Method, device and equipment for predicting electric load in target area

Abstract

The present invention provides a method, device, and equipment for predicting electric load in a target area, relating to the field of computer information technology. The method comprises: obtaining an electric load sequence and influencing factor data for the target area; performing outlier processing on the electric load sequence to obtain an electric load dataset; performing correlation analysis on the influencing factor data to obtain factor influence values; concatenating the electric load dataset and the factor influence values ​​to obtain electric load characteristic values; inputting the electric load characteristic values ​​into an electric load prediction model for electric load prediction processing, and modularly combining the output values ​​of the electric load prediction model to obtain an electric load prediction result for the target area. This solution can accurately predict electric load and achieve modular combination of load prediction values ​​for different zero-carbon industrial parks.

Description

Technical Field

[0001] The present invention relates to the field of computer information technology, and in particular to a method, device and equipment for predicting electric load in a target area. Background Art

[0002] As a vital hub for production and daily life, industrial parks concentrate a large number of industrial enterprises and equipment, generating enormous demands for energy, including electricity and heat. With the adjustment of energy structures and increasing environmental protection requirements, effectively managing industrial park energy systems, optimizing energy efficiency, and reducing carbon emissions have become critical issues that require urgent resolution.

[0003] The electricity load in a specific zero-carbon industrial park is affected by factors such as economic development level, industrial structure, season and climate, electricity price policy, energy substitution, and the application of energy-saving technologies. The accuracy of the prediction results cannot be guaranteed using traditional load forecasting methods. Summary of the Invention

[0004] The present invention provides a method, device and equipment for predicting electric load in a target area, so as to solve the problem of inaccurate prediction of electric load in a target area in the prior art.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0006] A method for predicting electric load in a target area, comprising:

[0007] Obtain the electric load sequence and influencing factor data of the target area;

[0008] performing outlier processing on the electric load sequence to obtain an electric load data set;

[0009] Performing correlation analysis on the influencing factor data to obtain factor influence values;

[0010] splicing the electric load data set and the factor influence value to obtain an electric load characteristic value;

[0011] The electric load characteristic value is input into the electric load prediction model to perform electric load prediction processing, and the output value of the electric load prediction model is modularly combined to obtain the electric load prediction result of the target area.

[0012] Optionally, performing outlier processing on the electric load sequence to obtain an electric load data set includes:

[0013] Calibrate the range of the electric load sequence to obtain first processed data;

[0014] Marking abnormal data on the first processed data to obtain second processed data;

[0015] Outlier correction and missing value filling are performed on the second processed data to obtain an electric load data set.

[0016] Optionally, performing a correlation analysis on the influencing factor data to obtain factor influence values ​​includes:

[0017] Adding a first disturbance amount to the influencing factor data to obtain first disturbance data;

[0018] Adding a second disturbance amount to the influencing factor data to obtain second disturbance data;

[0019] A factor influence value is obtained according to the first disturbance data and the second disturbance data.

[0020] Optionally, obtaining a factor influence value according to the first disturbance data and the second disturbance data includes:

[0021] Obtaining an average impact value according to the first disturbance data and the second disturbance data;

[0022] The average influence value is normalized to obtain the factor influence value.

[0023] Optionally, the electric load data set and the factor influence value are concatenated to obtain an electric load characteristic value, including:

[0024] The electric load data set and the factor influence values ​​are normalized to obtain an electric load characteristic value.

[0025] Optionally, inputting the electric load characteristic value into an electric load prediction model to obtain an electric load prediction result includes:

[0026] Inputting the electric load characteristic value into an electric load prediction model to perform electric load prediction processing to obtain a first output value and a second output value;

[0027] The first output value and the second output value are combined to obtain an electric load prediction result.

[0028] Optionally, the target area electric load prediction method further includes:

[0029] According to the evaluation index of the electric load forecast result, the electric load forecast evaluation result is obtained.

