consumption data are obtained; then, the unreasonable energy consumption of the m sub-areas is analyzed according to the m second real-time energy consumption data and the m first predicted energy consumption data, and t unreasonable energy consumption areas are determined; the energy consumption influencing factor data of the t suspected unreasonable energy consumption areas in the current time window are collected, and they are classified through t energy consumption anomaly classification models. class, and obtain t classification results; when there are p classification results indicating regional abnormalities, the corresponding p areas of suspected unreasonable energy consumption are determined as unreasonable energy consumption areas; using the above scheme, energy consumption prediction models and energy consumption anomaly classification models are trained for different areas respectively, and the next time window is predicted according to the energy consumption data of different areas in the current time window, and then the difference between the actual energy consumption data and the predicted results is combined to determine whether there is an abnormality. If there is an abnormality, a secondary judgment is further made according to the energy consumption anomaly classification model to accurately determine whether the area has an energy consumption abnormality problem; thereby solving the problem of the lack of an effective smart park energy consumption abnormality early warning method in the relevant technology.

[0156] Optionally, the prediction module 54 is further used to obtain historical energy consumption data of the m sub-areas; perform data preprocessing on the m historical energy consumption data to obtain m feature data; and train the m energy consumption prediction models according to the m feature data.

[0157] Optionally, the above-mentioned prediction module 54 is also used to obtain a timestamp sequence corresponding to n moments, and perform weighted merging processing on each timestamp to obtain an integrated time series, wherein n is a positive integer, the timestamp sequence includes multiple timestamps, and the integrated time series includes multiple integration times, and the multiple timestamps and the multiple integration times correspond one to one; the m historical energy consumption data are divided according to the time window T, and energy consumption features are extracted for the divided m historical energy consumption data according to the multiple integration times to obtain m energy consumption feature matrices, wherein the energy consumption features include: maximum value, minimum value, moving mean, and moving variance; the m energy consumption feature matrices are feature normalized to obtain the m feature data.

[0158] Optionally, the second acquisition module 58 is also used to collect the energy consumption influencing factor data of the m sub-areas; perform data preprocessing on the m energy consumption influencing factor data to obtain m energy consumption factor characteristic data; and train m energy consumption anomaly classification models according to the m energy consumption factor characteristic data, wherein the m energy consumption anomaly classification models include the t energy consumption anomaly classification models.

[0159] Optionally, the above-mentioned analysis module 56 is also used to determine the first-level alarm thresholds of the m sub-areas to obtain m first-level alarm thresholds; calculate the absolute value of the difference between the second real-time energy consumption data and the first predicted energy consumption data corresponding to each sub-area to obtain m absolute values; determine the t suspected unreasonable energy consumption areas based on the m first-level alarm thresholds and the m absolute values, wherein the absolute value of the suspected unreasonable energy consumption area is greater than the first-level alarm threshold.

[0160] Optionally, the second acquisition module 58 is further used to determine r areas of suspected unreasonable energy consumption corresponding to the r classification results as normal areas when r of the t classification results indicate that the area is normal, wherein r is a positive integer; and to correct the r first-level alarm thresholds corresponding to the r areas of suspected unreasonable energy consumption.

[0161] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0162] Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps:

[0163] S1, regularly collecting real-time energy consumption data of m sub-areas of the target area according to the time window T, and obtaining m first real-time energy consumption data, wherein m is a positive integer, and the first real-time energy consumption data is the real-time energy consumption data within the current time window;

[0164] S2, predicting the energy consumption data of the m sub-regions in the next time window respectively by using m energy consumption prediction models and the m first real-time energy consumption data, to obtain m first predicted energy consumption data, wherein the m energy consumption prediction models correspond one-to-one to the m sub-regions;

[0165] S3, performing unreasonable energy consumption analysis on the m sub-regions according to the m second real-time energy consumption data and the m first predicted energy consumption data, and determining t suspected unreasonable energy consumption regions in the m sub-regions, where t is a positive integer and t is less than or equal to m;

[0166] S4, collecting the energy consumption influencing factor data of the t energy consumption suspected unreasonable areas in the current time window, and classifying the t energy consumption influencing factor data respectively through t energy consumption abnormality classification models to obtain t classification results, wherein the t energy consumption abnormality classification models correspond one-to-one to the t energy consumption suspected unreasonable areas;

[0167] S5, when p classification results among the t classification results indicate that the area is abnormal, the p suspected unreasonable energy consumption areas corresponding to the p classification results are determined as unreasonable energy consumption areas, wherein p is a positive integer.

