A method and system for online evaluation of energy consumption data accuracy

Abstract

The application discloses a kind of energy consumption data accuracy online evaluation method and system, it is related to data evaluation technical field, including collection port connection customer energy data management system, collect and establish historical energy consumption database, and acquire energy consumption data according to preset time interval;The energy consumption data is preprocessed, and the BP neural network model parameter is initialized based on historical data sample;The data after pre-processing is input into BP neural network, and the BP neural network is trained, the model parameter is optimized by minimizing the training error of sample, and the BP neural network model satisfying preset condition;The trained BP neural network model is applied to predict the newly incoming energy consumption data, calculate the absolute error between predicted value and actual value, whether it is abnormal value according to preset threshold value is judged, while providing correction suggestion and dynamically adjusting abnormality judgment threshold value.The application reduces the potential loss caused by data error.

Description

Technical Field

[0001] This invention relates to the field of data evaluation technology, and in particular to an online method and system for evaluating the accuracy of energy consumption data. Background Technology

[0002] The patent, "A Method for Streaming Data Quality Inspection of Energy Equipment Based on Flinkx," provides a method for inspecting the quality of streaming data from energy equipment based on Flinkx. First, a Flinkx job is initialized, and a source reader is built to read the data to be inspected. Then, a rule engine processing end is created to process the data. Finally, a sink writer writes the verification result stream data and the statistical result stream data. The rule engine processing end of this invention uses the Rete algorithm for forward inference to quickly verify the data to be inspected, ensuring the stability of the verification under high throughput and low latency. For abnormal data, the system provides three inspection methods: null value inspection, value range inspection, and fluctuation inspection. However, due to the use of a one-size-fits-all value range inspection method, its accuracy needs improvement. The paper, "Research on Power Data Quality Assessment Based on Analytic Hierarchy Process," uses the analytic hierarchy process (AHP) to establish a data quality evaluation index system and evaluation model, achieving effective evaluation of power data quality. This helps power grid companies to promptly identify and resolve power data quality problems, improving the practicality of power metering systems and the level of power data quality management. It constructs evaluation rules to evaluate index scores and uses a "1-9" scaling method to process index weights. However, this method heavily relies on human experience in scoring and weighting indicators, incurring significant labor costs and resulting in low efficiency. Its accuracy and professionalism also need to be considered. The patent, "A Method for Evaluating the Quality of Online Power Quality Monitoring Data," proposes a method to effectively and conveniently filter out "distorted" data, ensuring the reliability of online power quality monitoring data. This method checks for outliers based on the rate of change of k data points before and after the check. This method requires manual determination of the screening threshold and k, which is subject to chance. Furthermore, the data used to calculate the rate of change is limited. The research paper, "Research on the Evaluation Method of Online Energy Consumption Monitoring Data Quality," constructs multiple dimensions of data evaluation indicators, including completeness, accuracy, and compliance. Each dimension is scored separately, and the weighted summation of the scores by different organizations is used as the final score. However, when evaluating data accuracy, the evaluation object is the quality of the measuring instruments, making the assessment of data accuracy insufficiently intuitive and accurate.

[0003] Current research in data quality monitoring primarily relies on relatively traditional methods. These methods often lack intuitiveness and effectiveness in assessing data accuracy. This lag can lead to overlooked or delayed processing of data errors and negatively impact the final data quality assessment results. Therefore, existing data quality monitoring methods urgently need improvement and innovation to identify and handle anomalies in data more efficiently and accurately, thereby ensuring the reliability and stability of system data quality. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides an online method and system for evaluating the accuracy of energy consumption data, which can solve the problems mentioned in the background art.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] In a first aspect, the present invention provides an online method for evaluating the accuracy of energy consumption data, which includes: connecting a collection port to a customer energy data management system, collecting and establishing a historical energy consumption database, and acquiring energy consumption data at preset time intervals;

[0009] The energy consumption data is preprocessed, and the parameters of the BP neural network model are initialized based on historical data samples;

[0010] The preprocessed data is input into the BP neural network and trained. The model parameters are optimized by minimizing the training error of the samples to achieve a BP neural network model that meets the preset conditions.