[0030] The present invention also provides an electric load prediction device for a target area, characterized by comprising:

[0031] An acquisition module is used to obtain the electric load sequence and influencing factor data of the target area;

[0032] The processing module is used to perform outlier processing on the electric load sequence to obtain an electric load data set; perform correlation analysis on the influencing factor data to obtain factor influence values; splice the electric load data set and the factor influence values ​​to obtain electric load characteristic values; input the electric load characteristic values ​​into an electric load prediction model to perform electric load prediction processing, and modularly combine the output values ​​of the electric load prediction model to obtain an electric load prediction result for a target area.

[0033] The present invention further provides a computing device, comprising: a processor and a memory storing a computer program, wherein the computer program executes the method described above when executed by the processor.

[0034] The present invention also provides a computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to execute the method described above.

[0035] The above solution of the present invention includes at least the following beneficial effects:

[0036] The above-mentioned solution of the present invention obtains the electric load sequence and influencing factor data of the target area; performs outlier processing on the electric load sequence to obtain an electric load data set; performs correlation analysis on the influencing factor data to obtain factor influence values; concatenates the electric load data set and the factor influence values ​​to obtain electric load characteristic values; inputs the electric load characteristic values ​​into the electric load prediction model for electric load prediction processing, and modularly combines the output values ​​of the electric load prediction model to obtain the electric load prediction results for the target area. The solution of the present invention can accurately predict electric load and realize modular combination of load prediction values ​​for different zero-carbon industrial parks. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a step diagram of a method for predicting electric load in a target area provided by an embodiment of the present invention;

[0038] Figure 2 is a specific flow chart of the electric load prediction method for the target area provided by an embodiment of the present invention;

[0039] Figure 3 This is a box plot of the electricity load data of Company A provided by a specific embodiment of the present invention;

[0040] Figure 4 This is a box plot of the electricity load data of Company B provided by a specific embodiment of the present invention;

[0041] Figure 5 A bar chart showing the degree of influence of influencing factors on Enterprise A provided by a specific embodiment of the present invention;

[0042] Figure 6A bar chart showing the degree of influence of influencing factors on Company B provided by a specific embodiment of the present invention;

[0043] Figure 7 This is a curve diagram of the optimization results of the electric load data of Enterprise A provided by a specific embodiment of the present invention;

[0044] Figure 8 This is a curve diagram of the optimization results of the electric load data of Enterprise B provided by a specific embodiment of the present invention;

[0045] Figure 9 This is a graph showing the electric load forecast results of Enterprise A provided by a specific embodiment of the present invention;

[0046] Figure 10 This is a graph showing the electric load forecast results of Company B provided by a specific embodiment of the present invention;

[0047] Figure 11 A modular combination result curve diagram of electric load forecasting results provided by a specific embodiment of the present invention;

[0048] Figure 12 It is a module schematic diagram of the electric load prediction device for the target area provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0050] like Figure 1 As shown, an embodiment of the present invention provides a method for predicting electric load in a target area, comprising:

[0051] Step 11, obtaining the electric load sequence and influencing factor data of the target area;

[0052] Step 12, performing outlier processing on the electric load sequence to obtain an electric load data set;

[0053] Step 13, performing correlation analysis on the influencing factor data to obtain factor influence values;

[0054] Step 14: combining the electric load data set and the factor influence value to obtain an electric load characteristic value;

[0055] Step 15: input the electric load characteristic value into the electric load prediction model to perform electric load prediction processing, and modularize the output value of the electric load prediction model to obtain the electric load prediction result of the target area.

[0056] In this example, a zero-carbon industrial park contains diverse industrial modules and facilities, such as polysilicon production, coal chemical industry, metallurgy, and petrochemical industry. The industrial park's electricity load is affected by a variety of factors. Specifically, the influencing factor data includes 13 factors: time characteristics such as date, hour, month, season, and day of the week; and meteorological characteristics such as dry-bulb temperature, dew point temperature, relative humidity, atmospheric pressure, global radiation, direct radiation, diffuse radiation, and wind speed. The electricity load data for different industrial facilities in the zero-carbon industrial park are arranged in a time series to generate an electricity load sequence for the target area. This data can reflect changes in electricity load over different time periods.

