Load prediction method based on neural network

Through the load prediction method based on neural network, the external influencing factors are integrated to construct and optimize the neural network model, the problem of low load prediction accuracy in the power system in the existing technology is solved, and more accurate and flexible short-term and long-term load prediction is achieved.

CN118982260AInactive Publication Date: 2024-11-19NANJING ZHONGHUI ELECTRIC TECH CO LTD

Patent Information

Application Number
CN202411149676.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-21
Publication Date
2024-11-19
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing power system load prediction methods are limited by the linear assumptions of the model and the simplified processing of external influencing factors, resulting in low prediction accuracy and difficult to meet short-term and long-term prediction needs.

Method used

Using a load prediction method based on neural network, a neural network model for load prediction of power system is constructed through data acquisition and preprocessing, and data of external influencing factors are integrated for training and optimization to achieve short-term and long-term load prediction.

Benefits of technology

It improves the accuracy and accuracy of load prediction in power system, can flexibly adapt to different types of load data and external influencing factors, enhances the adaptability and robustness of prediction, and meets the load prediction needs at different time scales.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982260A_ABST
    Figure CN118982260A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of power system load prediction, and discloses a load prediction method based on a neural network, and the method comprises the following steps: SP01, data collection and preprocessing: collecting historical load data and external influence factor data, and carrying out the preprocessing; sP02, constructing a neural network model used for power system load prediction; sP03, integrating external factors, combining external influence factor data, taking the external factors as additional input features, and adding the external factors into the neural network model for predicting the load of the power system; sP04, training and optimizing a neural network model used for power system load prediction; according to the power system load prediction method, the accuracy and precision of power system load prediction are improved, the neural network model can flexibly adapt to different types of load data and external influence factors, the power system load can be predicted at different time scales, and the power system load prediction accuracy is improved. And short-term and long-term prediction requirements on the load of the power system can be met.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of load forecasting for electric power systems, and more specifically, to a load forecasting method based on a neural network. Background Art

[0002] The function of the power system is to convert the primary energy in nature into electrical energy through power generation devices, and then supply the electrical energy to various users through transmission, transformation and distribution. To achieve this function, the power system also has corresponding information and control systems in various links and different levels to measure, regulate, control, protect, communicate and dispatch the production process of electrical energy to ensure that users obtain safe and high-quality electrical energy. However, the existing power system load forecasting method may be limited by the linear assumptions of the model and the simplified processing of external influencing factors, resulting in low accuracy of power system load forecasting, and often only a certain time scale can be predicted, which is difficult to meet the short-term and long-term forecasting needs of the power system load. Based on this, the present invention designs a load forecasting method based on a neural network to solve the above problems. Summary of the invention

[0003] In order to overcome the deficiencies of the prior art, the present invention provides a load forecasting method based on a neural network to solve the problems raised in the above background technology.

[0004] To achieve the above object, the present invention provides a load forecasting method based on a neural network, comprising the following steps: SP01: Data collection and preprocessing, collecting historical load data and external influencing factor data and preprocessing; SP02: Construct a neural network model for power system load forecasting; SP03: Integrate external factors, combine external influencing factor data, and add them as additional input features to the neural network model for power system load forecasting; SP04: Training and optimizing neural network models for power system load forecasting; SP05: Output the prediction results and use the trained neural network model for power system load forecasting to make short-term and long-term predictions of power system load.

[0005] As a preferred technical solution of the present invention, the data collection and preprocessing process includes the following steps: SP11: Use power system monitoring equipment and sensors to collect historical load data and external influencing factor data in real time; SP12: Process missing values ​​for the collected data; SP13: Perform outlier detection on data after missing value processing.

[0006] As a preferred technical solution of the present invention, the external influencing factor data specifically includes temperature, humidity, weather conditions, seasons and holidays influencing factor data.

[0007] As a preferred technical solution of the present invention, the missing value processing of the collected data is specifically to fill the missing part of the data with time and replace it with null values, convert the missing part into a null value area, split the data to process the continuous missing value part of the null value area, estimate the missing value based on the split data, evaluate and process the estimated value to determine whether it meets the backfilling condition, and backfill the estimated value that meets the backfilling condition.

[0008] As a preferred technical solution of the present invention, the outlier detection on the data after missing value processing is specifically to filter the data according to the filtering threshold, filter out the required data from the data after missing value processing, identify and eliminate the outlier data in the required data according to the detection conditions, so as to obtain normal value data.

[0009] As a preferred technical solution of the present invention, the process of constructing a neural network model for power system load forecasting includes the following steps: SP21: Design a neural network structure and parameters of a neural network model for power system load forecasting, the parameters including the number of neurons, the number of layers, and the activation function; SP22: Select an appropriate neural network structure based on data characteristics and forecasting requirements to build a neural network model for power system load forecasting.

[0010] As a preferred technical solution of the present invention, the neural network structure includes an MLP multi-layer perceptron, an RNN recursive neural network and an LSTM long short-term memory network.

