Regional electricity consumption prediction method based on Internet of Things

Abstract

The invention relates to a regional electricity consumption prediction method based on the Internet of Things. The method comprises the following steps: acquiring whole-city electricity consumption data; the whole-city power consumption data comprises power load data and related external data; inputting the whole-city power consumption data into the power consumption prediction model to obtain a whole-city power consumption prediction result; the power consumption prediction model is obtained by using the training set to train an LSTM model through forward and back propagation algorithms; the training set comprises historical city-wide power consumption data; and performing district-level power consumption prediction on the city-wide power consumption prediction result to obtain a district-level power consumption prediction result. According to the method, the consistency of hierarchical prediction results can be realized, the proportion decomposition hierarchical prediction method is adopted, the harmony and consistency of different hierarchical prediction results are ensured, and the hierarchical prediction effect and the practical application value are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electricity consumption prediction, and in particular to a method for regional electricity consumption prediction based on the Internet of Things. Background Art

[0002] With the rapid development of the power industry and smart grid technologies, power load forecasting is playing an increasingly critical role in energy planning, power generation, distribution, and power market reform. Load forecasting is not only fundamental to the efficient operation of power systems but also provides technical support for the in-depth application of the Energy Internet and Internet of Things (IoT) technologies. In recent years, the use of IoT devices in power systems has become increasingly widespread, becoming a crucial tool for data collection and monitoring. IoT devices can collect and transmit real-time power load data, including regional power consumption, equipment operation, weather, and other external factors, laying the foundation for the development of high-precision forecasting models.

[0003] Existing power load forecasting technologies are primarily categorized into statistical and machine learning methods. Traditional statistical methods, such as exponential smoothing, ARMA models, ARIMA models, and fuzzy logic, perform well when processing time series data, but have limitations when dealing with nonlinear and long-period fluctuating power load data. Machine learning methods, including support vector machines (SVMs), artificial neural networks (ANNs), and deep learning (such as long short-term memory networks (LSTMs), excel in modeling complex nonlinearities. The LSTM model, in particular, has been widely used in short-term power load forecasting due to its powerful time series modeling capabilities.

[0004] 1. Statistical methods: Statistical methods are widely used in power load forecasting. Classic methods include:

[0005] Exponential smoothing: This method uses past data to weight future data for smoothing. It is suitable for processing time series data with a certain degree of seasonality and trend. Its advantage is simple calculation, but the forecast accuracy may be low when faced with complex power load fluctuations.

[0006] Autoregressive Moving Average (ARMA) Model: The ARMA model is used to model stationary time series data and effectively captures data autocorrelation. It consists of an autoregressive (AR) component and a moving average (MA) component, building a forecasting model by analyzing past observations and forecast errors. While suitable for data without obvious trends or seasonal fluctuations, it may not provide accurate forecasts for complex power load data.

[0007] Autoregressive Integrated Moving Average (ARIMA) Model: The ARIMA model builds on the ARMA model by incorporating differencing for nonstationary data, making it suitable for time series with trends or seasonal variations. The ARIMA model is widely used for short- and medium-term power load forecasting, but it is still limited in its ability to handle nonlinear and long-term fluctuations.

[0008] Fuzzy logic: This method addresses uncertainty and ambiguity by establishing fuzzy rules and reasoning mechanisms. It is suitable for load forecasting with incomplete information. However, the fuzzy logic model has poor generalization capabilities when faced with large-scale power data.

[0009] Disadvantages:

[0010] When the forecast horizon is long, the accuracy of statistical methods decreases significantly, making it difficult to effectively capture the complex nonlinear relationships in power load. Statistical methods typically require the input data to meet certain stationarity assumptions, making traditional methods less adaptable to complex seasonal and holiday fluctuations in power load data.

[0011] 2. Machine Learning Methods: With the improvement of computing power and the development of big data technology, the application of machine learning methods in power load forecasting has been significantly improved. In particular, when faced with complex characteristics such as nonlinearity and multi-factor interference of power load, machine learning methods have shown strong advantages. Common machine learning methods include:

[0012] Support Vector Machine (SVM): SVM is a classic supervised learning method that excels at processing high-dimensional, nonlinear data. SVM maps input features into a higher-dimensional space and uses the maximum margin principle to find the optimal classification plane. Although SVM has achieved some success in some power load forecasting problems, it is computationally intensive and takes a long time to train when processing large amounts of data.

