A power consumption prediction method and system based on a smart power meter

Abstract

The application relates to the field of data processing, and relates to a power consumption prediction method and system based on a smart electric energy meter. The method comprises the following steps: acquiring an average temperature prediction value of a next date; acquiring a planned running duration of production equipment of the next date; acquiring real power consumption values of a continuous preset number of days by using the smart electric energy meter; inputting the average temperature prediction value of the next date and the planned running duration into a preset power consumption prediction model, so that a power consumption prediction value of the next date of the enterprise is obtained. The method can improve the accuracy of the power consumption prediction result of the enterprise.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and more specifically, to a method and system for predicting power consumption based on a smart energy meter. Background Art

[0002] A manufacturing enterprise is one that transforms input factors such as raw materials and parts into higher-value-added products through physical or chemical processing, assembly, and other production processes. Processing and assembly of raw materials and parts often requires the use of related production equipment. With the second round of electricity reforms, the electricity market has gradually evolved into a real-time market. To ensure the normal operation of production equipment within an enterprise, it is often necessary to forecast the company's electricity consumption so that the company can accurately purchase the appropriate amount of electricity in the real-time market. This prevents excessive power purchases from causing losses due to assessment deviations, or insufficient power purchases from production equipment that may not produce enough products, thus affecting the company's economic benefits.

[0003] Existing methods for predicting enterprise electricity consumption predict the enterprise's electricity consumption based on the enterprise's order volume or based on the enterprise's historical electricity consumption patterns. However, due to the relatively single factor considered in the prediction, this method results in poor prediction accuracy. For example, Chinese invention patent publication number CN 111275267 B discloses a method for predicting electricity consumption for a manufacturing enterprise. This method predicts the enterprise's electricity consumption by fitting a functional relationship between the enterprise's order volume and the enterprise's electricity consumption, and then predicts the electricity consumption based on the predicted order volume. Since this method only considers the order volume when predicting electricity consumption, in reality, not only the order volume but also other factors can affect the enterprise's electricity consumption, resulting in low accuracy in the electricity consumption prediction results of this method. Summary of the Invention

[0004] In order to solve the technical problem that the prediction results of the electricity consumption prediction methods of production enterprises in the prior art have low accuracy, the present invention provides solutions in the following aspects.

[0005] In a first aspect, the present invention provides a method for predicting power consumption based on a smart energy meter.

[0006] The method includes: obtaining the predicted average temperature value for the next date; obtaining the planned operating time of the enterprise's production equipment for the next date; using a smart electricity meter to obtain the actual value of the enterprise's electricity consumption for a consecutive number of preset days; inputting the predicted average temperature value for the next date and the planned operating time into a preset electricity consumption prediction model, thereby obtaining the predicted value of the enterprise's electricity consumption for the next date; the training set of the electricity consumption prediction model is composed of the actual electricity consumption value, average temperature and operating time for the consecutive preset days. When training the electricity consumption prediction model, the loss function corresponding to a certain date is positively correlated with the deviation between the actual electricity consumption value and the predicted value on that date, and is positively correlated with the ratio of the planned operating time to the actual operating time of the production equipment on that date.

[0007] Its beneficial effects are as follows: the power consumption prediction method based on the smart electricity meter of the present invention takes into account that higher or lower temperatures will increase the power consumption of the enterprise when predicting the power consumption of the enterprise; the longer the production equipment runs, the greater the power consumption of the enterprise; the neural network model is used to learn the coupling relationship between the average temperature of the day, the running time of the production equipment and the power consumption of the enterprise, and the power consumption of the enterprise is predicted based on the data of the two dimensions, thereby improving the accuracy of the prediction results; in addition, in the power consumption prediction model training process, the relationship between the ratio of the planned running time of the production equipment to the actual running time and the equipment aging is taken into account, and the aging of the production equipment will cause the power consumption of the production equipment to be high. By using this ratio to compensate for the loss function, the penalty for prediction error is reduced when the degree of equipment aging is relatively light, and the penalty for prediction error is increased when the degree of production equipment aging is relatively serious, so that the power consumption prediction model can better adapt to the situation of high actual power consumption caused by equipment aging, and effectively improve the prediction accuracy of the power consumption prediction model.