[0011] As a preferred technical solution of the present invention, the external factor integration process includes the following steps: SP31: Integrate external influencing factor data with load data to construct a complete input feature vector; SP32: Design the input layer and input feature representation model to effectively combine the external factor data with the load data and input them into the neural network model for power system load forecasting.

[0012] As a preferred technical solution of the present invention, the process of training and optimizing the neural network model for power system load forecasting includes the following steps: SP41: Using historical load data and external influencing factor data, the neural network model for power system load forecasting is trained and optimized through training algorithms; SP42: Use appropriate loss functions and optimization algorithms to adjust model parameters so that they can better fit historical data.

[0013] As a preferred technical solution of the present invention, the process of outputting prediction results includes the following steps: SP51: Output prediction results using the trained neural network model for power system load forecasting; SP52: Generate corresponding load forecast curve and time series data according to the forecast results, characterize the forecast results in time series, and sort the nodes according to the characterization results; SP53: Classify and concatenate the time series corresponding to each node in the order of hours, days, weeks, months and years to obtain the power system load forecast information corresponding to the hours, days, weeks, months and years, so that the power system load can be predicted at different time scales to make short-term and long-term forecasts of the power system load.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention uses a neural network algorithm to perform load forecasting, which can better capture the nonlinear and time-varying characteristics of load data, thereby improving the accuracy and precision of power system load forecasting. The neural network model can flexibly adapt to different types of load data and external influencing factors, has strong generalization ability, and can better adapt and handle complex and changeable power system loads, thereby improving the adaptability and robustness of the forecast.

[0015] The present invention can make the power system load predicted at different time scales, which is convenient for meeting the short-term and long-term prediction needs of the power system load, can more comprehensively consider the changing laws of the load at different time scales, and improve the comprehensiveness and operability of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of the steps of a load forecasting method based on a neural network according to the present invention; Figure 2 A schematic diagram of the steps of data collection and preprocessing process in a load forecasting method based on a neural network of the present invention; Figure 3 A schematic diagram of the steps of constructing a neural network model process for power system load forecasting in a neural network-based load forecasting method of the present invention; Figure 4 A schematic diagram of the steps of integrating external factors in a load forecasting method based on a neural network according to the present invention; Figure 5A schematic diagram of the steps of training and optimizing a neural network model process for power system load forecasting in a neural network-based load forecasting method of the present invention; Figure 6 The present invention is a schematic diagram of the steps of outputting the prediction result flow in a load prediction method based on a neural network. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] like Figures 1 to 6 As shown, the present invention provides a load forecasting method based on a neural network, comprising the following steps: SP01: Data collection and preprocessing, collecting historical load data and external influencing factor data and preprocessing; SP02: Construct a neural network model for power system load forecasting; SP03: Integrate external factors, combine external influencing factor data, and add them as additional input features to the neural network model for power system load forecasting; SP04: Training and optimizing neural network models for power system load forecasting; SP05: Output the prediction results and use the trained neural network model for power system load forecasting to make short-term and long-term predictions of power system load.

[0019] Among them, Figure 2 As shown, the data collection and preprocessing process includes the following steps: SP11: Use power system monitoring equipment and sensors to collect historical load data and external influencing factor data in real time; SP12: Process missing values ​​for the collected data; SP13: Perform outlier detection on data after missing value processing.

[0020] The external influencing factor data specifically include temperature, humidity, weather conditions, seasons and holidays.

[0021] Specifically, the missing value processing of the collected data includes filling the missing part of the data with time and replacing it with null values, converting the missing part into a null value area, splitting the data to process the continuous missing value part of the null value area, estimating the missing values ​​based on the split data, evaluating and processing the estimated values ​​to determine whether they meet the backfilling conditions, and backfilling the estimated values ​​that meet the backfilling conditions.

[0022] Among them, outlier detection on the data after missing value processing is specifically to filter the data according to the filtering threshold, filter out the required data from the data after missing value processing, identify and eliminate the outlier data in the required data according to the detection conditions, so as to obtain normal value data, so as to ensure data quality and availability.

[0023] Among them, Figure 3 As shown in the figure, the process of building a neural network model for power system load forecasting includes the following steps: SP21: Design the neural network structure and parameters of the neural network model for power system load forecasting, including the number of neurons, number of layers, and activation function; SP22: Select an appropriate neural network structure based on data characteristics and forecasting requirements to build a neural network model for power system load forecasting.

[0024] Among them, the neural network structure includes MLP multi-layer perceptron, RNN recursive neural network and LSTM long short-term memory network.

[0025] Among them, Figure 4 As shown, the external factor integration process includes the following steps: SP31: Integrate external influencing factor data with load data to construct a complete input feature vector; SP32: Design the input layer and input feature representation model to effectively combine the external factor data with the load data and input them into the neural network model for power system load forecasting.

[0026] This enables the neural network model used for power system load forecasting to more comprehensively consider the impact of external factors on load and improve the accuracy of forecasting.