[0011] The trained BP neural network model is used to predict newly input energy consumption data, calculate the absolute error between the predicted value and the actual value, determine whether it is an outlier based on a preset threshold, and provide correction suggestions and dynamically adjust the outlier judgment threshold.

[0012] As a preferred embodiment of the online energy consumption data accuracy assessment method of the present invention, the step of collecting and establishing a historical energy consumption database includes:

[0013] Establish an interface with the customer's energy data management system to acquire and store historical energy consumption data.

[0014] The historical energy consumption data includes electricity consumption, water consumption, coal consumption, natural gas consumption, and oil consumption.

[0015] According to preset time intervals, energy consumption data is updated from the customer's energy data management system to the historical energy consumption database.

[0016] As a preferred embodiment of the online energy consumption data accuracy assessment method of the present invention, the preprocessing of the obtained data includes:

[0017] The collected energy consumption data is normalized, and the normalization formula is as follows:

[0018]

[0019] in, x represents the normalized data. j (i) represents the original data, max(x) j ) and min(x j () represent the minimum and maximum values ​​in the data, respectively;

[0020] The normalized data is divided into training and testing sets according to a set ratio.

[0021] As a preferred embodiment of the online energy consumption data accuracy assessment method of the present invention, the step of training the BP neural network includes:

[0022] Initialize the learning rate, batch size, number of iterations, number of hidden layers, and number of nodes of the BP neural network;

[0023] Use the rand function to initialize all connection weights and thresholds such that w,θ∈(0,1); where w represents the connection weight and θ represents the threshold.

[0024] The hidden layer uses the sigmoid tangent function (tansig) as the transfer function, i.e.,

[0025]

[0026] Where x represents the input signal;

[0027] The output layer uses logsig as the transfer function, that is,

[0028]

[0029] The trainlm algorithm is used as the training function;

[0030] The optimization objective is defined as minimizing the training error of the samples, where the training error is calculated using the following formula:

[0031]

[0032] Where y represents the actual value, Indicates the predicted value;

[0033] The parameters are optimized using gradient descent, and the parameter update formula is as follows:

[0034]

[0035] Where w(t) represents the current weight, and η represents the learning rate. This represents the gradient of the training error.

[0036] As a preferred embodiment of the online energy consumption data accuracy assessment method of the present invention, the determination of whether it is an outlier includes:

[0037] The normalized output is then restored using the inverse normalization formula, as shown in the following formula:

[0038] x=x′·(x max -x min )+x min

[0039] Where x′ represents the normalized data;

[0040] Calculate the absolute error between the predicted and actual values, and calculate the root mean square error using the following formula:

[0041]

[0042] Among them, y i This represents the i-th actual value. Let represent the i-th predicted value, n represent the number of samples, and RMSE represent the root mean square error.

[0043] The outlier tolerance P is set, and the calculation formula is as follows:

[0044]

[0045] Where P represents the outlier tolerance;

[0046] Determine whether the value exceeds a preset threshold. If it does, the data point is considered an outlier; otherwise, it is not.

[0047] As a preferred embodiment of the online energy consumption data accuracy assessment method of the present invention, the provision of correction suggestions and data quality assessment reports includes,

[0048] When an outlier is detected, the system sends a warning message to the client and provides a predicted value.

[0049] Regularly generate key data quality reports for each system, analyze and display data quality fluctuations and rule proportions.

[0050] As a preferred embodiment of the online energy consumption data accuracy assessment method of the present invention, the dynamic adjustment of the anomaly judgment threshold includes:

[0051] Adjust the outlier tolerance P setting;

[0052] The data correction suggestions provided based on deep learning from historical data include,

[0053] Based on the learning results from historical data, corresponding numerical modifications are made according to the detected outliers.

[0054] Secondly, the present invention provides an online evaluation system for the accuracy of energy consumption data, which includes: a data acquisition and storage module, a model initialization module, a model training module, and an anomaly detection module;

[0055] The data acquisition and storage module is used to collect data from the customer's energy data management system via a port connection, collect and establish a historical energy consumption database, and acquire energy consumption data at preset time intervals.

[0056] The model training module is used to preprocess the energy consumption data and initialize the BP neural network model parameters based on historical data samples.

[0057] The model training module is used to input the preprocessed data into the BP neural network and train the BP neural network. By minimizing the training error of the samples, the model parameters are optimized to achieve a BP neural network model that meets the preset conditions.