[0008] Preferably, the loss function expression used during training is:

[0009] ;

[0010] Where, represents the loss function corresponding to the nth date, and are the weighted coefficients for underestimation error and overestimation error, respectively. , and are the actual operating time and planned operating time of the production equipment on the nth date, and are the actual value and predicted value of electricity consumption on the nth date respectively.

[0011] Its beneficial effect is: making the weighted coefficient of the underestimated error in the loss function greater than the weighted coefficient of the overestimated error, making the penalty for under-prediction greater and the penalty for over-prediction smaller, thereby making the electricity consumption prediction model more adaptable to the situation of high actual electricity consumption caused by equipment aging, and avoiding insufficient power supply for enterprises.

[0012] Preferably, the method further comprises: modifying the loss function, wherein the calculation expression of the modified loss function is:

[0013] ;

[0014] Where, is the probability of production equipment failure on the nth date, Represents the corrected loss function.

[0015] The beneficial effect is that by increasing the penalty for under-prediction when the probability of equipment failure increases, and reducing the penalty for under-prediction when the probability of equipment failure decreases, the prediction model becomes more sensitive to underestimation errors (i.e., electricity consumption prediction results are lower than actual electricity consumption) as the probability of equipment failure increases.

[0016] Preferably, the predicted value of the average temperature for the next date is obtained using a weather forecast.

[0017] Preferably, the weighted coefficient expression for the underestimation error is: ;

[0018] Where, is the penalty factor for underestimated error, The probability of a production device failure on the nth date.

[0019] The beneficial effect is that when production equipment fails, it needs to be repaired by other devices before being put into operation, and the repair time is usually not too long. After it is repaired, it can be put into operation according to the planned operating time, which will eventually lead to an increase in its electricity consumption on a certain day. By setting the weighted coefficient of the underestimated error to increase with the increase in the probability of failure, the electricity consumption prediction model can better adapt to the situation where the actual electricity consumption is high due to equipment failure.

[0020] Preferably, the weighted coefficient expression for the overestimation error is: ;

[0021] Where, is the penalty factor for overestimation error, The probability of a production device failure on the nth date.

[0022] The beneficial effects are: since electricity consumption on the corresponding date will increase when production equipment fails; by making the weighted coefficient of overestimation error negatively correlated with the probability of equipment failure, the penalty for over-prediction decreases as the probability of equipment failure increases, which helps to avoid power shortages in enterprises.

[0023] Preferably, the continuous preset number of days is 50 days.

[0024] Preferably, the method further includes: processing the electricity consumption data obtained by using the smart electricity meter, and the processing method includes:

[0025] Perform data cleaning on electricity consumption data; data cleaning includes: checking the format of electricity consumption data and detecting whether there are missing values, duplicate data and erroneous data in the electricity consumption data, and processing the detected missing values, duplicate data and erroneous data;

[0026] Perform outlier detection and correction on the cleaned electricity consumption data;

[0027] Perform data smoothing and filtering on electricity consumption data.

[0028] Preferably, linear interpolation is used to fill in the missing electricity consumption data.

[0029] In a second aspect, the present invention provides an electricity consumption prediction device based on a smart electricity meter, comprising a memory and a processor, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, the electricity consumption prediction method based on the smart electricity meter of the present invention is implemented.

[0030] In summary, the beneficial effect of the present invention is that the method of the present invention can improve the accuracy of enterprise electricity consumption prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a flowchart schematically illustrating a method for predicting power consumption based on a smart energy meter according to an embodiment of the present invention;

[0032] Figure 2 The figure schematically shows the structure of a power consumption prediction system based on a smart energy meter according to an embodiment of the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0034] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0035] Example of a method for predicting power consumption based on a smart energy meter:

[0036] like Figure 1 As shown, the power consumption prediction method based on the smart electric energy meter of the present invention includes:

[0037] S101, obtaining the average temperature forecast value for the next date;

[0038] In this embodiment, the value of m is 10. In other embodiments, other appropriate values ​​may be used.

[0039] S102, obtaining the planned operating time of the enterprise's production equipment on the next date; using a smart electricity meter to obtain the actual value of the enterprise's electricity consumption for a preset number of consecutive days;

[0040] In this embodiment, the continuous preset number of days is 50 days. In other embodiments, other suitable days may be used, such as 40 days, 60 days, etc. The planned operating time is less than or equal to the working time of one working day of the enterprise.