[0027] Among them, Figure 5 As shown in Figure 1, the process of training and optimizing the neural network model for power system load forecasting includes the following steps: SP41: Using historical load data and external influencing factor data, the neural network model for power system load forecasting is trained and optimized through training algorithms; SP42: Use appropriate loss functions and optimization algorithms to adjust model parameters so that they can better fit historical data.

[0028] Among them, Figure 6 As shown, the process of outputting prediction results includes the following steps: SP51: Output prediction results using the trained neural network model for power system load forecasting; SP52: Generate corresponding load forecast curve and time series data according to the forecast results, characterize the forecast results in time series, and sort the nodes according to the characterization results; SP53: Classify and concatenate the time series corresponding to each node in the order of hours, days, weeks, months and years to obtain the power system load forecast information corresponding to the hours, days, weeks, months and years, so that the power system load can be predicted at different time scales to make short-term and long-term forecasts of the power system load.

[0029] In summary, the present invention can better capture the nonlinear and time-varying characteristics of load data by adopting a neural network algorithm for load forecasting, thereby improving the accuracy and precision of power system load forecasting. The neural network model can flexibly adapt to different types of load data and external influencing factors, and has a strong generalization ability. When faced with complex and changeable power system loads, it can better adapt and handle them, thereby improving the adaptability and robustness of the forecast, and can enable the power system load to be predicted at different time scales, thereby facilitating the short-term and long-term forecasting needs of the power system load, and can more comprehensively consider the changing laws of loads at different time scales, thereby improving the comprehensiveness and operability of the forecast.

[0030] It should be noted that, in this article, the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or apparatus.

[0031] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A load forecasting method based on neural network, characterized in that: The following steps are involved: SP01: Data collection and preprocessing, collecting historical load data and external influencing factor data and preprocessing; SP02: Construct a neural network model for power system load forecasting; SP03: Integrate external factors, combine external influencing factor data, and add them as additional input features to the neural network model for power system load forecasting; SP04: Training and optimizing neural network models for power system load forecasting; SP05: Output forecast results, using the trained neural network model for power system load forecasting to make short-term and long-term forecasts of power system load; The data collection and preprocessing process includes the following steps: SP11: Use power system monitoring equipment and sensors to collect historical load data and external influencing factor data in real time; SP12: Process missing values ​​of the collected data, fill in the missing parts of the data with time and replace them with null values, convert the missing parts into null value areas, split the data to process the continuous missing value parts of the null value areas, estimate the missing values ​​based on the split data, evaluate the estimated values ​​to determine whether they meet the backfilling conditions, and backfill the estimated values ​​that meet the backfilling conditions; SP13: Perform outlier detection on the data after missing value processing, filter the data according to the filtering threshold, filter out the required data from the data after missing value processing, identify and eliminate the outlier data in the required data according to the detection conditions to obtain normal value data.

2. The load forecasting method based on neural network according to claim 1 is characterized in that: The process of constructing a neural network model for power system load forecasting includes the following steps: SP21: Design a neural network structure and parameters of a neural network model for power system load forecasting, the parameters including the number of neurons, the number of layers, and the activation function; SP22: Select an appropriate neural network structure based on data characteristics and forecasting requirements to build a neural network model for power system load forecasting.

3. The load forecasting method based on neural network according to claim 2 is characterized in that: The neural network structure includes an MLP multi-layer perceptron, an RNN recursive neural network and an LSTM long short-term memory network.

4. The load forecasting method based on neural network according to claim 2 is characterized in that: The process of integrating external factors includes the following steps: SP31: Integrate external influencing factor data with load data to construct a complete input feature vector; SP32: Design the input layer and input feature representation model to effectively combine the external factor data with the load data and input them into the neural network model for power system load forecasting.

5. The load forecasting method based on neural network according to claim 4 is characterized in that: The process of training and optimizing the neural network model for power system load forecasting includes the following steps: SP41: Using historical load data and external influencing factor data, the neural network model for power system load forecasting is trained and optimized through training algorithms; SP42: Use appropriate loss functions and optimization algorithms to adjust model parameters so that they can better fit historical data.

6. The load forecasting method based on neural network according to claim 5 is characterized in that: The output prediction result process includes the following steps: SP51: Output prediction results using the trained neural network model for power system load forecasting; SP52: Generate corresponding load forecast curve and time series data according to the forecast results, characterize the forecast results in time series, and sort the nodes according to the characterization results; SP53: Classify and concatenate the time series corresponding to each node in the order of hours, days, weeks, months and years to obtain the power system load forecast information corresponding to the hours, days, weeks, months and years, so that the power system load can be predicted at different time scales to make short-term and long-term forecasts of the power system load.

Citation Information

Patent Citations

  • Short-period power load prediction method based on T-S fuzzy Elman nerve network

    CN106971238A

  • Short-term load prediction method based on weighted grey correlation projection method

    CN117318025A

  • Method for improving power grid load prediction precision based on neural network

    CN117578441A

Cited By

  • Power load prediction method and device based on causal characteristics, and medium

    CN119315550A