CN110266002A - Method and apparatus for forecasting electrical load - Google Patents

Abstract

The embodiment of the present application discloses a method and a device for predicting electric load. A specific implementation of the method includes: obtaining the historical power load data, historical environmental data and future environmental data of the enterprise to be predicted; using a pre-trained encoder to encode the historical power load data and historical environmental data of the enterprise to be predicted to obtain the Predict the historical power load characteristics of the enterprise; use the pre-trained decoder to decode the historical power load characteristics and future environmental data of the enterprise to be predicted, and obtain the future power load forecast data of the enterprise to be predicted. This embodiment relates to the field of cloud computing, and provides a power load forecasting method based on the combination of an encoder and a decoder. The encoder is used to learn the historical power load characteristics of the enterprise, and the decoder is used to predict the future power load data of the enterprise, which improves the prediction accuracy of the power load.

Description

Method and apparatus for forecasting electrical load

technical field

The embodiments of the present application relate to the field of computer technology, and in particular to a method and device for predicting electric load.

Background technique

With the progress of domestic power reform, more and more enterprises have entered the power market. According to statistics, there are currently more than 20,000 electricity sales companies in China. Due to the particularity of electricity, the state requires electricity sales companies to estimate in advance the amount of electricity they will use and make a declaration. If the deviation between the actual electricity used in the end and the declared electricity exceeds a certain range, high deviation fines will be imposed on electricity sales companies. Therefore, power load forecasting is particularly important for electricity sales enterprises, which is related to the survival of electricity sales enterprises.

At present, commonly used power load forecasting methods include similar day method, time series method and so on. Among them, the similar day method is to predict the power load by selecting similar historical day data and then obtaining a weighted average from it. The time series method is to establish a data model of the power load changing with time based on the historical power load data of the enterprise, and predict the future power load on the basis of the model.

Contents of the invention

Embodiments of the present application propose a method and device for predicting electric load.

In the first aspect, the embodiment of the present application provides a method for predicting power load, including: obtaining the historical power load data, historical environment data and future environment data of the enterprise to be predicted; Encode the historical power load data and historical environmental data to obtain the historical power load characteristics of the enterprise to be predicted; use the pre-trained decoder to decode the historical power load characteristics and future environmental data of the enterprise to be predicted to obtain the future power load of the enterprise to be predicted forecast data.

In some embodiments, the encoder is a convolutional neural network and the decoder is a long short-term memory network.

In some embodiments, a pre-trained encoder is used to encode the historical power load data and historical environmental data of the enterprise to be predicted to obtain the historical power load characteristics of the enterprise to be predicted, including: based on the historical power load data and historical data of the enterprise to be predicted Environmental data, generating historical two-dimensional data of the enterprise to be tested, wherein the historical two-dimensional data includes time and space dimensions, the time dimension includes historical time points, and the spatial dimension includes power load data and environmental data at each historical time point; based on Historical two-dimensional data to generate the historical image of the enterprise to be predicted; input the historical image of the enterprise to be predicted to the convolutional neural network to obtain the historical power load characteristics of the enterprise to be predicted.

In some embodiments, the convolutional neural network includes an input network, a residual neural network, and an output network, and the operations involved in the input network include at least one of the following: convolution, batch normalization, and activation function transformation, and the operations involved in the residual neural network Including at least one of the following: convolution, random deactivation, activation function transformation, batch normalization and maximum pooling, and the operations involved in the output network include at least one of the following: batch normalization and average pooling.

In some embodiments, the residual neural network merges the output of the input network with the output of the residual neural network after the maximum pooling through direct connection.

In some embodiments, the LSTM network includes a plurality of LSTM units, and each LSTM unit includes an input gate, a forget gate, and an output gate.

In some embodiments, the encoder and the decoder are trained through the following steps: obtaining a set of training samples, wherein each training sample includes the following three parts: the first preset length of sample power load data and sample environment data, the second The sample environment data of the preset length and the sample power load data of the second preset length; for the training samples in the training sample set, the sample power load data and the sample environment data of the first preset length in the training samples are coded as The input of the encoder, the output of the encoder and the sample environment data of the second preset length in the training sample are used as the input of the decoder, and the sample power load data of the second preset length in the training sample is used as the decoder. Output, train the encoder and decoder.