[0041] In this embodiment, when collecting electricity usage data for an enterprise, ensure that the smart meter used supports long-term data storage and remote data access. The smart meter must be connected to the data collection platform via wired or wireless communication to upload data. After installing the smart meter, configure its basic parameters. For example, set the data logging interval to 15 minutes, 20 minutes, or once an hour, and set the duration for storing historical data to 60 days.

[0042] S103. Obtain the predicted value of the enterprise's electricity consumption for the next date, specifically: input the predicted value of the average temperature for the next date and the planned operating time into a preset electricity consumption prediction model, thereby obtaining the predicted value of the enterprise's electricity consumption for the next date; the training set of the electricity consumption prediction model is composed of the actual electricity consumption values, average temperatures and operating time of production equipment for the consecutive preset days. When training the electricity consumption prediction model, the loss function corresponding to a certain date is positively correlated with the deviation between the actual electricity consumption value and the predicted value on that date, and is positively correlated with the ratio of the planned operating time to the actual operating time of the production equipment on that date.

[0043] The manufacturing enterprise in this example refers to a manufacturing enterprise in the machining industry, whose production equipment operates in intermittent production mode. Throughout the day, the equipment may be in four states: maintenance, repair, operation, and shutdown. The equipment's daily operating time is related to order volume. When order volume is sufficient, the equipment will operate for a longer period. When order volume is low or during the off-season, the equipment will operate for a shorter period.

[0044] If the ratio of the planned operating time of production equipment to the actual operating time increases, it means that the equipment has aged, and the aging of production equipment will lead to higher electricity consumption of production equipment. By using this ratio to compensate the loss function, when the planned operating time of production equipment is close to the actual operating time (that is, the degree of equipment aging is relatively mild), the penalty for prediction error is reduced. When the planned operating time of production equipment deviates greatly from the actual operating time (that is, the degree of equipment aging is relatively serious), the penalty for prediction error is increased. This makes the electricity consumption prediction model more adaptable to the situation of higher actual electricity consumption caused by equipment aging, and effectively improves the prediction accuracy of the electricity consumption prediction model.

[0045] If the temperature is high or low during the day, in order to ensure that the staff are in a comfortable working environment, the company's air-conditioning equipment will be used for a longer period of time, and the air-conditioning will operate at high power consumption, thereby increasing the company's electricity consumption; for example: when the temperature is too high, the air-conditioning will turn on the cooling mode, and the air-conditioning temperature will be adjusted lower, and the wind speed will be adjusted higher, resulting in an increase in the power of the air-conditioning; when the temperature is too low, the air-conditioning will turn on the heating mode, and the air-conditioning temperature will be adjusted higher, and the wind speed will be adjusted higher, which will also lead to an increase in the power of the air-conditioning and an increase in power consumption per unit time.

[0046] If production equipment runs for a long time, the company's electricity consumption will also increase. The electricity consumption prediction method based on smart electricity meters of the present invention comprehensively considers the impact of the average daily temperature on the company's electricity consumption and the impact of the production equipment running time on the company's electricity consumption when predicting the company's electricity consumption. Because the neural network model can learn the coupling relationship between the average daily temperature, the production equipment running time and the company's electricity consumption, it uses data from these two dimensions combined with the neural network model to predict the company's electricity consumption, thereby making the electricity consumption prediction results more accurate.

[0047] In the present embodiment, the production equipment can be a numerically controlled machine tool. In other embodiments, it can also be other types of production equipment.

[0048] In one embodiment, the loss function expression used when training the power consumption prediction model is:

[0049] ;

[0050] Where, represents the loss function corresponding to the nth date, and are the weighted coefficients for underestimation error and overestimation error, respectively. , and are the actual operating time and planned operating time of the production equipment on the nth date, and are the actual value and predicted value of electricity consumption on the nth date respectively.