[0013] Artificial Neural Networks (ANNs): Neural networks have powerful nonlinear modeling capabilities and can capture complex patterns in power load data. However, traditional neural networks typically require large amounts of data for training and are sensitive to the choice of network structure and parameters, making them prone to overfitting.

[0014] Deep Learning: In recent years, deep learning methods have become mainstream in the field of power load forecasting. Deep learning can automatically extract data features and is particularly well-suited for processing complex nonlinear relationships. Recurrent neural networks (RNNs) and their variants, such as long short-term memory (LSTM) and gated recurrent units (GRUs), have been widely used in the field of short-term power load forecasting, achieving excellent results.

[0015] Strengths: Machine learning methods can effectively capture complex nonlinear relationships in data. Deep learning methods, in particular, demonstrate powerful feature extraction and adaptability in modeling large-scale power load data. Recurrent neural network (RNN) variants, such as long short-term memory (LSTM) and gated recurrent unit (GRU), are particularly adept at processing time series data and can capture long-term dependencies in the data, resulting in their excellent performance in power load forecasting.

[0016] Disadvantages: Machine learning methods usually require a large amount of data to train models. Especially in power load forecasting, large-scale historical data requires higher computing resources and training time. Summary of the Invention

[0017] In order to solve the problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a regional electricity consumption prediction method based on the Internet of Things. By using the LSTM deep learning model, very detailed features (such as the power consumption of specific equipment and the electricity consumption in different regions) can be collected through the Internet of Things devices. These data are rich in layers, which helps the model to more comprehensively understand the changing pattern of power load.

[0018] To achieve the above object, the present invention provides the following solutions:

[0019] A regional electricity consumption prediction method based on the Internet of Things, comprising:

[0020] Obtaining city-wide electricity consumption data; the city-wide electricity consumption data includes: power load data and related external data;

[0021] Inputting the city-wide electricity consumption data into a power consumption prediction model to obtain a city-wide power consumption prediction result; the power consumption prediction model is obtained by training an LSTM model using a training set through forward and back propagation algorithms; the training set includes: historical city-wide electricity consumption data;

[0022] Perform district-level power consumption forecast on the city-wide power consumption forecast result to obtain district-level power consumption forecast results.

[0023] Optionally, obtaining the training set includes:

[0024] The city-wide electricity consumption data is one-hot encoded and feature data is standardized, and effective features that have an impact on power load are selected to construct the training set.

[0025] Optionally, training the LSTM model further includes: using an Adam optimizer and a mean square error (MSE) as a loss function to compile the LSTM model.

[0026] Optionally, the power consumption prediction model includes:

[0027] Input gate, used to control the input information to be written into the memory unit;

[0028] Forget gate, used to decide to discard invalid information from the memory unit;

[0029] an output gate, configured to control the hidden state at the current moment and output the predicted result of the city's power consumption;

[0030] The input gate and the forget gate communicate with the gating mechanism of the output gate through their respective gating mechanisms.

[0031] Optionally, controlling the input information to be written into the memory unit includes:

[0032] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0033]

[0034] Among them, i t represents the output of the input gate, W i represents the weight matrix of the input gate, b i represents the bias term, h t-1 Indicates the hidden state of the previous moment, x t Indicates the current input, represents the candidate memory unit, W C represents the weight matrix, b C represents the bias term, and tanh represents the activation function.

[0035] Optionally, deciding to discard invalid information from the memory unit includes:

[0036] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0037] Among them, f t represents the output of the forget gate, h t-1 Indicates the hidden state of the previous moment, x t Indicates the current input, W f represents the weight matrix, b f represents the bias term, and σ represents the sigmoid activation function.

