Short-term load prediction method and system for nested residual error correction

Abstract

The invention discloses a nesting residual error correction-based short-term load prediction method and system, and the method comprises the steps: obtaining a power load sequence at K moments in historical data, carrying out the load prediction of the power load at the K moments through a neural network prediction model, and obtaining an initial prediction value; performing subtraction on the power load sequence and the initial predicted value to obtain a residual sequence; predicting the residual error sequence by using a neural network prediction model to obtain a predicted value of the residual error sequence; adding the predicted value of the initial value and the predicted value of the residual error sequence to obtain a final predicted value, taking an error between the final predicted value and an actual load as training loss, training network parameters of a prediction neural network, and obtaining a trained nested residual error correction prediction model; and performing short-term load prediction through the trained nested residual error correction prediction model. Prediction is carried out through an error correction method, an initial predicted value is subjected to predictive correction by adopting a predicted value of a residual sequence, and the short-term load prediction effect of the power system is improved.

Description

Technical Field

[0001] The invention relates to power load forecasting, and in particular to a short-term load forecasting method and system with nested residual error correction. Background Art

[0002] The tight power supply and greater demand for electricity have put forward higher power supply requirements for the power grid. However, power generation, power supply and power consumption are carried out at the same time, and it is still difficult to store large amounts of electricity. Therefore, in order to match the power generation and power consumption levels as much as possible and formulate the most reasonable power generation plan, it is necessary to accurately predict the power load of the power grid in order to make a reasonable power production plan. In the short term, the day is used as the unit, so that the power generation unit can adjust the power generation strategy according to the week and month. The prediction of short-term power load is the basis for the good operation of power generation units, which helps to customize power generation plans and power dispatch planning.

[0003] Considering the non-stationary, nonlinear and highly random characteristics of regional power load time series data, a single model has poor feature extraction capability and low accuracy in directly predicting load. Summary of the invention

[0004] Purpose of the invention: In view of the above shortcomings, the present invention provides a short-term load forecasting method and system with nested residual error correction that has good short-term load forecasting effect for power systems.

[0005] Technical solution: To solve the above problems, the present invention adopts a short-term load forecasting method with nested residual error correction, which includes the following steps:

[0006] Step 1: Obtain the power load sequence of K moments in the historical data, and use the neural network prediction model to predict the power load at K moments based on the historical data to obtain the initial value prediction value;

[0007] Step 2: Subtract the power load sequence from the initial predicted value to obtain the residual sequence;

[0008] Step 3: Use the neural network prediction model to predict the residual sequence and obtain the predicted value of the residual sequence;

[0009] Step 4: Add the initial value prediction value and the prediction value of the residual sequence to obtain the final prediction value, use the error between the final prediction value and the actual load as the training loss, train the network parameters of the prediction neural network, and obtain the prediction model with trained nested residual error correction;

[0010] Step 5: Perform short-term load forecasting by training the nested residual error correction forecasting model.

[0011] Furthermore, in step 3, the residual sequence is predicted using a neural network prediction model to obtain the predicted value of the residual sequence, which specifically includes:

[0012] Step 3.1: Perform empirical mode EMD decomposition on the residual sequence to obtain n intrinsic mode components and 1 residual component;

[0013] Step 3.2: Add the n intrinsic mode components to obtain the stationary component of the residual sequence;

[0014] Step 3.3: Use the neural network prediction model to predict the stationary component of the residual sequence to obtain the predicted value of the stationary component of the residual sequence; at the same time, use the neural network prediction model to predict the residual component of the residual sequence to obtain the predicted value of the residual component of the residual sequence;

[0015] Step 3.4: Add the predicted value of the stationary component of the residual sequence to the predicted value of the residual component of the residual sequence to obtain the predicted value of the residual sequence.

[0016] Furthermore, the neural network prediction model in step 1 adopts a bidirectional long short-term memory network BiLSTM. The neural network prediction model in step 3.3 adopts a bidirectional long short-term memory network BiLSTM.