[0057] A boxplot is used to check the abnormal data of the electric load sequence, and the outliers are corrected and missing values ​​are filled; secondly, the importance of 13 external influencing factors that affect the temporal characteristics and meteorological characteristics of the industrial park electric load is evaluated, and the factor influence values ​​of the key factors are obtained; data optimization processing of the industrial park electric load sequence can effectively extract the electric load signal with multi-scale characteristics and adaptively optimize the decomposition, reducing the impact of noise on the prediction, thereby improving the accuracy and robustness of the load prediction; the key external factors are spliced ​​with the processed industrial park electric load sequence, and the electric load prediction model is used for prediction, and the prediction results are evaluated using the fitting curve; finally, the electric load prediction results in different zero-carbon industrial parks are obtained, and a modular combination of different zero-carbon industrial parks is realized.

[0058] The electric load prediction method for the target area of ​​this embodiment can accurately predict the electric load. The constructed model network can be directly applied to the electric load prediction and obtain accurate prediction results, thereby realizing the modular combination of load prediction values ​​for different zero-carbon industrial parks.

[0059] In an optional embodiment of the present invention, step 12 may include:

[0060] Step 121, calibrating the range of the electric load sequence to obtain first processed data;

[0061] Step 122: Mark the first processed data as abnormal data to obtain second processed data;

[0062] Step 123 : performing outlier correction and missing value filling on the second processed data to obtain an electric load data set.

[0063] In this embodiment, the median, lower quartile (Q1), upper quartile (Q3), and interquartile range (I QR ), whiskers and outliers.

[0064] The median of the electric load sequence is determined by the center line of the box plot; the lower quartile (Q1) is determined by the lower edge of the box; the upper quartile (Q3) is determined by the upper edge of the box; and the interquartile range (I QR ); According to the box boundary (i.e., the upper quartile Q3 and the lower quartile Q1) extending to the maximum and minimum non-outlier value in the data set, determine the upper and lower whiskers (Whiskers), set Data outside the upper and lower whiskers are considered outliers, and data points outside the upper and lower whiskers represent values ​​far from the data center and are identified as outliers.

[0065] The interquartile range method is used to set the range of the electric load series, I QR = Q3- Q1; By flexibly adjusting the quantile range to optimize the detection effect, the robustness is enhanced and the impact of extreme values ​​is reduced.

[0066] Detect abnormal values ​​within the set range of the electric load sequence: or higher The data are considered as outliers; and through the formula: , correct the outliers and obtain the electric load data set.

[0067] in, is an outlier, is the electric load value, and K is the number of electric load samples in the set range of the electric load sequence.

[0068] In an optional embodiment of the present invention, step 13 may include:

[0069] Step 131, adding a first disturbance value to the influencing factor data to obtain first disturbance data;

[0070] Step 132: adding a second disturbance value to the influencing factor data to obtain second disturbance data;

[0071] Step 133: Obtain a factor influence value according to the first disturbance data and the second disturbance data.

[0072] In this embodiment, a disturbance is added or subtracted from the original sequence of each external factor affecting the electric load, and the order of magnitude of the disturbance is 1e -5 , generating a positive perturbation and negative disturbances , set the disturbance ratio to 0.1.

[0073] The electric load with positive and negative disturbances is predicted separately to obtain the result of generating positive disturbance. and negative perturbation results By comparing these two sets of prediction results, the average impact of external factors on load forecasting is obtained. The calculation formula of the average impact value (MIV) is:

[0074] ;

[0075] in, is the number of samples of electrical load;

[0076] The calculated factor influence values ​​are positive and negative. To evaluate the influence of external factors on the electric load, the absolute value of the factor influence value is normalized to between 0 and 1, and the result is:

[0077] ;

[0078] Where, is the factor influence value, is the absolute value of the average impact value, is the minimum value of the average impact value, The maximum value of the average impact value, where 1 indicates the greatest impact and 0 indicates the least impact.