In some embodiments, obtaining a training sample set includes: obtaining historical power load data and historical environmental data of a sample enterprise; using a sliding window to segment the historical electric load data and historical environmental data of a sample enterprise to generate a training sample set, wherein , the length of the sliding window is equal to the sum of the first preset length and the second preset length, and each training sample includes the following three parts: the historical power load data and historical environmental data of the first preset length before, and the second preset length after The historical environmental data of a length and the historical power load data of a second preset length.

In the second aspect, the embodiment of the present application provides a device for forecasting power load, including: an acquisition unit configured to acquire historical power load data, historical environment data and future environment data of the enterprise to be predicted; an encoding unit configured to It is configured to use a pre-trained encoder to encode the historical power load data and historical environmental data of the enterprise to be predicted to obtain the historical power load characteristics of the enterprise to be predicted; the decoding unit is configured to use a pre-trained decoder to predict the history of the enterprise The power load characteristics and future environmental data are decoded to obtain the future power load forecast data of the enterprise to be forecasted.

In some embodiments, the encoder is a convolutional neural network and the decoder is a long short-term memory network.

In some embodiments, the encoding unit is further configured to: generate historical two-dimensional data of the enterprise to be measured based on historical power load data and historical environment data of the enterprise to be predicted, wherein the historical two-dimensional data includes a time dimension and a spatial dimension, The time dimension includes historical time points, and the spatial dimension includes power load data and environmental data at each historical time point; based on historical two-dimensional data, the historical image of the enterprise to be predicted is generated; the historical image of the enterprise to be predicted is input to the convolutional neural network , to get the historical power load characteristics of the enterprise to be predicted.

In some embodiments, the convolutional neural network includes an input network, a residual neural network, and an output network, and the operations involved in the input network include at least one of the following: convolution, batch normalization, and activation function transformation, and the operations involved in the residual neural network Including at least one of the following: convolution, random deactivation, activation function transformation, batch normalization and maximum pooling, and the operations involved in the output network include at least one of the following: batch normalization and average pooling.

In some embodiments, the residual neural network merges the output of the input network with the output of the residual neural network after the maximum pooling through direct connection.

In some embodiments, the LSTM network includes a plurality of LSTM units, and each LSTM unit includes an input gate, a forget gate, and an output gate.

In some embodiments, the encoder and the decoder are trained through the following steps: obtaining a set of training samples, wherein each training sample includes the following three parts: the first preset length of sample power load data and sample environment data, the second The sample environment data of the preset length and the sample power load data of the second preset length; for the training samples in the training sample set, the sample power load data and the sample environment data of the first preset length in the training samples are coded as The input of the encoder, the output of the encoder and the sample environment data of the second preset length in the training sample are used as the input of the decoder, and the sample power load data of the second preset length in the training sample is used as the decoder. Output, train the encoder and decoder.

In some embodiments, obtaining a training sample set includes: obtaining historical power load data and historical environmental data of a sample enterprise; using a sliding window to segment the historical electric load data and historical environmental data of a sample enterprise to generate a training sample set, wherein , the length of the sliding window is equal to the sum of the first preset length and the second preset length, and each training sample includes the following three parts: the historical power load data and historical environmental data of the first preset length before, and the second preset length after The historical environmental data of a length and the historical power load data of a second preset length.

In the third aspect, the embodiment of the present application provides an electronic device, the electronic device includes: one or more processors; a storage device, on which one or more programs are stored; when one or more programs are used by one or more processors, so that one or more processors implement the method described in any implementation manner in the first aspect.

In a fourth aspect, an embodiment of the present application provides a computer-readable medium, on which a computer program is stored, and when the computer program is executed by a processor, the method described in any implementation manner in the first aspect is implemented.

The method and device for predicting power load provided by the embodiments of the present application firstly obtain the historical power load data, historical environment data and future environment data of the enterprise to be predicted; then use the pre-trained encoder to obtain the historical power load data of the enterprise to be predicted Coding with historical environmental data to obtain the historical power load characteristics of the enterprise to be predicted; finally, use the pre-trained decoder to decode the historical power load characteristics and future environmental data of the enterprise to be predicted to obtain the future power load forecast of the enterprise to be predicted data. A power load forecasting method based on the combination of encoder and decoder is provided. The encoder is used to learn the historical power load characteristics of the enterprise, and the decoder is used to predict the future power load data of the enterprise, which improves the prediction accuracy of the power load.