[0051] Since aging of production equipment will lead to higher electricity consumption of production equipment, it may cause insufficient electricity supply for the enterprise and affect the production efficiency of products. Therefore, the consequences of underestimating electricity consumption are more serious. By making the weighted coefficient of the underestimated error in the loss function greater than the weighted coefficient of the overestimation error, the penalty for under-prediction is increased and the penalty for over-prediction is reduced, so that the electricity consumption prediction model can better adapt to the situation of higher actual electricity consumption caused by aging of equipment, further improve the accuracy of electricity consumption prediction results, and avoid electricity consumption prediction being lower than actual electricity consumption, thereby avoiding insufficient electricity supply for enterprises.

[0052] In one embodiment, the method further includes: modifying the loss function, wherein the modified loss function calculation expression is:

[0053] ;

[0054] Where, is the probability of production equipment failure on the nth date, Represents the corrected loss function. is a weighting factor based on the probability of equipment failure, by using The loss function is modified so that the penalty for under-prediction is larger when the probability of equipment failure is higher. As the probability of equipment failure increases, the prediction model becomes more sensitive to underestimation errors (i.e., the power consumption prediction result is lower than the actual power consumption), making the power consumption prediction model more adaptable to the situation where the actual power consumption is higher due to equipment failure.

[0055] Under normal circumstances, the electricity consumption of production equipment is determined by its operating time. However, if a production equipment failure occurs, it must be repaired using other power-consuming devices, which consumes additional energy. Therefore, the penalty for under-prediction increases with the probability of equipment failure, making the prediction model more sensitive to underestimation errors. This makes the electricity consumption prediction model more adaptable to situations where actual electricity consumption is higher due to equipment failures. Using the modified loss function to train the electricity consumption prediction model can further improve its prediction accuracy.

[0056] In this embodiment, there are various methods for obtaining the probability of production equipment failure on the nth date. For example, an LSTM neural network model can be used to obtain the failure probability of production equipment on each date. Alternatively, the equipment failure prediction method disclosed in the invention patent with authorization publication number CN112395178B can be used to predict the probability of production equipment failure on the next date.

[0057] LSTM neural network, or long short-term memory network, is a time-recurrent neural network that can solve the long-term dependency problem of general RNN (recurrent neural network).

[0058] If the LSTM neural network model is used to obtain the failure probability of production equipment on each date, the training process of the LSTM neural network model is as follows:

[0059] S201. Construct a training set and a test set: obtain equipment operating parameters, maintenance records, and environmental parameters of production equipment in a historical period, and perform missing value processing and outlier detection on the obtained data.

[0060] In this embodiment, if the production equipment is a machine tool, the equipment operating parameters include current, voltage, spindle speed, spindle temperature, and spindle vibration signal. In other embodiments, other types of operating parameters may also be included.

[0061] S202: Extract the time window features of the preprocessed data and convert the fault occurrence time into a binary label (1 indicates fault, 0 indicates normal) for supervised learning.

[0062] You can set the equipment operating parameters, maintenance records, and environmental parameters of production equipment over a historical period as time series data and set a sliding window to traverse this time series data. Time window features can include mean, maximum, and minimum values ​​to reflect short-term trends. The time window can be set to 24 hours or another suitable duration. The data in one time window is considered a sample.

[0063] S203, define the LSTM model and initialize the parameters of the LSTM model,

[0064] The constructed LSTM network consists of an input layer, an LSTM layer, a fully connected layer, and an output layer. The input layer receives time window features. The LSTM layer consists of one to three stacked LSTM units, each containing 64 to 256 neurons, to capture temporal dependencies. The fully connected layer maps the LSTM output to a failure probability, using a sigmoid activation function. The output layer consists of a single neuron that outputs a failure probability value, ranging from 0 to 1.

[0065] Use Xavier initialization for weights and biases, all initialized to 0. Xavier (also known as Glorot initialization) is a neural network weight initialization method proposed by Xavier Glorot and Yoshua Bengio. Its core goal is to maintain consistent variance in activation values ​​across network layers, thereby accelerating training convergence and preventing vanishing or exploding gradients.

[0066] S204. Define the loss function and optimizer.

[0067] You can choose the binary cross-entropy loss function as the loss function and Adam as the optimizer. The binary cross-entropy loss function (BCE) is a core loss function for binary classification tasks in deep learning. It measures the difference between the probability distribution predicted by the model and the true label.

[0068] The Adam optimizer is an adaptive learning rate optimization algorithm that combines the concepts of momentum and RMSProp. It normalizes parameter updates to ensure that the magnitude of each parameter update is similar, effectively improving training results.