[0168] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0169] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0170] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0171] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0172] S1, regularly collecting real-time energy consumption data of m sub-areas of the target area according to the time window T, and obtaining m first real-time energy consumption data, wherein m is a positive integer, and the first real-time energy consumption data is the real-time energy consumption data within the current time window;

[0173] S2, predicting the energy consumption data of the m sub-regions in the next time window respectively by using m energy consumption prediction models and the m first real-time energy consumption data, to obtain m first predicted energy consumption data, wherein the m energy consumption prediction models correspond one-to-one to the m sub-regions;

[0174] S3, performing unreasonable energy consumption analysis on the m sub-regions according to the m second real-time energy consumption data and the m first predicted energy consumption data, and determining t suspected unreasonable energy consumption regions in the m sub-regions, where t is a positive integer and t is less than or equal to m;

[0175] S4, collecting the energy consumption influencing factor data of the t energy consumption suspected unreasonable areas in the current time window, and classifying the t energy consumption influencing factor data respectively through t energy consumption abnormality classification models to obtain t classification results, wherein the t energy consumption abnormality classification models correspond one-to-one to the t energy consumption suspected unreasonable areas;

[0176] S5, when p classification results among the t classification results indicate that the area is abnormal, the p suspected unreasonable energy consumption areas corresponding to the p classification results are determined as unreasonable energy consumption areas, wherein p is a positive integer.

[0177] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0178] An embodiment of the present application further provides a computer program product, comprising a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program product, and when the computer program is executed by a processor, the steps of the method described in each embodiment of the present application are implemented.

[0179] Optionally, in this embodiment, the above computer program may be configured to implement the following steps when executed by a processor:

[0180] S1, regularly collecting real-time energy consumption data of m sub-areas of the target area according to the time window T, and obtaining m first real-time energy consumption data, wherein m is a positive integer, and the first real-time energy consumption data is the real-time energy consumption data within the current time window;

[0181] S2, predicting the energy consumption data of the m sub-regions in the next time window respectively by using m energy consumption prediction models and the m first real-time energy consumption data, to obtain m first predicted energy consumption data, wherein the m energy consumption prediction models correspond one-to-one to the m sub-regions;

[0182] S3, performing unreasonable energy consumption analysis on the m sub-regions according to the m second real-time energy consumption data and the m first predicted energy consumption data, and determining t suspected unreasonable energy consumption regions in the m sub-regions, where t is a positive integer and t is less than or equal to m;

[0183] S4, collecting the energy consumption influencing factor data of the t energy consumption suspected unreasonable areas in the current time window, and classifying the t energy consumption influencing factor data respectively through t energy consumption abnormality classification models to obtain t classification results, wherein the t energy consumption abnormality classification models correspond one-to-one to the t energy consumption suspected unreasonable areas;

[0184] S5, when p classification results among the t classification results indicate that the area is abnormal, the p suspected unreasonable energy consumption areas corresponding to the p classification results are determined as unreasonable energy consumption areas, wherein p is a positive integer.

[0185] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0186] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0187] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for determining an unreasonable energy consumption area, characterized in that: include: Regularly collect real-time energy consumption data of m sub-areas of the target area according to the time window T to obtain m first real-time energy consumption data, wherein m is a positive integer, and the first real-time energy consumption data is the real-time energy consumption data within the current time window; The energy consumption data of the m sub-regions in the next time window are predicted respectively by using the m energy consumption prediction models and the m first real-time energy consumption data to obtain m first predicted energy consumption data, wherein the m energy consumption prediction models correspond to the m sub-regions one by one; Performing unreasonable energy consumption analysis on the m sub-regions according to the m second real-time energy consumption data and the m first predicted energy consumption data, and determining t suspected unreasonable energy consumption regions in the m sub-regions, where t is a positive integer and t is less than or equal to m; Collecting the energy consumption influencing factor data of the t energy consumption suspected unreasonable areas in the current time window, and classifying the t energy consumption influencing factor data respectively through t energy consumption abnormality classification models to obtain t classification results, wherein the t energy consumption abnormality classification models correspond one-to-one to the t energy consumption suspected unreasonable areas; When p classification results among the t classification results indicate that the area is abnormal, the p suspected unreasonable energy consumption areas corresponding to the p classification results are determined as unreasonable energy consumption areas, where p is a positive integer.