[0058] The anomaly detection module is used to predict newly input energy consumption data using a trained BP neural network model, calculate the absolute error between the predicted value and the actual value, determine whether it is an anomaly based on a preset threshold, and provide correction suggestions and dynamically adjust the anomaly detection threshold.

[0059] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of an online method for evaluating the accuracy of energy consumption data.

[0060] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements the steps of an online method for evaluating the accuracy of energy consumption data.

[0061] Compared with existing technologies, the beneficial effects of this invention are as follows: This patent aims to provide users with a highly customized and accurate information accuracy assessment solution. It not only cleverly utilizes massive historical information data of enterprises as a foundation to more accurately judge anomalies in their energy consumption monitoring data, but also supports timely updates to anomaly assessment thresholds based on the enterprise's development status, thereby ensuring the continuous effectiveness and applicability of the method. The system can accurately examine various types of energy consumption data, supporting energy big data centers to better achieve multi-source and multi-domain data fusion, realizing efficient integration, fusion, and management of energy data, helping to build a highly efficient and technologically advanced energy big data center, providing support for data sharing and in-depth value mining, providing comprehensive analysis support for the energy industry, and helping upstream and downstream enterprises quickly build data decision-making and analysis capabilities. Simultaneously, it helps serve government scientific decision-making and power grid enterprise energy statistical supervision, facilitating targeted management by governments and power grid enterprises whose energy consumption does not meet national policy requirements, and contributing to the implementation of the "dual carbon" policy. More advancedly, this patent goes beyond anomaly warning; it further provides customers with suggestions for modifying anomalies. These suggestions are based on deep learning and understanding of historical data, aiming to help enterprise users more effectively correct data errors and improve overall data quality. In this way, the company's data quality management process can operate more efficiently and smoothly, reducing potential losses caused by data errors. Attached Figure Description

[0062] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 A flowchart illustrating an online method and system for assessing the accuracy of energy consumption data, provided in one embodiment of the present invention;

[0064] Figure 2 This is an internal structural diagram of a computer device for an online energy consumption data accuracy assessment method and system provided in one embodiment of the present invention. Detailed Implementation

[0065] To make the above-mentioned objects, features, and advantages of the present invention more readily understood, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0066] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0067] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0068] Example 1, referring to Figures 1-2 This is the first embodiment of the present invention, which provides an online method for evaluating the accuracy of energy consumption data, including:

[0069] This application provides a method that can effectively solve the problems mentioned above. The following will describe in detail how to achieve the online assessment method for the accuracy of energy consumption data with reference to several embodiments.

[0070] Figure 1 A flowchart illustrating an online method and system for assessing the accuracy of energy consumption data is provided, including:

[0071] S1: The collection port connects to the customer's energy data management system, collects and establishes a historical energy consumption database, and acquires energy consumption data at preset time intervals;

[0072] Furthermore, collecting and establishing a historical energy consumption database includes,

[0073] Establish an interface with the customer's energy data management system to acquire and store historical energy consumption data.

[0074] Historical energy consumption data includes electricity consumption, water consumption, coal consumption, natural gas consumption, and oil consumption;

[0075] According to preset time intervals, energy consumption data is updated from the customer's energy data management system to the historical energy consumption database.

[0076] S2: Preprocess the energy consumption data and initialize the BP neural network model parameters based on historical data samples;

[0077] Furthermore, the preprocessing of the acquired data includes,

[0078] The collected energy consumption data is normalized, and the normalization formula is as follows:

[0079]

[0080] in, x represents the normalized data. j (i) represents the original data, max(x) j ) and min(x j () represent the minimum and maximum values ​​in the data, respectively;

[0081] The normalized data is divided into training and testing sets according to a set ratio.

[0082] S3: Input the preprocessed data into the BP neural network and train the BP neural network. Optimize the model parameters by minimizing the training error of the samples to obtain a BP neural network model that meets the preset conditions.

[0083] Furthermore, training the BP neural network includes,

[0084] Initialize the learning rate, batch size, number of iterations, number of hidden layers, and number of nodes of the BP neural network;

[0085] Use the rand function to initialize all connection weights and thresholds such that w,θ∈(0,1); where w represents the connection weight and θ represents the threshold.