[0079] This embodiment uses the average impact value analysis method to process the electric load data, which can provide better explanation and practicality by measuring the average impact of input variables on output results.

[0080] In an optional embodiment of the present invention, step 14 may include:

[0081] The electric load data set and the factor influence values ​​are normalized to obtain an electric load characteristic value.

[0082] In this embodiment, the electric load data set is processed by variational mode decomposition (AVMD) with the optimization objectives of minimizing the modal components and maximizing the spectrum concentration. The formula is:

[0083] ;

[0084] ;

[0085] in, is the optimized electric load data set, K is the number of modes, For each modal component, is the center frequency of the k-th factor influence value.

[0086] When performing variational modal decomposition (AVMD) on an electric load dataset, the following parameters are used: alpha represents the bandwidth of the controlled modes. A larger value results in a narrower modal bandwidth. The alpha search range is set between 500 and 10,000 to ensure accurate decomposition results. A smaller bandwidth helps extract subtle features in the signal, thereby improving the model's predictive power. K represents the number of modes and can be set between 2 and 10 to accommodate different data characteristics. Choosing a reasonable K value can avoid overfitting or underfitting, thereby optimizing the signal decomposition. The modal convergence rate, tau, is set between 0 and 1 to ensure efficient algorithm operation. tol represents the convergence accuracy of the control algorithm and is set to 1e-6. A smaller tolerance allows the algorithm to find a more accurate solution. DC determines whether to retain the DC component of the signal. Setting it to 1 retains the DC component, making the superposition of IMFs closer to the original signal.

[0087] Then, the electric load data set after variational modal decomposition is globally dynamically optimized. The Bayesian algorithm (TPE) is used to adjust parameters such as bandwidth, number of modes, and convergence speed (alpha, K, tau). This allows the multi-scale characteristics of the load signal to be accurately captured, improves decomposition accuracy and noise resistance, simplifies model complexity, and improves prediction stability. The formula is:

[0088]

[0089] in, is the power spectral density of the jth modal component, To avoid numerical calculation errors, the smoothing factor (valued as ). H(alpha, K, tau) represents the spectral entropy. The above parameter space is sampled using the Bayesian algorithm, with the maximum number of evaluations set to 100, to minimize the spectral entropy H(alpha, K, tau).

[0090] This embodiment adopts a two-layer optimization method combining AVMD and TPE, and uses an adaptive algorithm to find the optimal modal parameters, which can reduce the calculation difficulty and improve the prediction accuracy.

[0091] In an optional embodiment of the present invention, step 15 may include:

[0092] Step 151: inputting the electric load characteristic value into an electric load prediction model to perform electric load prediction processing to obtain a first output value and a second output value;

[0093] Step 152: Combine the first output value and the second output value to obtain an electric load prediction result.

[0094] In this embodiment, the factor influence value X(t) in the above steps is combined with the optimized electric load data set The electric load characteristic value Z(t) is spliced ​​as the prediction model, and the electric load prediction model of the time-based bidirectional recurrent neural network is used to predict the electric load.

[0095] Specifically, the factor influence value X(t) is compared with the electric load data set For splicing, the formula is:

[0096] ;

[0097] right Perform maximum and minimum normalization to obtain the electric load characteristic value Z(t), which is:

[0098] .

[0099] To build an electric load forecasting model, we first build the neural network layer:

[0100] Input Gate :

[0101] ;

[0102] in, is the weight matrix of the input gate; is the hidden state at the previous moment; is the bias vector of the input gate; As the activation function, the sigmoid function is used, and the interval range is .

[0103] Candidate value :

[0104] ;

[0105] in, is the weight matrix of candidate values; is the bias vector of the candidate value; is the hyperbolic tangent activation function, which maps the output value to .

[0106] Unit Status :

[0107] ;

[0108] in, Is the output of the forget gate, used to control the unit state at the previous moment The forgetting process; is the output of the input gate; is the candidate value at the current moment, Represents element-wise multiplication.