Description of drawings

Other characteristics, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

FIG. 1 is an exemplary system architecture to which the present application can be applied;

FIG. 2 is a flowchart of one embodiment of a method for forecasting electrical loads according to the present application;

Fig. 3 is a schematic structural diagram of an encoder and a decoder;

FIG. 4 is a flow chart of one embodiment of a method for training an encoder and a decoder according to the present application;

FIG. 5 is a schematic structural diagram of an embodiment of a device for predicting electric load according to the present application;

FIG. 6 is a schematic structural diagram of a computer system suitable for implementing the electronic device of the embodiment of the present application.

Detailed ways

The application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain related inventions, rather than to limit the invention. It should also be noted that, for the convenience of description, only the parts related to the related invention are shown in the drawings.

It should be noted that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and embodiments.

FIG. 1 shows an exemplary system architecture 100 to which embodiments of the method for forecasting electric load or the apparatus for predicting electric load of the present application can be applied.

As shown in FIG. 1 , a system architecture 100 may include a terminal device 101 , a network 102 and a server 103 . The network 102 is used as a medium for providing a communication link between the terminal device 101 and the server 103 . Network 102 may include various connection types, such as wires, wireless communication links, or fiber optic cables, among others.

The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages and the like. Various client software, such as information prediction applications, can be installed on the terminal device 101 .

The terminal device 101 may be hardware or software. When the terminal device 101 is hardware, it may be various electronic devices that have a display screen and support information prediction. This includes, but is not limited to, smartphones, tablets, laptops, and desktops, among others. When the terminal device 101 is software, it can be installed in the above-mentioned electronic device. It can be implemented as a plurality of software or software modules, or as a single software or software module. No specific limitation is made here.

The server 103 may be a server that provides various services. For example, information prediction server. The information forecast server can analyze and process the acquired historical power load data, historical environmental data and future environmental data of the enterprise to be predicted, generate processing results (such as the future power load forecast data of the enterprise to be predicted), and process The result is pushed to the terminal device 101.

It should be noted that the server 103 may be hardware or software. When the server 103 is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server. When the server 103 is software, it may be implemented as multiple software or software modules (for example, for providing distributed services), or as a single software or software module. No specific limitation is made here.

It should be noted that the method for predicting electric load provided by the embodiment of the present application is generally executed by the server 103 , and correspondingly, the device for predicting electric load is generally disposed in the server 103 .

It should be understood that the numbers of terminal devices, networks and servers in Fig. 1 are only illustrative. According to the implementation needs, there can be any number of terminal devices, networks and servers.

Continuing to refer to FIG. 2 , which shows a flow 200 of an embodiment of a method for forecasting electric load according to the present application. The method for forecasting electric load includes the following steps:

Step 201, acquiring historical power load data, historical environmental data and future environmental data of the enterprise to be predicted.

In this embodiment, the execution subject of the method for forecasting electric load (for example, the server 103 shown in FIG. 1 ) can obtain the historical electric load data, historical environmental data and future environmental data.

Here, the enterprise to be predicted can be any electricity sales enterprise. The historical power load data may be real power load data of the enterprise to be predicted within a historical time period (for example, within the previous two years). The historical environmental data may be real environmental data of the enterprise to be predicted in a historical time period. The future environmental data may be predicted environmental data of the enterprise to be predicted within a future time period (for example, within one month in the future). The power load data may be the sum of the electric power drawn from the power system by all electric equipment of the enterprise to be predicted within a certain period of time. The environmental data can be the external environmental data of the enterprise to be predicted in a certain period of time, including but not limited to the production situation of the enterprise to be predicted, the geographic information of the area where the enterprise to be predicted is located, and the area where the enterprise to be predicted is located in a certain period of time. The temperature in the time period, and the holidays in the time period, etc.

Step 202, using a pre-trained encoder to encode the historical power load data and historical environment data of the enterprise to be predicted to obtain the historical power load characteristics of the enterprise to be predicted.

In this embodiment, the execution subject can use a pre-trained encoder (Encoder) to encode the historical power load data and historical environment data of the enterprise to be predicted, so as to obtain the historical power load characteristics of the enterprise to be predicted.

Here, the encoder can be used to learn the historical power load characteristics of the enterprise, and characterize the correspondence between the historical power load data and the historical environmental data of the enterprise and the historical power load characteristics of the enterprise.