[0038] Optionally, the power consumption prediction model further includes:

[0039] An updating memory unit, configured to update a memory unit state according to output results of the input gate and the forget gate;

[0040] Updating the memory unit state includes:

[0041]

[0042] Among them, C t Represents the memory unit at the current moment, C t-1 Represents the memory unit of the previous moment.

[0043] Optionally, outputting the city-wide power consumption forecast result includes:

[0044] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0045] Among them, t Represents the output of the output gate, W o represents the weight matrix, b o represents the bias term.

[0046] Optionally, obtaining the district-level power consumption forecast result includes:

[0047] The power consumption forecast results of the entire city are used to make district-level power consumption forecasts, and the power consumption forecast values of each district are predicted by calculating the proportion of the city's power consumption data.

[0048] The beneficial effects of the present invention are:

[0049] The present invention can transmit data in real time through the Internet of Things devices, so that the model can instantly capture the load changes and fluctuations of external factors, and promptly reflect them in the prediction results. The introduction of this real-time data source makes the prediction more timely and adaptable. Very detailed features (such as the power consumption of specific equipment and the electricity consumption in different regions) can be collected through the Internet of Things devices. These data are rich in layers and help the model to understand the changing pattern of power load more comprehensively. Through the multi-dimensional and high-frequency characteristics of Internet of Things data, the LSTM model can capture complex nonlinear patterns and more accurately handle sudden fluctuations and non-periodic changes in power load, significantly improving the accuracy of load forecasting. Through the regional load data collected by Internet of Things devices, a hierarchical consistency correction model can be constructed to effectively control the prediction error when performing regional predictions, so that the overall prediction and hierarchical prediction results are harmonious and consistent.

[0050] The present invention solves the problem of insufficient accuracy of traditional methods in nonlinear and long-term power load prediction by using an LSTM deep learning model, avoids unnecessary feature redundancy, and improves model performance through reasonable feature selection and multi-level prediction.

[0051] The present invention improves prediction accuracy by selecting effective features in appropriate quantities to prevent the influence of feature redundancy on model performance.

[0052] The present invention can achieve consistency in hierarchical prediction results, adopts a hierarchical prediction method of proportional decomposition, ensures the harmony and consistency of prediction results at different levels, and improves the effect and practical application value of hierarchical prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0054] Figure 1 This is a flow chart of a method for predicting regional electricity consumption based on the Internet of Things according to an embodiment of the present invention;

[0055] Figure 2 1 is a comparison diagram of the actual value and the predicted value of the total social electricity consumption according to an embodiment of the present invention; wherein (a) is a first comparison diagram of the actual value and the predicted value, and (b) is a second comparison diagram of the actual value and the predicted value. DETAILED DESCRIPTION

[0056] 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 the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0057] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0058] like Figure 1As shown, this embodiment discloses a regional electricity consumption prediction method based on the Internet of Things, including: obtaining city-wide electricity consumption data; the city-wide electricity consumption data includes: power load data and related external data; inputting the city-wide electricity consumption data into a power consumption prediction model to obtain a city-wide electricity consumption prediction result; the power consumption prediction model is obtained by training an LSTM model through forward and backward propagation algorithms using a training set; the training set includes: historical city-wide electricity consumption data; performing district-level electricity consumption prediction on the city-wide electricity consumption prediction result to obtain a district-level electricity consumption prediction result.

[0059] Specifically:

[0060] IoT data preprocessing and feature selection: By analyzing historical data of power load, key features (such as temperature, time, social activities, etc.) are extracted, normalized and smoothed, and effective features with a greater impact on power load are selected.

[0061] Multi-layer LSTM model construction: Design a multi-layer LSTM neural network, combine it with an appropriate number of neurons and regularization method, and build a prediction model to capture the long-term dependence and short-term fluctuation characteristics in time series data.

[0062] Hierarchical forecasting and proportional decomposition: A top-down hierarchical forecasting strategy is adopted to first forecast the overall electricity consumption at the municipal level, and then allocate it to each district through proportional decomposition to ensure the overall consistency of the forecast.

[0063] Model training and optimization: Use historical data to conduct multiple model trainings and optimize the model's prediction accuracy by adjusting parameters such as the number of LSTM layers and the number of neurons.