[0017] The present invention also adopts a short-term load forecasting system with nested residual error correction, comprising:

[0018] The data acquisition module is used to obtain the power load sequence of K moments in the historical data, and at the same time, use the neural network prediction model to perform load prediction on the power load at K moments according to the historical data to obtain the initial value prediction value;

[0019] The residual sequence calculation module is used to make a difference between the power load sequence and the initial value prediction value to obtain the residual sequence;

[0020] The residual sequence prediction module is used to predict the residual sequence using a neural network prediction model to obtain the predicted value of the residual sequence;

[0021] The prediction model training module is used to add the initial prediction value and the prediction value of the residual sequence to obtain the final prediction value, use the error between the final prediction value and the actual load as the training loss, train the network parameters of the prediction neural network, and obtain the prediction model with nested residual error correction.

[0022] The forecasting module is used to perform short-term load forecasting by training a nested residual error-corrected forecasting model.

[0023] The present invention also adopts a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0024] The present invention also adopts a computer-readable storage medium on which a computer program is stored, and the computer program implements the steps of the above method when executed by a processor.

[0025] Beneficial effect: Compared with the prior art, the significant advantage of the present invention is that most of the current regional short-term prediction methods are direct prediction methods based on models, and their prediction accuracy is often affected by the randomness of data. The present invention proposes an error correction method for prediction, and uses the prediction value of the residual sequence to make a progressive correction to the initial prediction value, thereby improving the effect of short-term load prediction of the power system. The present invention embeds a new residual error correction in the residual sequence prediction, decomposes the residual sequence into a stationary component and a residual component using the empirical mode EMD, and separately uses neural networks to model and predict these two components. The sum of the prediction values ​​of the two networks is used as the prediction value of the residual sequence, and the prediction accuracy is higher. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a schematic diagram of the process of error correction in short-term load forecasting of the present invention. DETAILED DESCRIPTION

[0027] Example 1

[0028] In this embodiment, a short-term load forecasting method with nested residual error correction includes the following steps:

[0029] Step 1: For the K-time power load sequence [x t-K+1 , …, x t ], the current load is x t , use the bidirectional long short time network BiLSTM (denoted as function f) to perform load forecasting and obtain the initial value forecast value

[0030] Step 2: Transform the power load sequence [x t-K+1 , …, x t ] and the initial predicted value Subtract and get the residual sequence

[0031] Step 3: Use the neural network prediction model to predict the residual sequence and obtain the predicted value of the residual sequence. Specifically include:

[0032] Step 3.1: For the residual sequence Perform EMD decomposition to obtain n intrinsic mode components and 1 residual component. The i-th eigenmode component is recorded as The residual component is recorded as [r t-K+1 ,…,r t ], then

[0033] Step 3.2: Add the n intrinsic mode components to obtain the stationary component of the residual sequence, denoted as

[0034] Step 3.3: For the stationary component of the residual sequence [c t-K+1 , …, c t ], use the bidirectional long short time network BiLSTM (denoted as function g) to make predictions and obtain the predicted value c of the stationary component of the residual sequence pre =g([c t-K+1 , …, c t ]). The residual component of the residual sequence [r t-K+1 ,…,r t ], use the bidirectional long short time network BiLSTM (represented as function h) to make predictions and get the predicted value r of the residual component of the residual sequence pre =h([r t-K+1 ,…,r t ]).

[0035] Step 3.4: Set the predicted value c of the stationary component of the residual sequence pre and the predicted value r of the residual component of the residual sequence pre Add together to get the predicted value of the residual sequence That is, the prediction of the residual sequence is error corrected.

[0036] Step 4: The initial value prediction value obtained in step 1 The predicted value of the residual sequence obtained in step 3 Add together to get the final predicted value That is, the error correction is performed again on the final load forecast value. Calculate the final load forecast value y pre The error from the actual load y, the actual load is also recorded as x t+1 =y, calculate the training loss L = (y pre -y) 2 , train the network parameters f, g and h of the prediction neural network, and obtain the trained nested residual error corrected prediction model.

[0037] Step 5: Perform short-term load forecasting by training the nested residual error correction forecasting model.

[0038] Example 2

[0039] In this embodiment, a short-term load forecasting system with nested residual error correction includes:

[0040] The data acquisition module is used to obtain the power load sequence of K moments in the historical data, and at the same time, use the neural network prediction model to perform load prediction on the power load at K moments according to the historical data to obtain the initial value prediction value;

[0041] The residual sequence calculation module is used to make a difference between the power load sequence and the initial value prediction value to obtain the residual sequence;

[0042] The residual sequence prediction module is used to predict the residual sequence using a neural network prediction model to obtain the predicted value of the residual sequence;

[0043] The prediction model training module is used to add the initial prediction value and the prediction value of the residual sequence to obtain the final prediction value, use the error between the final prediction value and the actual load as the training loss, train the network parameters of the prediction neural network, and obtain the prediction model with nested residual error correction.