Here, the historical power load feature can represent the historical power load law of the enterprise, that is, the correlation between the historical power load data and the historical environmental data (including but not limited to the power load trend of the enterprise in the past period of time, the historical power load of the enterprise and the historical environmental data). The correlation of holidays, the correlation between the historical power load of the enterprise and the temperature, etc.).

In this embodiment, the encoder can be trained in various ways.

In some optional implementations of this embodiment, the above-mentioned executive body may pre-collect historical power load data and historical environment data of a large number of enterprises, and manually analyze the historical power load characteristics of these enterprises. Subsequently, the historical power load data, historical environmental data and historical power load characteristics of these enterprises are correspondingly stored to generate a correspondence table as an encoder. In this way, the above-mentioned executive body can first calculate the similarity between the historical power load data and historical environment data of the enterprise to be predicted and the historical power load data and historical environment data of each enterprise in the correspondence table; then based on the calculated similarity degree, and determine the historical power load characteristics of the enterprise to be predicted from the corresponding relationship table. For example, the executive body may select the historical power load characteristics of the enterprise with the highest similarity with the historical power load data and historical environmental data of the enterprise to be predicted from the correspondence table as the historical power load characteristics of the enterprise to be predicted.

In some optional implementation manners of this embodiment, the encoder may be obtained by performing supervised training on an existing machine learning model by using a machine learning method and training samples. Typically, the encoder can be a Convolutional Neural Network (CNN). Convolutional neural networks have made significant progress in the field of images and have outperformed humans in image classification problems. Therefore, the above execution subject can treat the historical power load data and historical environmental data of the enterprise to be predicted as an image and be processed by the convolutional neural network.

In some optional implementations of this embodiment, the executive body may first generate historical two-dimensional data of the enterprise to be tested based on the historical power load data and historical environmental data of the enterprise to be predicted; then based on the historical two-dimensional data, generate The historical image of the enterprise to be predicted; finally, the historical image of the enterprise to be predicted is input to the convolutional neural network to obtain the historical power load characteristics of the enterprise to be predicted. Wherein, the historical two-dimensional data may include a time dimension and a space dimension. The time dimension can include historical points in time. The spatial dimension may include power load data and environmental data at each historical point in time. In this way, when the convolutional neural network learns the historical power load characteristics of the enterprise, it can take into account both time and space, that is, the horizontal and vertical dimensions of the input data.

Step 203, using a pre-trained decoder to decode the historical power load characteristics and future environmental data of the enterprise to be predicted, to obtain the future power load forecast data of the enterprise to be predicted.

In this embodiment, the execution subject can use a pre-trained decoder (Decoder) to decode the historical power load characteristics and future environmental data of the enterprise to be predicted, so as to obtain the future power load forecast data of the enterprise to be predicted.

Here, the decoder can be used to predict the future power load data of the enterprise, and characterize the correspondence between the historical power load characteristics of the enterprise and the future environmental data and the future power load forecast data of the enterprise.

In this embodiment, the decoder can be trained in various ways.

In some optional implementations of this embodiment, the execution subject may pre-collect historical power load characteristics, historical environment data, and historical power load data of a large number of enterprises, and store correspondingly to generate a correspondence table as a decoder. In this way, the above-mentioned executive body can first calculate the similarity between the historical power load characteristics and future environmental data of the enterprise to be predicted and the historical power load characteristics and historical environmental data of each enterprise in the correspondence table; then based on the calculated similarity degree, and determine the future power load forecast data of the enterprise to be forecasted from the correspondence table. For example, the executive body may select the historical power load data of the enterprise with the highest similarity with the historical power load characteristics and future environmental data of the enterprise to be predicted from the correspondence table as the future power load forecast data of the enterprise to be predicted.

In some optional implementation manners of this embodiment, the decoder may be obtained by performing supervised training on an existing machine learning model by using a machine learning method and training samples. On the problem of sequence decoding, Recurrent Neural Network (RNN) has achieved good results, especially Long Short-Term Memory (LSTM). Therefore, the decoder can be a long short-term memory network.

In this embodiment, the above execution subject may use an encoder-decoder mode to predict power load data. For ease of understanding, FIG. 3 shows a schematic structural diagram of an encoder and a decoder.