[0086] The hidden layer uses the sigmoid tangent function (tansig) as the transfer function, i.e.,

[0087]

[0088] Where x represents the input signal;

[0089] The output layer uses logsig as the transfer function, that is,

[0090]

[0091] The trainlm algorithm is used as the training function;

[0092] The optimization objective is defined as minimizing the training error of the samples, where the training error is calculated using the following formula:

[0093]

[0094] Where y represents the actual value, Indicates the predicted value;

[0095] The parameters are optimized using gradient descent, and the parameter update formula is as follows:

[0096]

[0097] Where w(t) represents the current weight, and η represents the learning rate. This represents the gradient of the training error.

[0098] S4: Apply the trained BP neural network model to predict newly input energy consumption data, calculate the absolute error between the predicted value and the actual value, determine whether it is an outlier based on the preset threshold, and provide correction suggestions and dynamically adjust the outlier judgment threshold.

[0099] Further, determining whether a value is an outlier includes:

[0100] The normalized output is then restored using the inverse normalization formula, as shown in the following formula:

[0101] x=x′·(x max -x min )+x min

[0102] Where x′ represents the normalized data;

[0103] Calculate the absolute error between the predicted and actual values, and calculate the root mean square error using the following formula:

[0104]

[0105] Among them, y i This represents the i-th actual value. Let represent the i-th predicted value, n represent the number of samples, and RMSE represent the root mean square error.

[0106] The outlier tolerance P is set, and the calculation formula is as follows:

[0107]

[0108] Where P represents the outlier tolerance;

[0109] Determine whether the value exceeds a preset threshold. If it does, the data point is considered an outlier; otherwise, it is not.

[0110] Furthermore, providing corrective suggestions and data quality assessment reports includes,

[0111] When an outlier is detected, the system sends a warning message to the client and provides a predicted value.

[0112] Regularly generate key data quality reports for each system, analyze and display data quality fluctuations and rule proportions.

[0113] Furthermore, dynamically adjusting the anomaly assessment threshold includes,

[0114] Adjust the outlier tolerance P setting;

[0115] The data correction suggestions provided based on deep learning from historical data include:

[0116] Based on the learning results from historical data, corresponding numerical modifications are made according to the detected outliers.

[0117] Furthermore, this embodiment also provides an online energy consumption data accuracy assessment system, including:

[0118] The system includes a data acquisition and storage module, a model initialization module, a model training module, and an anomaly detection module.

[0119] The data acquisition and storage module is used to collect data from the customer's energy data management system via the port connection, collect and build a historical energy consumption database, and acquire energy consumption data at preset time intervals.

[0120] The model training module is used to preprocess energy consumption data and initialize the parameters of the BP neural network model based on historical data samples.

[0121] The model training module is used to input the preprocessed data into the BP neural network and train the BP neural network. It optimizes the model parameters by minimizing the training error of the samples to create a BP neural network model that meets the preset conditions.

[0122] The anomaly detection module is used to predict newly input energy consumption data using a trained BP neural network model, calculate the absolute error between the predicted value and the actual value, determine whether it is an anomaly based on a preset threshold, and provide correction suggestions and dynamically adjust the anomaly judgment threshold.

[0123] This embodiment also provides a computer device, which may be a terminal, and its internal structure diagram may be as follows. Figure 2 As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements an online method for assessing the accuracy of energy consumption data. The display screen can be an LCD screen or an e-ink display. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0124] This embodiment also provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it performs the following steps: connecting the collection port to the customer's energy data management system, collecting and establishing a historical energy consumption database, and acquiring energy consumption data at preset time intervals.

[0125] The energy consumption data is preprocessed, and the parameters of the BP neural network model are initialized based on historical data samples;

[0126] The preprocessed data is input into the BP neural network and trained. The model parameters are optimized by minimizing the training error of the samples to achieve a BP neural network model that meets the preset conditions.

[0127] The trained BP neural network model is used to predict newly input energy consumption data, calculate the absolute error between the predicted value and the actual value, determine whether it is an outlier based on a preset threshold, and provide correction suggestions and dynamically adjust the outlier judgment threshold.