[0044] The forecasting module is used to perform short-term load forecasting by training a nested residual error-corrected forecasting model.

[0045] Example 3

[0046] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0047] The present invention is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0048] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0049] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0050] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A short-term load forecasting method with nested residual error correction, characterized in that: The following steps are involved: Step 1: Obtain the power load sequence of K moments in the historical data, and use the neural network prediction model to predict the power load at K moments based on the historical data to obtain the initial value prediction value; Step 2: Subtract the power load sequence from the initial predicted value to obtain the residual sequence; Step 3: Use the neural network prediction model to predict the residual sequence and obtain the predicted value of the residual sequence; Step 4: Add the initial value prediction value and the prediction value of the residual sequence to obtain the final prediction value, use the error between the final prediction value and the actual load as the training loss, train the network parameters of the prediction neural network, and obtain the prediction model with trained nested residual error correction; Step 5: Perform short-term load forecasting by training the nested residual error correction forecasting model.

2. The short-term load forecasting method with nested residual error correction according to claim 1, characterized in that: In step 3, the residual sequence is predicted using a neural network prediction model to obtain the predicted value of the residual sequence, which specifically includes: Step 3.1: Perform empirical mode EMD decomposition on the residual sequence to obtain n intrinsic mode components and 1 residual component; Step 3.2: Add the n intrinsic mode components to obtain the stationary component of the residual sequence; Step 3.3: Use the neural network prediction model to predict the stationary component of the residual sequence to obtain the predicted value of the stationary component of the residual sequence; at the same time, use the neural network prediction model to predict the residual component of the residual sequence to obtain the predicted value of the residual component of the residual sequence; Step 3.4: Add the predicted value of the stationary component of the residual sequence to the predicted value of the residual component of the residual sequence to obtain the predicted value of the residual sequence.

3. The short-term load forecasting method with nested residual error correction according to claim 1, characterized in that: The neural network prediction model in step 1 adopts a bidirectional long short-term memory network BiLSTM.

4. The small sample power load forecasting method based on transfer learning according to claim 2 is characterized in that: The neural network prediction model in step 3.3 adopts a bidirectional long short-term memory network BiLSTM.

5. A short-term load forecasting system with nested residual error correction, characterized in that: include: The data acquisition module is used to obtain the power load sequence of K moments in the historical data, and at the same time, use the neural network prediction model to perform load prediction on the power load at K moments according to the historical data to obtain the initial value prediction value; The residual sequence calculation module is used to make a difference between the power load sequence and the initial value prediction value to obtain the residual sequence; The residual sequence prediction module is used to predict the residual sequence using a neural network prediction model to obtain the predicted value of the residual sequence; The prediction model training module is used to add the initial prediction value and the prediction value of the residual sequence to obtain the final prediction value, use the error between the final prediction value and the actual load as the training loss, train the network parameters of the prediction neural network, and obtain the prediction model with nested residual error correction. The forecasting module is used to perform short-term load forecasting by training a nested residual error-corrected forecasting model.

6. The short-term load forecasting system with nested residual error correction according to claim 5, characterized in that: The residual sequence prediction module uses a neural network prediction model to predict the residual sequence, and the predicted value of the residual sequence is obtained specifically including: Perform empirical mode EMD decomposition on the residual sequence to obtain n intrinsic mode components and 1 residual component; Add the n intrinsic mode components to obtain the stationary component of the residual sequence; The stationary component of the residual sequence is predicted using a neural network prediction model to obtain the predicted value of the stationary component of the residual sequence; at the same time, the residual component of the residual sequence is predicted using a neural network prediction model to obtain the predicted value of the residual component of the residual sequence; Add the predicted value of the stationary component of the residual sequence to the predicted value of the residual component of the residual sequence to obtain the predicted value of the residual sequence.

7. The short-term load forecasting system with nested residual error correction according to claim 5, characterized in that: The neural network prediction model in the data acquisition module adopts a bidirectional long short-term memory network BiLSTM.

8. The system according to claim 6, characterized in that The neural network prediction model in the residual sequence prediction module adopts a bidirectional long short-term memory network BiLSTM.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images