The left half of Fig. 3 shows a schematic structural diagram of the encoder. Here, the convolutional neural network may include an input network, a residual neural network (Resnet), and an output network. Generally, there may be multiple residual neural networks in the convolutional neural network, and FIG. 3 only schematically shows one residual neural network.

Here, the operations involved in the input network may include but not limited to at least one of the following: convolution (Convolution), batch normalization (BatchNorm), activation function transformation (ReLU), and the like. Batch normalization can make the network stable while training. The input to the network may be historical power load data and historical environmental data at multiple historical time points. Fig. 3 schematically shows historical power load data and historical environment data at w historical time points.

Here, the operations involved in the residual neural network may include, but are not limited to, at least one of the following: convolution, random inactivation (Dropout), activation function transformation, batch normalization, and maximum pooling (Max pool) and the like. The number of convolutional layers in a residual neural network is usually two. Dropout prevents overfitting. The residual neural network can merge the output of the input network with the output of the residual neural network after a direct connection (shortcut). Through the structure of the residual neural network, the network depth of the encoder can be greatly increased, so that the encoder can learn more features.

Here, the operations involved in the output network may include but not limited to at least one of the following: batch normalization, average pooling (Average pool) and so on. The output of the output network can be a one-dimensional vector, which can be used to represent the historical power load characteristics of the enterprise.

The right half of Fig. 3 shows a schematic structural diagram of the decoder. Here, the LSTM network may include multiple LSTM units. Each LSTM unit may include an input gate, a forget gate and an output gate.

Here, the input of the decoder may be the output of the encoder and future environment data at multiple future time points. Fig. 3 schematically shows future environmental data at p future time points. The output of the decoder may be future electrical load data for multiple future points in time. Fig. 3 schematically shows future power load data at p future time points.

Usually, the long-term short-term memory network adds a "processor" to the algorithm to judge whether information is useful or not. The structure of this "processor" is the long-term short-term memory unit, also known as a cell. Three gates are placed in a cell, which are called input gate, forget gate and output gate. When a piece of information enters the long-term short-term memory network, it can be judged according to the rules whether it is useful or not. Only the information that conforms to the algorithm authentication will be left, and the information that does not match will be forgotten through the forget gate.

The method for predicting power load provided by the embodiment of the present application first obtains the historical power load data, historical environmental data and future environmental data of the enterprise to be predicted; then uses the pre-trained encoder to obtain the historical power load data and historical The environmental data is encoded to obtain the historical power load characteristics of the enterprise to be predicted; finally, the pre-trained decoder is used to decode the historical power load characteristics and future environmental data of the enterprise to be predicted to obtain the future power load forecast data of the enterprise to be predicted. A power load forecasting method based on the combination of encoder and decoder is provided. The encoder is used to learn the historical power load characteristics of the enterprise, and the decoder is used to predict the future power load data of the enterprise, which improves the prediction accuracy of the power load.

Continuing to refer to FIG. 4 , which shows a flow 400 of an embodiment of a method for training an encoder and a decoder according to the present application. The method for training an encoder and a decoder comprises the following steps:

Step 401, acquire a training sample set.

In this embodiment, the execution subject of the method for training the encoder and the decoder (for example, the server 103 shown in FIG. 1 ) can obtain the training sample set.

Wherein, each training sample may include the following three parts: sample power load data and sample environment data of a first preset length, sample environment data of a second preset length, and sample power load data of a second preset length.

In some optional implementations of this embodiment, the training sample set can be obtained through the following steps:

First, obtain the historical power load data and historical environmental data of the sample enterprises.

Then, use the sliding window to segment the historical power load data and historical environmental data of the sample enterprises to generate a training sample set.

Wherein, the length of the sliding window may be equal to the sum of the first preset length and the second preset length. Each training sample may include the following three parts: historical power load data and historical environment data of a first preset length before, historical environment data of a second preset length later, and historical power load data of a second preset length later.

Step 402, for a training sample in the training sample set, use the sample power load data and sample environment data of the first preset length in the training sample as the input of the encoder, and use the output of the encoder and the first preset length of the sample environment data in the training sample The sample environment data of two preset lengths is used as the input of the decoder, and the sample power load data of the second preset length in the training samples is used as the output of the decoder to train the encoder and the decoder.

