A combined electricity sales quantity prediction method, device, equipment and medium

Abstract

The application relates to a combined electricity sales quantity prediction method, device, equipment and medium, wherein the method comprises the following steps: collecting power technology parameters, associated parameters and environment adaptation parameters in real time and preprocessing; taking the preprocessed power technology parameters, associated parameters and environment adaptation parameters as inputs, outputting a first sub-period electricity sales quantity prediction value by using an LSTM neural network, and introducing a power error penalty factor to impose an additional penalty on the peak-valley period prediction error in the LSTM neural network training process; taking the sub-period electricity sales quantity as an explained variable, the preprocessed associated parameters as core explained variables, and introducing a power grid power transmission efficiency, a transformer loss coefficient and an extreme weather influence coefficient as power regulation factors to construct a generalized additive model and output a second sub-period electricity sales quantity prediction value; and calculating a fusion weight, fusing the first and second sub-period electricity sales quantity prediction values, and obtaining a final electricity sales quantity prediction result. Compared with the prior art, the application has the advantages of high prediction accuracy.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and in particular to a combined electricity sales forecasting method, apparatus, equipment and medium. Background Technology

[0002] With the deepening of power market reform and the construction of new power systems, time-of-use electricity sales forecasting serves as a core foundation for grid dispatching, power planning, and supply-demand balance control. Its forecasting accuracy directly impacts the safety, economy, and stability of grid operation. Accurate time-of-use electricity sales forecasting provides grid dispatching departments with a scientific basis for decision-making, effectively avoiding problems such as insufficient power supply during peak hours and energy waste during off-peak hours. It also supports the integration of new energy sources, the dispatching of energy storage systems, and the optimization of electricity market transactions, thus becoming a research hotspot in the field of power systems.

[0003] Currently, the main methods for predicting electricity sales include: 1. Traditional statistical models: Common models include time series analysis methods (such as the autoregressive integral moving average model ARIMA and the seasonal autoregressive integral moving average model SARIMA), generalized additive models (GAM), etc., which are based on linear assumptions, have simple structures, strong interpretability, and were widely used in electricity sales forecasting in the early days.

[0004] 2. Machine learning models: Common models include Long Short-Term Memory Neural Network (LSTM), Support Vector Machine (SVM), Random Forest (RF), etc. Due to their strong nonlinear fitting ability and time series feature extraction ability, they have gradually become the mainstream methods for electricity sales forecasting.

[0005] Chinese patent CN112418545A discloses a method and system for electricity sales forecasting based on load characteristics and model fusion. This method employs load characteristics and model fusion, decomposing electricity load characteristics into trend, seasonal, and residual components through data acquisition, feature engineering, model construction, and parameter optimization. It combines time characteristics and weather factors, using the Prophet model and gradient boosting model for short-, medium-, and long-term forecasts, and improves forecast accuracy through model fusion. However, this method fails to consider the differentiated impact of peak-valley forecasting errors in electricity scenarios, treating all time-period errors indiscriminately. This results in forecasting accuracy during critical peak-valley periods failing to meet grid dispatching requirements. Furthermore, it neglects core grid operating parameters, limiting weather considerations to temperature correlation coefficients and ignoring the abrupt impact of extreme weather on electricity consumption patterns, leading to a disconnect between forecast results and actual grid operating conditions. Summary of the Invention

[0006] The purpose of this invention is to overcome the deficiencies of the prior art by providing a combined electricity sales forecasting method, device, equipment, and medium that takes into account the impact of peak and off-peak periods, extreme weather, and power grid operation, thereby improving the accuracy of electricity sales forecasting.