[0069] S205, forward propagation: Input the time window features into the LSTM model to obtain the failure probability and calculate the loss.

[0070] S206, Backpropagation: Calculate gradients using the chain rule.

[0071] S207, parameter update: The optimizer adjusts the weights according to the gradient.

[0072] S208. Iteratively perform forward propagation, backward propagation, and parameter update until a preset number of times is reached.

[0073] In one embodiment, in order to improve the quality of the collected real value of electricity consumption, the method further includes: processing the electricity consumption data obtained by using the smart electricity meter, and the processing method includes:

[0074] S301. Clean the electricity consumption data to handle format errors, missing values, duplicate records, and unreasonable data.

[0075] Data cleaning includes: checking the format of electricity consumption data and detecting whether there are missing values, duplicate data and erroneous data in the electricity consumption data, and processing the detected missing values, duplicate data and erroneous data.

[0076] For missing electricity consumption data, linear interpolation, mean filling, spline interpolation or time-weighted weighting can be used to fill the missing values ​​to ensure data integrity.

[0077] For duplicate data, delete the data recorded repeatedly at the same time point to ensure the uniqueness of the data at each time point.

[0078] For erroneous data, you can delete them or replace them with reasonable data values.

[0079] A reasonable data value can be the mean of the neighboring data.

[0080] The electricity consumption data is output and cleaned to ensure data quality and handle format errors, missing values, duplicate records and unreasonable data.

[0081] S302: Detect and correct abnormal values ​​of the cleaned electricity consumption data.

[0082] The Z-score method or box plot method can be used to detect outliers. The corresponding outliers can be eliminated or repaired by using the mean of the previous and next data points.

[0083] The process of detecting outliers using the Z-score method is as follows: the Z-score of each data point is calculated. If it exceeds a set threshold (such as 3 or -3), the point is considered an outlier.

[0084] The process of detecting outliers using the box plot method is as follows: outliers are identified using the interquartile range (IQR) method.

[0085] By performing outlier detection and correction on the cleaned electricity consumption data, outliers caused by factors such as sensor failure and power grid fluctuations can be detected and corrected.

[0086] S303: Perform data smoothing and filtering on the power consumption data.

[0087] The method for smoothing electricity consumption data can use simple moving average (SMA) or weighted moving average (WMA).

[0088] The process of using a simple moving average to smooth electricity consumption data is to apply a simple moving average to the data and take the mean of the data within the window for smoothing. For example, you can use a 7-day or 30-day window to smooth electricity consumption data to remove short-term fluctuations.

[0089] The process of using weighted moving average to smooth the electricity consumption data is: assigning different weights to each data point. Generally, the data points closer to the current time have larger weights, thereby reducing the interference of historical data on the current trend.

[0090] The method for filtering the electricity consumption data can be the median filter method, and the process is: for isolated abnormal values, use the median filter to replace them.

[0091] Median filtering is very effective in removing sudden power fluctuations and smoothing out short-term anomalies in the data.

[0092] By smoothing and filtering the electricity consumption data, short-term noise in the data can be removed and a smoother long-term trend can be extracted.

[0093] In one embodiment, the predicted average temperature value for the next date is obtained using a weather forecast.

[0094] In one embodiment, the weighting coefficient expression for the underestimation error is:

[0095] ;

[0096] Where, is the penalty factor for underestimated error, is the probability of production equipment failure on the nth date, The specific value of can be obtained through experiments.

[0097] When production equipment fails, it needs to be repaired before it can be put back into operation. Therefore, its electricity consumption on a certain day will increase (i.e., the company's electricity consumption on that day will increase). By setting the weighting coefficient of the underestimated error to increase with the increase in the probability of failure, the prediction model will become more sensitive to the underestimated error as the probability of equipment failure increases, making the electricity consumption prediction model more adaptable to the situation where the actual electricity consumption is higher due to equipment failure.

[0098] In one embodiment, the weighting coefficient expression for the overestimation error is:

[0099] ;

[0100] Where, is the penalty factor for overestimation error, is the probability of production equipment failure on the nth date, The specific value of can be obtained through experiments.