In this embodiment, for the training samples in the training sample set, the execution subject can use the sample power load data and sample environment data of the first preset length in the training samples as the input of the encoder, and the output of the encoder And the sample environment data of the second preset length in the training sample is used as the input of the decoder, the sample power load data of the second preset length in the training sample is used as the output of the decoder, and the encoder and the decoder are trained .

Specifically, the above execution subject can train the encoder and decoder based on the training sample set through the following steps:

First, initialize the encoder and decoder.

Here, the above execution subject can initialize the encoder and decoder by setting initial network parameters for each layer of the encoder and decoder.

Afterwards, for a training sample in the training sample set, the sample power load data and the sample environment data of the first preset length in the training sample are input to the encoder to obtain the sample power load feature corresponding to the training sample.

Then, the sample power load feature corresponding to the training sample and the sample environment data of the second preset length in the training sample are input to the decoder to obtain the sample power load prediction data of the second preset length corresponding to the training sample.

Finally, network parameters of each layer of the encoder and decoder are adjusted based on the sample power load prediction data of the second preset length corresponding to the training sample and the sample power load data of the second preset length in the training sample.

Usually, the above execution subject can calculate the prediction accuracy of the encoder and the decoder based on the sample power load prediction data of the second preset length corresponding to the training sample and the sample power load data of the second preset length in the training sample . When the prediction accuracy meets the preset threshold, the encoder and decoder training is completed; when the prediction accuracy does not meet the preset threshold, adjust the network parameters of each layer of the encoder and decoder, and use the training sample set to continue encoding The encoder and decoder are trained, and the cycle repeats until the prediction accuracy meets the preset threshold, and the encoder and decoder training is completed.

Further referring to FIG. 5 , as an implementation of the methods shown in the above-mentioned figures, the present application provides an embodiment of a device for predicting electric load, which corresponds to the method embodiment shown in FIG. 2 , The device can be specifically applied to various electronic devices.

As shown in FIG. 5 , the apparatus 500 for predicting electric load in this embodiment may include: an acquiring unit 501 , an encoding unit 502 and a decoding unit 503 . Among them, the acquisition unit 501 is configured to acquire the historical power load data, historical environmental data and future environmental data of the enterprise to be predicted; the encoding unit 502 is configured to use a pre-trained encoder to obtain the historical power load data and historical The environmental data is encoded to obtain the historical power load characteristics of the enterprise to be predicted; the decoding unit 503 is configured to use a pre-trained decoder to decode the historical power load characteristics and future environmental data of the enterprise to be predicted to obtain the future power of the enterprise to be predicted load forecast data.

In this embodiment, in the device 500 for predicting power load: the specific processing of the acquisition unit 501, the encoding unit 502, and the decoding unit 503 and the technical effects brought about by them can refer to step 201 in the corresponding embodiment in FIG. 2 , step 202 and step 203 related descriptions will not be repeated here.

In some optional implementation manners of this embodiment, the encoder is a convolutional neural network, and the decoder is a long short-term memory network.

In some optional implementations of this embodiment, the encoding unit 502 is further configured to: generate historical two-dimensional data of the enterprise to be tested based on the historical power load data and historical environment data of the enterprise to be predicted, wherein the historical two-dimensional The data includes time dimension and space dimension, the time dimension includes historical time points, and the spatial dimension includes power load data and environmental data at each historical time point; based on historical two-dimensional data, the historical image of the enterprise to be predicted is generated; the enterprise to be predicted The historical image is input to the convolutional neural network to obtain the historical power load characteristics of the enterprise to be predicted.

In some optional implementations of this embodiment, the convolutional neural network includes an input network, a residual neural network, and an output network, and the operations involved in the input network include at least one of the following: convolution, batch normalization, and activation function transformation, The operations involved in the residual neural network include at least one of the following: convolution, random deactivation, activation function transformation, batch normalization and maximum pooling, and the operations involved in the output network include at least one of the following: batch normalization and average pooling.

In some optional implementation manners of this embodiment, the residual neural network merges the output of the input network with the output of the residual neural network after the maximum pooling through direct connection.

In some optional implementation manners of this embodiment, the long-short-term memory network includes multiple long-short-term memory units, and each long-short-term memory unit includes an input gate, a forget gate, and an output gate.