[0128] Example 2, refer to Figure 1 - Figure 2 This is the second embodiment of the present invention, which provides an online method for evaluating the accuracy of energy consumption data. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0129] To assess the accuracy of energy consumption data for a manufacturing company in Guizhou Province, hourly historical energy consumption data for the company from 1:00 AM on January 1, 2022 to 12:00 AM on September 30, 2024 was extracted from the system. The included fields are time point and energy consumption (including electricity consumption, water consumption, coal consumption, natural gas consumption, diesel and gasoline consumption, etc., which can be adjusted according to specific energy consumption types).

[0130] Taking the company's energy consumption data at 1:00 AM on January 1, 2022 as an example, the data structure is shown below:

[0131] Table 1: Daily Energy Consumption Data of a Certain Enterprise at 1:00 AM on January 1, 2022

[0132]

[0133] The following table shows the historical hourly energy consumption data of a certain enterprise from 1:00 AM on January 1, 2022 to 12:59 PM on September 30, 2024:

[0134] Table 2: Data Anomalies of a Certain Enterprise from 1:00 AM on January 1, 2022 to 12:59 PM on September 30, 2024

[0135]

[0136] The categorical variables were encoded at specific time points, and the numerical data were normalized. The training and test sets were then divided in an 8:2 ratio.

[0137] Table 3: Processed Data

[0138] Time point Water consumption Electricity consumption Coal mine consumption Natural gas consumption Oil consumption 0.08 0.72 0.43 0.58 0.23 0.74

[0139] Next, a backpropagation (BP) neural network is constructed for model training.

[0140] First, initialization parameters were set: the learning rate η was set to 0.003, epochs to 5000, and batch size to 300. Next, the number of hidden layers and hidden nodes were determined. The neuron transfer function and transmission function of the hidden layers were both set to tansig. The training function was trainlm, and the loss function was set to mean squared error. The training errors of network structures with dimensions of 6*64*5, 6*64*64*5, 6*64*64*64*5, and 6*64*64*64*64*5 were compared. Experiments showed that the training error was minimized when the network structure was 6*64*64*64*5. Then, the number of hidden nodes was determined using a grid search method. When the network structure was 6*64*128*64*5, the training error was further reduced.

[0141] Next, the activation functions tansig, logsig, ReLU, linear, tanh, and sigmoid are compared. The training error is minimized when the activation function for each layer is set to tansig, tansig, tansig, and logsig, respectively. Then, the optimization functions GDS, MGDS, ADAM, and trainlm are compared. The trainlm optimization function converges fastest and achieves the best results.

[0142] The `rand` function is used to initialize the network's connection weights. Data is then input into the network for training. Since the network output data is normalized numerical value, it is restored to its normalized form to obtain the predicted data.

[0143] The root mean square error between the predicted data and the actual data was calculated, and the results are shown in the table below:

[0144] Table 4: Root Mean Square Error of Each Variable

[0145] variable Water consumption Electricity consumption Coal consumption Natural gas consumption Oil consumption Root mean square error 2.433 3.324 2.234 3.75 3.40

[0146] Set a tolerance value, calculate Recall and FAR. The best results are achieved when the tolerance value is 3.25. The model prediction performance on the test set is shown in the table below:

[0147] Table 5: Recall and FAR data for each variable

[0148] variable Water consumption / % Electricity consumption / % Coal consumption / % Natural gas consumption / % Oil consumption / % Recall 96.4 97.5 94.9 95.7 98.3 FAR 0.7 1.2 0.9 2.3 2.1

[0149] Outliers are generally flagged, and the system sends a warning to the client, suggesting that the predicted value be used instead of the uploaded value. It also provides the accuracy assessment results of the data and can generate a data quality assessment report.