[0101] By making the weighting coefficient of overestimation error negatively correlated with the probability of equipment failure, the penalty for over-forecasting is reduced when the probability of equipment failure increases, which helps to avoid power supply shortages for enterprises.

[0102] Example of power consumption prediction device based on smart energy meter:

[0103] The present invention also provides a device for predicting power consumption based on a smart energy meter. Figure 2As shown, the power consumption prediction device based on the smart power meter includes a processor and a memory, and the memory stores computer program instructions. When the computer program instructions are executed by the processor, a power consumption prediction method based on the smart power meter according to the above embodiment of the present invention is implemented.

[0104] The power consumption prediction device based on the smart energy meter also includes other components well known to those skilled in the art, such as a communication bus and a communication interface. Their settings and functions are known in the art and will not be described in detail here.

[0105] In the description of this specification, “a plurality of” means at least two, for example, two, three or more, etc., unless otherwise clearly defined.

[0106] While several embodiments of the present invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Numerous modifications, variations, and alternatives will occur to those skilled in the art without departing from the concept and spirit of the present invention. It should be understood that various alternatives to the embodiments of the present invention described herein may be employed in practicing the present invention.

Claims

1. A method for predicting electricity consumption based on a smart energy meter, characterized in that: include: Get the average temperature forecast for the next date; Get the planned operating time of the enterprise's production equipment on the next date; Use smart electricity meters to obtain the actual value of the enterprise's electricity consumption for a preset number of consecutive days; Inputting the average temperature forecast value for the next date and the planned operating time into a preset electricity consumption forecast model to obtain the enterprise electricity consumption forecast value for the next date; The prediction model is used to learn the coupling relationship between the average temperature of the day, the operating time of production equipment, and the actual value of the enterprise's electricity consumption, and the enterprise's electricity consumption is predicted based on the data of these two dimensions; The training set of the power consumption prediction model consists of the actual power consumption values, average temperatures, and operating hours for the consecutive preset days. When training the power consumption prediction model, the loss function corresponding to a certain date is positively correlated with the deviation between the actual power consumption value and the predicted value on that date, and is positively correlated with the ratio of the planned operating hours to the actual operating hours of the production equipment on that date. The expression of the loss function used in training is: ； Where, represents the loss function corresponding to the nth date, and are the weighted coefficients for underestimation error and overestimation error, respectively. , and are the actual operating time and planned operating time of the production equipment on the nth date, and are the actual value and predicted value of electricity consumption on the nth date respectively; The loss function is modified, and the calculation expression of the modified loss function is: ； Where, is the probability of production equipment failure on the nth date, Represents the corrected loss function.

2. The method for predicting power consumption based on a smart energy meter according to claim 1, wherein: The predicted average temperature for the next date is obtained using the weather forecast.

3. The method for predicting power consumption based on a smart energy meter according to claim 1, wherein: The weighted coefficient expression for the underestimated error is: ; Where, is the penalty factor for underestimated error, The probability of a production device failure on the nth date.

4. The method for predicting power consumption based on a smart electric energy meter according to claim 1, wherein: The weighted coefficient expression for overestimation error is: ; Where, is the penalty factor for overestimation error, The probability of a production device failure on the nth date.

5. The method for predicting power consumption based on a smart energy meter according to claim 1, wherein: The preset consecutive number of days is 50 days.

6. The method for predicting power consumption based on a smart electric energy meter according to any one of claims 1 to 5, characterized in that: Also includes: The power consumption data obtained by using the smart energy meter is processed, and the processing method includes: Perform data cleaning on electricity consumption data; data cleaning includes: checking the format of electricity consumption data and detecting whether there are missing values, duplicate data and erroneous data in the electricity consumption data, and processing the detected missing values, duplicate data and erroneous data; Perform outlier detection and correction on the cleaned electricity consumption data; Perform data smoothing and filtering on electricity consumption data.

7. The method for predicting power consumption based on a smart energy meter according to claim 6, characterized in that: Linear interpolation is used to fill in the missing electricity consumption data.

8. A power consumption prediction device based on a smart energy meter, comprising a memory and a processor, wherein the memory stores computer program instructions, characterized in that: When the computer program instructions are executed by a processor, the power consumption prediction method based on a smart electric energy meter according to any one of claims 1 to 7 is implemented.

Images