[0109] Output Gate :

[0110] ;

[0111] in, is the weight matrix of the output gate; is the bias vector of the output gate; As the activation function, the sigmoid function is used.

[0112] Output value (hidden state) :

[0113] ;

[0114] in, Compress the value of the cell state to , To control the output ratio of the hidden state, Represents element-wise multiplication.

[0115] The two neural network layers are combined to construct an electric load forecasting model. The first neural network layer processes the forward information of the input sequence, and the second neural network layer processes the reverse information. Finally, the output of each neural network layer is merged into a single output. The data is divided into 80% training set and 20% test set. The formula is:

[0116] Forward neural network layer output :

[0117] ;

[0118] Reverse neural network layer output :

[0119] ;

[0120] Combine the outputs to get the electric load forecast value :

[0121] .

[0122] Finally, the power load forecast values ​​of different industries in the industrial park are The combined results are superimposed to form a modular combination result, and the electric load forecast result of the target area is obtained. A curve chart is drawn based on the electric load forecast result of the target area.

[0123] In the above steps, the parameters of the electric load forecasting model are set as follows:

[0124] Neural network layer configuration:

[0125] Number of units: The number of units in each neural network layer is set to 100 to provide the model with sufficient learning capacity to capture complex patterns in time series data.

[0126] Return sequences: The “return_sequences” parameter of the first neural network layer is set to True to ensure that the layer returns the output at each time step, which can be processed by subsequent layers.

[0127] Regularization:

[0128] To prevent the model from overfitting, a Dropout layer was added after each neural network layer, and the dropout rate was set to 0.2, that is, 20% of the neurons were randomly dropped during the training process.

[0129] Training process parameters:

[0130] Epochs: Set to 50 to control the number of epochs the model learns.

[0131] Batch Size: Set to 32, using 32 samples for each parameter update.

[0132] Optimizer and loss function:

[0133] The Adam optimizer is used, which performs well in dealing with complex problems and has an adaptive learning rate feature.

[0134] The mean squared error (MSE) is used as the loss function, which is suitable for regression tasks.

[0135] Callback mechanism:

[0136] Early Stopping: Monitor the validation loss (val_loss). If there is no improvement within 5 consecutive epochs, stop training and restore to the optimal weights.

[0137] Learning rate decay (ReduceLROnPlateau): If the validation loss does not improve within 3 epochs, the learning rate is reduced to 20% of the original value (factor = 0.2). The minimum learning rate is set to 0.001.

[0138] Data preprocessing:

[0139] All input features are normalized using MinMaxScaler so that their values ​​range from 0 to 1 to accelerate model convergence.

[0140] In an optional embodiment of the present invention, the target area electric load prediction method further includes:

[0141] Step 16: Obtain an electric load forecast evaluation result based on the evaluation index of the electric load forecast result.

[0142] In this embodiment, the load prediction results obtained in step 15 are used to perform an index evaluation, and the evaluation indicators include root mean square error (RMSE), mean absolute error (MAE), fitting coefficient (R 2 ):

[0143] RMSE= ;

[0144] MAE= ;

[0145] R 2 =1- ;

[0146] in, is the true value, is the predicted value, is the mean of the true value, n is the number of electric load forecast results, R 2 The closer the value is to 1, the stronger the explanatory power of the model is and the more accurate the electricity load forecast results are.

[0147] In a specific embodiment of the present invention, Figure 2 The figure shows a flow chart for predicting the electric load of a zero-carbon industrial park. The temporal and meteorological characteristics of the external factors and the electric load data of the zero-carbon industrial park are derived from the low-voltage power distribution area where the park is located. The park has a weather station information collection device installed.

[0148] Step S1: Collect the hourly electricity load of companies A and B throughout the year and the time characteristics of date, hour, month, season, and week, as well as external influencing factors such as dry-bulb temperature, dew point temperature, relative humidity, atmospheric pressure, total radiation, direct radiation, scattered radiation, and wind speed.