In some optional implementations of this embodiment, the encoder and the decoder are trained through the following steps: obtaining a training sample set, wherein each training sample includes the following three parts: sample power load data of a first preset length And the sample environment data, the sample environment data of the second preset length and the sample power load data of the second preset length; for the training samples in the training sample set, the sample power load of the first preset length in the training samples The data and the sample environment data are used as the input of the encoder, the output of the encoder and the sample environment data of the second preset length in the training sample are used as the input of the decoder, and the sample of the second preset length in the training sample The power load data is used as the output of the decoder to train the encoder and decoder.

In some optional implementations of this embodiment, obtaining the training sample set includes: obtaining historical power load data and historical environmental data of the sample enterprise; using a sliding window to segment the historical electric load data and historical environmental data of the sample enterprise , to generate a training sample set, wherein the length of the sliding window is equal to the sum of the first preset length and the second preset length, and each training sample includes the following three parts: the historical power load data of the first preset length and the historical environment data, the historical environmental data of the second preset length, and the historical power load data of the second preset length.

Referring now to FIG. 6 , it shows a schematic structural diagram of a computer system 600 suitable for implementing the electronic device (such as the server 103 shown in FIG. 1 ) of the embodiment of the present application. The electronic device shown in FIG. 6 is only an example, and should not limit the functions and scope of use of this embodiment of the present application.

As shown in FIG. 6 , a computer system 600 includes a central processing unit (CPU) 601 that can be programmed according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage section 608 into a random-access memory (RAM) 603 Instead, various appropriate actions and processes are performed. In the RAM 603, various programs and data necessary for the operation of the system 600 are also stored. The CPU 601 , ROM 602 , and RAM 603 are connected to each other through a bus 604 . An input/output (I/O) interface 605 is also connected to the bus 604 .

The following components are connected to the I/O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage section 608 including a hard disk, etc. and a communication section 609 including a network interface card such as a LAN card, a modem, or the like. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I/O interface 605 as needed. A removable medium 611 such as a magnetic disk, optical disk, magneto-optical disk, semiconductor memory, etc. is mounted on the drive 610 as necessary so that a computer program read therefrom is installed into the storage section 608 as necessary.

In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, where the computer program includes program codes for executing the methods shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from a network via communication portion 609 and/or installed from removable media 611 . When the computer program is executed by the central processing unit (CPU) 601, the above-mentioned functions defined in the method of the present application are performed.

It should be noted that the computer-readable medium described in this application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, electrical connections with one or more wires, portable computer diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable Programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, however, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, in which computer-readable program codes are carried. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. . Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Computer program code for carrying out the operations of this application can be written in one or more programming languages, or combinations thereof, including object-oriented programming languages—such as Java, Smalltalk, C++, and conventional A procedural programming language—such as "C" or a similar programming language. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or electronic device. In cases involving a remote computer, the remote computer can be connected to the user computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (such as through an Internet service provider). Internet connection).

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in a flowchart or block diagram may represent a module, program segment, or portion of code that contains one or more logical functions for implementing specified executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved. It should also be noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by a dedicated hardware-based system that performs the specified functions or operations , or may be implemented by a combination of dedicated hardware and computer instructions.

The units involved in the embodiments described in the present application may be implemented by means of software or by means of hardware. The described units may also be set in a processor, for example, it may be described as: a processor includes an acquisition unit, an encoding unit, and a decoding unit. Among them, the names of these units do not constitute a limitation of the unit itself in this case, for example, the acquisition unit can also be described as "the unit that acquires the historical power load data, historical environmental data and future environmental data of the enterprise to be predicted" .

As another aspect, the present application also provides a computer-readable medium. The computer-readable medium may be included in the electronic device described in the above-mentioned embodiments; or it may exist independently without being assembled into the electronic device. middle. The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed by the electronic device, the electronic device: acquires historical power load data, historical environmental data, and future environmental data of the enterprise to be predicted; Use the pre-trained encoder to encode the historical power load data and historical environmental data of the enterprise to be predicted to obtain the historical power load characteristics of the enterprise to be predicted; use the pre-trained decoder to encode the historical power load characteristics and future environmental data of the enterprise to be predicted Decode to obtain the future power load forecast data of the enterprise to be forecasted.

The above description is only a preferred embodiment of the present application and an illustration of the applied technical principle. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solution formed by the specific combination of the above-mentioned technical features, and should also cover the technical solutions formed by the above-mentioned technical features or without departing from the above-mentioned inventive concept. Other technical solutions formed by any combination of equivalent features. For example, a technical solution formed by replacing the above-mentioned features with technical features with similar functions disclosed in this application (but not limited to).