[0064] Consistency correction: Use proportional adjustment to make consistency corrections to the forecast values of each district to ensure that the forecast values of the region and the city are consistent.

[0065] Furthermore, obtaining a training set includes: performing one-hot encoding on the city's electricity consumption data and standardizing feature data, and selecting effective features that affect power load to construct a training set; wherein standardization includes: normalization and smoothing.

[0066] Furthermore, training the LSTM model further includes: using an Adam optimizer and a mean square error (MSE) as a loss function to compile the LSTM model.

[0067] Furthermore, the power consumption prediction model includes: an input gate, which is used to control the writing of input information into the memory unit; a forget gate, which is used to decide whether to discard invalid information from the memory unit; and an output gate, which is used to control the hidden state at the current moment and output the city's power consumption prediction result; wherein, the input gate and the forget gate communicate information with the gating mechanism of the output gate through their respective gating mechanisms.

[0068] LSTM (Long Short-Term Memory) is an improved recurrent neural network (RNN) architecture specifically designed for processing and predicting sequence data, particularly excelling in problems with long temporal dependencies. The key advantage of LSTM lies in its ability to effectively capture long-term dependencies through a gating mechanism and avoid the vanishing or exploding gradient problems found in traditional RNNs.

[0069] The basic structure of LSTM consists of three gating units: the input gate, the forget gate, and the output gate. The design of these three gating mechanisms enables LSTM to maintain effective memory capabilities over long time steps.

[0070] Forget Gate:

[0071] The forget gate determines how much information in the memory unit (memory state) at the previous moment should be forgotten. Its output value range is 0 to 1, where 0 means completely forgotten and 1 means completely retained. The calculation formula of the forget gate is:

[0072] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0073] Among them, f t is the output of the forget gate, h t-1 is the hidden state of the previous moment, x t is the current input, W f is the weight matrix, b f is the bias term, and σ is the sigmoid activation function.

[0074] Input Gate:

[0075] The input gate controls how much information of the current input needs to be written into the current memory unit. Its calculation formula is:

[0076] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0077] Among them, i t is the output of the input gate, W i is the weight matrix of the input gate, b i is the bias term.

[0078] Candidate Memory Cell:

[0079] In order to update the memory unit, LSTM also calculates a candidate memory unit, which determines what information should be memorized at the current moment. The calculation formula for the candidate memory unit is:

[0080]

[0081] in, is a candidate memory unit, W C is the weight matrix, b C is the bias term, and tanh is the tanh activation function.

[0082] Memory Cell Update:

[0083] Through the forget gate and input gate, LSTM will update the current memory cell state. The new memory cell state is calculated by the following formula:

[0084]

[0085] Among them, C t is the memory unit at the current moment, C t-1 It is the memory unit of the previous moment.

[0086] Output Gate:

[0087] The output gate controls the hidden state at the current moment and determines the output at the current moment. The calculation formula of the output gate is:

[0088] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0089] Among them, t is the output of the output gate, W o is the weight matrix, b o is the bias term.

[0090] The update formula for the hidden state is:

[0091] h t =o t tanh(C t )

[0092] Among them, h t is the hidden state at the current moment, C t is the state of the memory cell.

[0093] Through these gating mechanisms, LSTM can decide which information should be retained and which should be forgotten, thereby effectively capturing long-term dependencies:

[0094] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0095] Among them, t Represents the output of the output gate, W o represents the weight matrix, b o represents the bias term.

[0096] Furthermore, obtaining district-level power consumption forecast results includes:

[0097] The district-level power consumption forecast is carried out based on the city-wide power consumption forecast results. The power consumption forecast value of each district is predicted by calculating the proportion of the city's power consumption data.

[0098] This embodiment discloses a method for predicting regional electricity consumption based on the Internet of Things, including:

[0099] A multi-layer LSTM network was constructed, and the model was trained through layer-by-layer forward propagation and backpropagation algorithms to finally generate an LSTM model for power consumption prediction.