[0149] Step S2: For the hourly electricity load sequence of companies A and B throughout the year, set Q1 to 10% and Q2 to 90%. Use the 80% range set by IQR to identify and mark abnormal electricity load data using a box plot, such as Figure 3 and Figure 4 As shown in the figure, the K-nearest neighbor algorithm is used to correct outliers and fill missing values, and the new hourly electricity load series of enterprises A and B throughout the year are obtained;

[0150] Step S3: Use the mean impact value (MIV) method to evaluate the impact of 13 external factors of companies A and B on the power load of the industrial park, such as Figure 5 and Figure 6As shown in the figure, the order of the impact of external factors on Company A is date, total radiation, month, diffuse radiation, atmospheric pressure, season, direct radiation, wind speed, hour, week, dry bulb temperature, relative humidity, and dew point temperature; the order of the impact of external factors on Company B is dry bulb temperature, dew point temperature, relative humidity, atmospheric pressure, wind speed, total radiation, season, diffuse radiation, date, direct radiation, month, hour, and week.

[0151] Step S4: The electric load data of companies A and B are decomposed into multiple modal functions using a two-layer optimization method combining variational mode decomposition (AVMD) and Bayesian algorithm (TPE), such as Figure 7 and Figure 8 shown.

[0152] Step S5: The external factors that have a greater impact on companies A and B are combined with the power load signals processed by AVMD as training data for the power load prediction model to predict the power load for 8760 hours in the next year. Figure 9 and Figure 10 As shown in the figure, the RMSE of the predicted value and the true value of Company A is 1.46%, MAE is 1.07%, and R2 is 0.99; the RMSE of the predicted value and the true value of Company B is 0.577%, MAE is 0.44%, and R2 is 0.99;

[0153] Step S6: Modularize the industrial enterprises A and B. Figure 11 As shown, the load forecast of this park is obtained.

[0154] The target area electricity load forecasting method described in the above embodiments of the present invention addresses the impact of multiple nonlinear factors (such as season, temperature, and economic activity) on industrial park electricity load. By constructing a hierarchical electricity load forecasting model, it is possible to learn hidden complex patterns and handle complex nonlinear relationships, thereby improving forecasting accuracy. Industrial park electricity load is highly volatile, and this electricity load forecasting method can adapt to data changes, demonstrating strong robustness and generalization capabilities to dynamically fluctuating electricity demand. Given the large volume of electricity load data in industrial parks, the electricity load forecasting model performs well in big data environments. By training on extensive historical data, the model can more accurately predict future electricity loads. This provides a basis for industrial park power system scheduling and energy management, contributing to the realization of zero-carbon industrial parks.

[0155] like Figure 12 As shown, an embodiment of the present invention further provides an electric load prediction device 120 for a target area, comprising:

[0156] An acquisition module 121 is used to acquire the electric load sequence and influencing factor data of the target area;

[0157] The processing module 122 is used to perform outlier processing on the electric load sequence to obtain an electric load data set; perform correlation analysis on the influencing factor data to obtain factor influence values; concatenate the electric load data set and the factor influence values ​​to obtain electric load characteristic values; input the electric load characteristic values ​​into the electric load prediction model for electric load prediction processing, and modularly combine the output values ​​of the electric load prediction model to obtain the electric load prediction results of the target area.

[0158] Optionally, performing missing value processing on the electric load data to obtain an electric load data set includes:

[0159] Calibrate the range of the electric load sequence to obtain first processed data;

[0160] Marking abnormal data on the first processed data to obtain second processed data;

[0161] Outlier correction and missing value filling are performed on the second processed data to obtain an electric load data set.

[0162] Optionally, performing a correlation analysis on the influencing factor data to obtain factor influence values ​​includes:

[0163] Adding a first disturbance amount to the influencing factor data to obtain first disturbance data;

[0164] Adding a second disturbance amount to the influencing factor data to obtain second disturbance data;

[0165] A factor influence value is obtained according to the first disturbance data and the second disturbance data.

[0166] Optionally, obtaining a factor influence value according to the first disturbance data and the second disturbance data includes:

[0167] Obtaining an average impact value according to the first disturbance data and the second disturbance data;

[0168] The average influence value is normalized to obtain the factor influence value.