2. The method according to claim 1, characterized in that Before respectively predicting the energy consumption data of the m sub-areas in the next time window by using the m energy consumption prediction models and the m first real-time energy consumption data to obtain the m first predicted energy consumption data, the method further includes: Obtaining historical energy consumption data of the m sub-areas; Performing data preprocessing on the m historical energy consumption data to obtain m characteristic data; The m energy consumption prediction models are trained respectively according to the m feature data.

3. The method according to claim 2, characterized in that The m historical energy consumption data are preprocessed to obtain m characteristic data, including: Obtain a timestamp sequence corresponding to n moments, and perform weighted merging processing on each timestamp to obtain an integrated time sequence, wherein n is a positive integer, the timestamp sequence includes multiple timestamps, the integrated time sequence includes multiple integration times, and the multiple timestamps correspond to the multiple integration times in one-to-one correspondence; Dividing the m historical energy consumption data according to the time window T, and extracting energy consumption features from the divided m historical energy consumption data according to the multiple integration times, to obtain m energy consumption feature matrices, wherein the energy consumption features include: maximum value, minimum value, moving mean, and moving variance; The m energy consumption feature matrices are normalized to obtain the m feature data.

4. The method according to claim 1, characterized in that The method further comprises: classifying the t energy consumption influencing factor data respectively by using t energy consumption abnormality classification models to obtain t classification results: Collecting energy consumption influencing factor data of the m sub-areas; Performing data preprocessing on the m energy consumption influencing factor data to obtain m energy consumption factor characteristic data; According to the m energy consumption factor characteristic data, m energy consumption anomaly classification models are trained respectively, wherein the m energy consumption anomaly classification models include the t energy consumption anomaly classification models.

5. The method according to claim 1, characterized in that Performing unreasonable energy consumption analysis on the m sub-regions according to the m second real-time energy consumption data and the m first predicted energy consumption data, and determining t suspected unreasonable energy consumption regions in the m sub-regions, including: Determine the first-level alarm thresholds of the m sub-areas to obtain m first-level alarm thresholds; Calculate the absolute value of the difference between the second real-time energy consumption data and the first predicted energy consumption data corresponding to each sub-region to obtain m absolute values; The t areas of suspected unreasonable energy consumption are determined according to the m first-level alarm thresholds and the m absolute values, wherein the absolute values ​​of the areas of suspected unreasonable energy consumption are greater than the first-level alarm thresholds.

6. The method according to claim 5, characterized in that After classifying the t energy consumption influencing factor data respectively through t energy consumption abnormality classification models to obtain t classification results, the method further includes: When r of the t classification results indicate that the area is normal, r areas of suspected unreasonable energy consumption corresponding to the r classification results are determined as normal areas, where r is a positive integer; The r first-level alarm thresholds corresponding to the r areas of suspected unreasonable energy consumption are corrected.

7. A device for determining unreasonable energy consumption areas, characterized in that: include: A first collection module is used to periodically collect real-time energy consumption data of m sub-areas of the target area according to a time window T, and obtain m first real-time energy consumption data, wherein m is a positive integer, and the first real-time energy consumption data is the real-time energy consumption data within the current time window; A prediction module, used to predict the energy consumption data of the m sub-regions in the next time window respectively through m energy consumption prediction models and the m first real-time energy consumption data, and obtain m first predicted energy consumption data, wherein the m energy consumption prediction models correspond to the m sub-regions one by one; an analysis module, used to perform unreasonable energy consumption analysis on the m sub-regions according to the m second real-time energy consumption data and the m first predicted energy consumption data, and determine t suspected unreasonable energy consumption regions in the m sub-regions, wherein t is a positive integer, and t is less than or equal to m; The second acquisition module is used to collect the energy consumption influencing factor data of the t energy consumption suspected unreasonable areas in the current time window, and classify the t energy consumption influencing factor data respectively through t energy consumption abnormality classification models to obtain t classification results, wherein the t energy consumption abnormality classification models correspond one-to-one to the t energy consumption suspected unreasonable areas; A determination module is used to determine the p suspected unreasonable energy consumption areas corresponding to the p classification results as unreasonable energy consumption areas when p classification results among the t classification results indicate that the area is abnormal, wherein p is a positive integer.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program executes the method according to any one of claims 1 to 6 when executed.

9. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 6 through the computer program.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images