[0100] Data loading and formatting: Load preprocessed data and separate it into lagged features and current features. Perform one-hot encoding on the monthly electricity consumption data, standardize the feature data, and convert it into a format suitable for LSTM input.

[0101] Model Construction: Build a multi-layer LSTM network. Each LSTM layer contains a certain number of neurons and uses the ReLU activation function. The final layer uses a fully connected layer for output. Compile the model using the Adam optimizer and the mean squared error (MSE) as the loss function.

[0102] Model Training and Prediction: During model training, we first predict global electricity consumption, then industrial electricity consumption. During training, we use batch stochastic gradient descent (Batch SGD) to optimize model parameters. After training, we calculate and output the mean squared error (MSE) and mean absolute percentage error (MAPE) to evaluate model performance.

[0103] Result Saving and Visualization: Plot the predicted values against the actual values to demonstrate the model's prediction performance. Also, save the prediction results as an Excel file for subsequent analysis.

[0104] Hierarchical Forecasting and Proportional Decomposition: Based on the city-wide electricity consumption forecast, district-level electricity consumption forecasts are performed. Proportional calculations are performed on historical data to derive forecasts for each district. The specific steps include data loading, proportional calculations, district-level forecasts, and error assessment. Finally, the district-level electricity consumption forecasts are saved and the associated forecast accuracy metrics are output. The city-wide electricity consumption forecast is used as the top-level forecast and allocated to each district using proportional decomposition. This decomposition is adjusted based on the initial forecast proportions to achieve a top-down forecast.

[0105] Data loading: City-wide electricity consumption data: This data is collected and aggregated from IoT devices at each factory, including but not limited to electricity consumption in each area, power consumption of various types of electrical equipment, weather data, price indices, etc. The data needs to be arranged in chronological order, and each record must include a timestamp (year, month, and day) and all relevant feature information. District-level electricity consumption data: This data is collected and aggregated from IoT devices at each factory, ensuring that it also includes the corresponding timestamp and can be aggregated by month or quarter.

[0106] Merge and format: Merge the city-wide electricity consumption data and the district-level electricity consumption data to generate unified time series data. Auxiliary features such as temperature and number of working days are aligned according to time.

[0107] Feature generation: Perform one-hot encoding on data such as months, and finally convert the data into a format suitable for LSTM input.

[0108] Proportional Calculation: The purpose of proportional calculation is to allocate electricity demand to each district (or banner) based on historical proportions based on the city-wide electricity consumption forecast. This ensures that the accuracy of the city-level forecast is carried forward to the district-level forecast.

[0109] Rolling ratio calculation: Historical ratio calculation: First, based on historical data (such as electricity consumption in the past 12 months), a rolling ratio is calculated for each district. For example, the proportion of electricity consumption in each district to the total electricity consumption in the city in the past 12 months is:

[0110]

[0111] Sliding window: To account for fluctuations between seasons and months, you can set a sliding window (such as 6 months or 12 months) to calculate the rolling ratio of each zone, which helps smooth data fluctuations and improve the stability of the forecast.

[0112] Proportional allocation: Based on the city's electricity consumption forecast, allocate it to each district in proportion. Specifically, assuming the city's electricity consumption forecast is P city , then the power consumption prediction value P of a certain area j isj for:

[0113] P j =P city × Ratio j

[0114] District-level forecast: The district-level power consumption forecast mainly uses the city-wide power consumption forecast results and combines them with the historical proportions of each district to make allocations. The specific steps are as follows:

[0115] District-level data loading: Read historical electricity consumption data for each district (flag area). This data needs to be aggregated by time (e.g., by month) and needs to include each district's historical total electricity consumption and corresponding characteristic information (e.g., weather data, price index, etc.).

[0116] Power consumption forecast: City-wide forecast results: Obtain city-wide power consumption forecast results from the LSTM model.

[0117] District-level allocation: The city's predicted electricity consumption is allocated to each district in proportion to its historical proportion. For each district i, the predicted electricity consumption is:

[0118]

[0119] Error Assessment: For each zone's prediction results, metrics such as mean squared error (MSE) and mean absolute percentage error (MAPE) can be used to assess the accuracy of the predictions. If the prediction errors for some zones are large, the model or scaling method can be adjusted based on these errors.