[0169] Optionally, the electric load data set and the factor influence value are concatenated to obtain an electric load characteristic value, including:

[0170] The electric load data set and the factor influence values ​​are normalized to obtain an electric load characteristic value.

[0171] Optionally, inputting the electric load characteristic value into an electric load prediction model to obtain an electric load prediction result includes:

[0172] Inputting the electric load characteristic value into an electric load prediction model to perform electric load prediction processing to obtain a first output value and a second output value;

[0173] The first output value and the second output value are combined to obtain an electric load prediction result.

[0174] It should be noted that the device is a device corresponding to the above method, and all implementation methods in the above method embodiments are applicable to the embodiments of the device and can achieve the same technical effects.

[0175] An embodiment of the present invention further provides a computing device comprising: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described in the above embodiment. All implementations of the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0176] The present invention also provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to execute the method described in the above embodiment. All implementations of the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0177] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0178] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0179] In the embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interface, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0180] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0181] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0182] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, ROM, RAM, a magnetic disk, or an optical disk.

[0183] In addition, it should be pointed out that in the apparatus and method of the present invention, it is obvious that each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent solutions of the present invention. Moreover, the steps of performing the above-mentioned series of processing can naturally be performed in chronological order according to the order of description, but they do not necessarily need to be performed in chronological order, and some steps can be performed in parallel or independently of each other. For those of ordinary skill in the art, it can be understood that all or any steps or components of the method and apparatus of the present invention can be implemented in hardware, firmware, software or a combination thereof in any computing device (including a processor, storage medium, etc.) or a network of computing devices. This can be achieved by those of ordinary skill in the art using their basic programming skills after reading the description of the present invention.

[0184] Therefore, the purpose of the present invention can also be achieved by running a program or a group of programs on any computing device. The computing device can be a well-known general-purpose device. Therefore, the purpose of the present invention can also be achieved simply by providing a program product containing program code that implements the method or device. That is to say, such a program product also constitutes the present invention, and the storage medium storing such a program product also constitutes the present invention. Obviously, the storage medium can be any well-known storage medium or any storage medium developed in the future. It should also be pointed out that in the device and method of the present invention, it is obvious that each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present invention. In addition, the steps of performing the above-mentioned series of processing can naturally be performed in chronological order according to the order of description, but do not necessarily need to be performed in chronological order. Certain steps can be performed in parallel or independently of each other.

[0185] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for predicting electric load in a target area, characterized in that: include: Obtain the electric load sequence and influencing factor data of the target area; performing outlier processing on the electric load sequence to obtain an electric load data set; Performing correlation analysis on the influencing factor data to obtain factor influence values; splicing the electric load data set and the factor influence value to obtain an electric load characteristic value; Inputting the electric load characteristic value into an electric load prediction model to perform electric load prediction processing to obtain a first output value and a second output value; Combining the first output value and the second output value to obtain an electric load prediction result; The electric load forecasting model is constructed through the following process: Input gate of a neural network layer for: ； in, is the weight matrix of the input gate; is the hidden state at the previous moment; is the bias vector of the input gate; As the activation function, the sigmoid function is used, and the interval range is ; is the characteristic value of electric load; Candidate value for: ； in, is the weight matrix of candidate values; is the bias vector of the candidate value; is the hyperbolic tangent activation function, which maps the output value to ; Unit Status for: ； in, Is the output of the forget gate, used to control the unit state at the previous moment The forgetting process; is the output of the input gate; is the candidate value at the current moment, represents element-wise multiplication; Output Gate for: ； in, is the weight matrix of the output gate; is the bias vector of the output gate; As the activation function, the sigmoid function is used; Output value for: ； in, Compress the value of the cell state to , To control the output ratio of the hidden state, represents element-wise multiplication; The electric load forecasting model is obtained by combining two neural network layers. The first neural network layer processes the forward information of the input sequence, and the second neural network layer processes the reverse information. Finally, the output of each neural network layer is combined into a single output to obtain the electric load forecast value. The correlation analysis of the influencing factor data to obtain the factor impact value includes: Adding a first disturbance amount to the influencing factor data to obtain first disturbance data; Adding a second disturbance amount to the influencing factor data to obtain second disturbance data; Obtaining a factor influence value according to the first disturbance data and the second disturbance data; Wherein, obtaining the factor influence value according to the first disturbance data and the second disturbance data includes: Obtaining an average impact value according to the first disturbance data and the second disturbance data; The average influence value is normalized to obtain the factor influence value.