[0150] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0151] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0152] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0153] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0154] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0155] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0156] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. An online energy consumption data accuracy evaluation method, characterized in that: The method comprises the following steps: The energy consumption data is preprocessed, and the BP neural network model parameters are initialized based on historical data samples; The preprocessed data is input into the BP neural network, and the BP neural network is trained, the model parameters are optimized by minimizing the training error of the samples, and the BP neural network model meeting the preset condition is obtained; The trained BP neural network model is applied to predict new incoming energy consumption data, the absolute error between the predicted value and the actual value is calculated, it is judged whether it is an abnormal value according to the preset threshold value, and a correction suggestion is provided and the abnormal judgment threshold value is dynamically adjusted; The obtained data is preprocessed, including, The normalized data is divided into a training set and a test set according to a set proportion; wherein, denotes normalized data, x j (i) denotes raw data, max(x j ) and min(x j ) denote the minimum and maximum values in the data, respectively; The judgment whether it is an abnormal value includes, The output result after normalization is applied to the inverse normalization formula for restoration, and the calculation formula is as follows: Where x' represents the normalized data; x = x' • (x max - x min + x min The absolute error between the predicted value and the actual value is calculated, and the root mean square error is calculated, and the calculation formula is as follows: The tolerance P of the abnormal value is set, and the calculation formula is as follows: where y i represents the ith actual value, represents the ith predicted value, n represents the number of samples, and RMSE represents the root mean square error; It is judged whether it exceeds the preset threshold value, if it exceeds the preset threshold value, the data point is judged as an abnormal value, otherwise it is not. where P represents the tolerance of abnormal values; y represents the actual value, represents the predicted value, represents the absolute value of the difference between the actual value and the predicted value; RMSE represents the root mean square error; The collected historical energy consumption database includes, 2. The online energy consumption data accuracy evaluation method of claim 1, wherein: An interface with the customer energy data management system is established for obtaining historical energy consumption data and storing, The historical energy consumption data includes electricity consumption, water consumption, coal consumption, natural gas consumption and oil consumption; According to the preset time interval, the energy consumption data is updated from the customer energy data management system to the historical energy consumption database. The training of the BP neural network includes, 3. The online energy consumption data accuracy evaluation method of claim 2, wherein: The learning rate, batch size, iteration number, and hidden layer number and node number of the BP neural network are initialized; All connection weights and thresholds are initialized using the rand function, so that w, θ ∈ (0, 1); Where w represents the connection weight, and θ represents the threshold value; The hidden layer adopts the S-type tangent function tansig as the transfer function, that is, Where x represents the input signal; The output layer adopts logsig as the transmission function, that is, The trainlm algorithm is used as the training function; The optimization target is determined to be the minimization of the training error of the samples, and the training error calculation formula is as follows: The parameters are optimized by the gradient descent method, and the parameter update formula is as follows: wherein y represents the actual value, represents the predicted value; The provision of the correction suggestion and the data quality evaluation report includes, where w(t) denotes the current weight, and η denotes the learning rate, denotes the gradient of the training error.

4. The online energy consumption data accuracy evaluation method of claim 3, wherein: When an abnormal value is detected, the system sends a warning message to the client and provides a predicted value; Periodically generate key data quality reports for each system, analyze and display data quality fluctuation and rule proportion. The dynamic adjustment of the abnormal judgment threshold value includes, 5. The online energy consumption data accuracy evaluation method of claim 4, wherein: Adjust the set value of the abnormal value tolerance P; The provision of the data correction suggestion based on historical data deep learning includes, Based on the learning result of the historical data, the corresponding numerical modification is made according to the detected abnormal value. The method comprises the following steps:

6. An online energy consumption data accuracy evaluation system based on the online energy consumption data accuracy evaluation method according to any one of claims 1 to 5, characterized in that: The method comprises the following steps: The data collection storage module is configured to collect port connection customer energy data management system, collect and establish a historical energy consumption database, and obtain energy consumption data according to a preset time interval; The model training module is configured to pre-process the energy consumption data, and initialize BP neural network model parameters based on historical data samples; The model training module is configured to input the pre-processed data into the BP neural network, and train the BP neural network, optimize model parameters by minimizing sample training errors, and obtain a BP neural network model meeting a preset condition; The anomaly judgment module is configured to apply the trained BP neural network model to predict newly transmitted energy consumption data, calculate an absolute error between a predicted value and an actual value, determine whether the absolute error is an abnormal value according to a preset threshold, and provide a correction suggestion and dynamically adjust an anomaly judgment threshold. 7.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to implement the steps of the energy consumption data accuracy online evaluation method of any one of claims 1-5.

8. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to implement the steps of the energy consumption data accuracy online evaluation method of any one of claims 1-5.

Images