[0120] Visualization of prediction results: Use the matplotlib library to visualize and present the prediction results, such as Figure 2 As shown in (a)-(b).

[0121] The LSTM model predicts the electricity consumption in Ordos City with a percentage deviation (mape) of less than 4%, as shown in Table 1.

[0122] Table 1

[0123]

[0124] The LSTM model’s prediction deviation percentage for electricity consumption in various districts of Ordos City is around 6.5%, as shown in Table 2.

[0125] Table 2

[0126]

[0127] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by persons skilled in the art should fall within the scope of protection defined by the claims of the present invention.

Claims

1. A regional electricity consumption prediction method based on the Internet of Things, characterized in that: include: Obtain city-wide electricity consumption data; The city-wide electricity consumption data includes: power load data and related external data; Inputting the city-wide electricity consumption data into a power consumption prediction model to obtain a city-wide power consumption prediction result; the power consumption prediction model is obtained by training an LSTM model using a training set through forward and back propagation algorithms; the training set includes: historical city-wide electricity consumption data; Perform district-level power consumption forecast on the city-wide power consumption forecast result to obtain district-level power consumption forecast results.

2. The method for predicting regional electricity consumption based on the Internet of Things according to claim 1, characterized in that: Obtaining the training set includes: The city-wide electricity consumption data is one-hot encoded and feature data is standardized, and effective features that have an impact on power load are selected to construct the training set.

3. The method for predicting regional electricity consumption based on the Internet of Things according to claim 1, characterized in that: Training the LSTM model further includes: using an Adam optimizer and a mean square error (MSE) as a loss function to compile the LSTM model.

4. The method for predicting regional electricity consumption based on the Internet of Things according to claim 1, characterized in that: The power consumption prediction model includes: Input gate, used to control the input information to be written into the memory unit; Forget gate, used to decide to discard invalid information from the memory unit; an output gate, configured to control the hidden state at the current moment and output the predicted result of the city's power consumption; The input gate and the forget gate communicate with the gating mechanism of the output gate through their respective gating mechanisms.

5. The method for predicting regional electricity consumption based on the Internet of Things according to claim 4, characterized in that: Controlling the input information to be written into the memory unit includes: i t =σ(W i ·[h t-1 ,x t ]+b i ) Among them, i t represents the output of the input gate, W i represents the weight matrix of the input gate, b i represents the bias term, h t-1 Indicates the hidden state of the previous moment, x t Indicates the current input, represents the candidate memory unit, W C represents the weight matrix, b C represents the bias term, and tanh represents the activation function.

6. The method for predicting regional electricity consumption based on the Internet of Things according to claim 4, characterized in that: Deciding to discard invalid information from the memory unit includes: f t =σ(W f ·[h t-1 ,x t ]+b f ) Among them, f t represents the output of the forget gate, h t-1 Indicates the hidden state of the previous moment, x t Indicates the current input, W f represents the weight matrix, b f represents the bias term, and σ represents the sigmoid activation function.

7. The method for predicting regional electricity consumption based on the Internet of Things according to claim 5, characterized in that: The power consumption prediction model also includes: An updating memory unit, configured to update a memory unit state according to output results of the input gate and the forget gate; Updating the memory unit state includes: Among them, C t Represents the memory unit at the current moment, C t-1 Represents the memory unit of the previous moment.

8. The method for predicting regional electricity consumption based on the Internet of Things according to claim 4, characterized in that: Outputting the city-wide electricity consumption forecast results includes: the t =σ(W o ·[h t-1 ,x t ]+b o ) Among them, t Represents the output of the output gate, W o represents the weight matrix, b o represents the bias term.

9. The method for predicting regional electricity consumption based on the Internet of Things according to claim 1, characterized in that: Obtaining the district-level power consumption forecast result includes: The power consumption forecast results of the entire city are used to make district-level power consumption forecasts, and the power consumption forecast values of each district are predicted by calculating the proportion of the city's power consumption data.

Images