2. The method for predicting electric load in a target area according to claim 1, characterized in that: Performing outlier processing on the electric load sequence to obtain an electric load data set includes: Calibrate the range of the electric load sequence to obtain first processed data; Marking abnormal data on the first processed data to obtain second processed data; Outlier correction and missing value filling are performed on the second processed data to obtain an electric load data set.

3. The method for predicting electric load in a target area according to claim 1, characterized in that: The electric load data set and the factor influence value are combined to obtain the electric load characteristic value, including: The electric load data set and the factor influence values ​​are normalized to obtain an electric load characteristic value.

4. The method for predicting electric load in a target area according to claim 1, wherein: Also includes: According to the evaluation index of the electric load forecast result, the electric load forecast evaluation result is obtained.

5. A device for predicting electric load in a target area, characterized in that: include: An acquisition module is used to obtain the electric load sequence and influencing factor data of the target area; A processing module is used to perform outlier processing on the electric load sequence to obtain an electric load data set; and perform correlation analysis on the influencing factor data to obtain factor influence values; splicing the electric load data set and the factor influence value to obtain an electric load characteristic value; Inputting the electric load characteristic value into the electric load prediction model to perform electric load prediction processing, and modularly combining the output values ​​of the electric load prediction model to obtain an electric load prediction result for the target area; Inputting the electric load characteristic value into an electric load prediction model to perform electric load prediction processing to obtain a first output value and a second output value; Combining the first output value and the second output value to obtain an electric load prediction result; The electric load forecasting model is constructed through the following process: Input gate of a neural network layer for: ； in, is the weight matrix of the input gate; is the hidden state at the previous moment; is the bias vector of the input gate; As the activation function, the sigmoid function is used, and the interval range is ; is the characteristic value of electric load; Candidate value for: ； in, is the weight matrix of candidate values; is the bias vector of the candidate value; is the hyperbolic tangent activation function, which maps the output value to ; Unit Status for: ； in, Is the output of the forget gate, used to control the unit state at the previous moment The forgetting process; is the output of the input gate; is the candidate value at the current moment, represents element-wise multiplication; Output Gate for: ； in, is the weight matrix of the output gate; is the bias vector of the output gate; As the activation function, the sigmoid function is used; Output value for: ； in, Compress the value of the cell state to , To control the output ratio of the hidden state, represents element-wise multiplication; The electric load forecasting model is obtained by combining two neural network layers. The first neural network layer processes the forward information of the input sequence, and the second neural network layer processes the reverse information. Finally, the output of each neural network layer is combined into a single output to obtain the electric load forecast value. The correlation analysis of the influencing factor data to obtain the factor impact value includes: Adding a first disturbance amount to the influencing factor data to obtain first disturbance data; Adding a second disturbance amount to the influencing factor data to obtain second disturbance data; Obtaining a factor influence value according to the first disturbance data and the second disturbance data; Wherein, obtaining the factor influence value according to the first disturbance data and the second disturbance data includes: Obtaining an average impact value according to the first disturbance data and the second disturbance data; The average influence value is normalized to obtain the factor influence value.

6. A computing device, characterized in that include: A processor and a memory storing a computer program, wherein when the computer program is executed by the processor, the method according to any one of claims 1 to 4 is performed.

7. A computer-readable storage medium, characterized in that: The device stores instructions, which, when executed on a computer, enable the computer to execute the method according to any one of claims 1 to 4.

Images