Claims (11)

1. a kind of method for predicting electric load, comprising:

Obtain history Power system load data, history environment data and the FUTURE ENVIRONMENT data of enterprise to be predicted；

It is carried out using history Power system load data and history environment data of the encoder trained in advance to the enterprise to be predicted Coding, obtains the history electric load feature of the enterprise to be predicted；

It is carried out using history electric load feature and FUTURE ENVIRONMENT data of the decoder trained in advance to the enterprise to be predicted Decoding, obtains the future electrical energy load prediction data of the enterprise to be predicted.

2. the decoder is length according to the method described in claim 1, wherein, the encoder is convolutional neural networks Phase memory network.

3. according to the method described in claim 2, wherein, utilization encoder trained in advance is to the enterprise to be predicted History Power system load data and history environment data are encoded, and the history electric load feature of the enterprise to be predicted is obtained, Include:

History Power system load data and history environment data based on the enterprise to be predicted generate the history of the enterprise to be measured 2-D data, wherein the history 2-D data includes time dimension and Spatial Dimension, and the time dimension includes historical time Point, the Spatial Dimension include the Power system load data and environmental data of each historical time point；

Based on the history 2-D data, the history image of the enterprise to be predicted is generated；

The history image of the enterprise to be predicted is input to the convolutional neural networks, obtains the history of the enterprise to be predicted Electric load feature.

4. according to the method described in claim 2, wherein, the convolutional neural networks include input network, residual error neural network With output network, the operation that the input network is related to includes at least one of the following: that convolution, batch standardization and activation primitive become It changes, the operation that the residual error neural network is related to includes at least one of the following: convolution, random inactivation, activation primitive transformation, batch mark Standardization and maximum pond, the operation that the output network is related to include at least one of the following: batch standardization and average pond.

5. according to the method described in claim 4, wherein, the residual error neural network is by direct-connected by the defeated of the input network Merge behind maximum pond with the output of the residual error neural network out.

6. according to the method described in claim 2, wherein, the shot and long term memory network includes multiple shot and long term memory units, Each shot and long term memory unit includes input gate, forgets door and out gate.

7. method described in one of -6 according to claim 1, wherein the encoder and the decoder are instructed as follows It gets:

Obtain training sample set, wherein each training sample includes following three parts: the sample power load of the first preset length The sample electric load number of lotus data and sample environment data, the sample environment data of the second preset length and the second preset length According to；

For the training sample in the training sample set, by the sample power load of the first preset length in the training sample The input of lotus data and sample environment data as the encoder, by the output of the encoder and the training sample Input of the sample environment data of two preset lengths as the decoder, by the sample of the second preset length in the training sample Output of this Power system load data as the decoder, trains the encoder and the decoder.

8. according to the method described in claim 7, wherein, the acquisition training sample set, comprising:

Obtain the history Power system load data and history environment data of sample companies；

It is split using history Power system load data and history environment data of the sliding window to the sample companies, generates training Sample set, wherein the length of the sliding window is equal to the sum of first preset length and second preset length, each Training sample includes following three parts: the history Power system load data and history environment data of preceding first preset length, rear second The history Power system load data of the history environment data of preset length and rear second preset length.

9. a kind of for predicting the device of electric load, comprising:

Acquiring unit is configured to obtain history Power system load data, history environment data and the FUTURE ENVIRONMENT of enterprise to be predicted Data；

Coding unit, be configured to using encoder trained in advance to the history Power system load data of the enterprise to be predicted and History environment data are encoded, and the history electric load feature of the enterprise to be predicted is obtained；

Decoding unit, be configured to using decoder trained in advance to the history electric load feature of the enterprise to be predicted and FUTURE ENVIRONMENT data are decoded, and obtain the future electrical energy load prediction data of the enterprise to be predicted.

10. a kind of electronic equipment, comprising:

One or more processors；

Storage device is stored thereon with one or more programs,

When one or more of programs are executed by one or more of processors, so that one or more of processors are real Now such as method described in any one of claims 1-8.

11. a kind of computer-readable medium, is stored thereon with computer program, wherein the computer program is held by processor Such as method described in any one of claims 1-8 is realized when row.