[0007] The objective of this invention can be achieved through the following technical solutions: According to a first aspect of the present invention, a combined electricity sales forecasting method is provided, the method comprising the following steps: Real-time acquisition and preprocessing of power technical parameters, related parameters, and environmental adaptation parameters; The preprocessed power technical parameters, correlation parameters, and environmental adaptation parameters are used as inputs. The first time-segmented electricity sales prediction value is output using an LSTM neural network. During the training process of the LSTM neural network, an electricity error penalty factor is introduced to impose an additional penalty on the prediction error during peak and valley periods. Using time-of-use electricity sales as the dependent variable, preprocessed correlation parameters as the core explanatory variables, and introducing grid transmission efficiency, transformer loss coefficient and extreme weather impact coefficient as power regulation factors, a generalized additive model is constructed to output the second time-of-use electricity sales forecast value. Calculate the fusion weight and fuse the electricity sales forecast values ​​of the first and second time periods to obtain the final electricity sales forecast result.

[0008] The aforementioned power technical parameters include total electricity sales within the region, time-of-use load, and voltage stability indicators.

[0009] The associated parameters include the installed capacity of industrial electrical equipment, the duration of electricity consumption in the tertiary sector, and the number of residential electricity terminals.

[0010] The environmental adaptation parameters include seasonality and extreme weather indicators.

[0011] The specific details of introducing an electricity error penalty factor during LSTM neural network training to impose additional penalties on peak and valley prediction errors are as follows: Introducing a power error penalty factor to construct the loss function: ; ; in, For loss function, To predict the total number of items in different time periods within the forecast period, express Power error penalty factor for the time period express Forecast error for a given period of time express Electricity consumption time period attributes, Indicates peak electricity consumption periods. Indicates off-peak electricity consumption periods. Indicates the time period of the level. express The relative deviation of electricity load during a given time period.

[0012] The generalized additive model is expressed as follows: ; in, To sell electricity in different time periods, Indicates the first A smooth function, Indicates the installed capacity of industrial electrical equipment. Indicates the duration of electricity consumption in the tertiary sector. Indicates the number of residential electricity terminals. Indicates the power grid transmission efficiency. This represents the transformer loss coefficient. The extreme weather impact coefficient is calculated based on seasonal coefficients and extreme weather indicators. This is the random error term.

[0013] The fusion weights are determined using the entropy method, specifically including the following steps: The electricity sales forecasts for the first and second time periods were standardized and shifted respectively. The weight of the predicted electricity sales for the first and second time periods after processing is calculated using the entropy method, and the information entropy is calculated based on the weight. The difference coefficient is calculated based on the information entropy, and the difference coefficient is normalized to obtain the final fusion weight.

[0014] According to a second aspect of the present invention, a combined electricity sales forecasting device is provided, the device comprising: Data acquisition module: Real-time collection and preprocessing of power technology parameters, related parameters, and environmental adaptation parameters; The first prediction module takes the preprocessed power technical parameters, related parameters and environmental adaptation parameters as input, uses the LSTM neural network to output the first time-segmented electricity sales prediction value, and introduces a power error penalty factor during the training process of the LSTM neural network to impose an additional penalty on the prediction error during peak and valley periods. The second prediction module uses the time-segmented electricity sales volume as the explained variable, the preprocessed correlation parameters as the core explanatory variables, and introduces the power grid transmission efficiency, transformer loss coefficient and extreme weather impact coefficient as power regulation factors to construct a generalized additive model and output the predicted value of the second time-segmented electricity sales volume. Output module: Calculates the fusion weight, merges the electricity sales forecast values ​​of the first and second time periods, and obtains the final electricity sales forecast result.

[0015] According to a third aspect of the present invention, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described thereon.

[0016] According to a fourth aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method described thereon.

[0017] Compared with the prior art, the present invention has the following beneficial effects: (1) In the training process of LSTM neural network, the present invention introduces power error penalty factor to impose additional penalty on the prediction error during peak and valley periods. This can preferentially correct the prediction deviation during peak and valley periods, solve the defect of traditional model that treats all time period errors indiscriminately, and improve the prediction accuracy.

[0018] (2) This invention introduces three major power regulation factors into the generalized additive model. Among them, the power grid transmission efficiency and transformer loss coefficient capture the actual loss in the process of power transmission and consumption. The extreme weather impact coefficient accurately adapts to the sudden impact of sudden weather such as rainstorms and high temperature extremes on the electricity consumption pattern, so that the prediction results can effectively match the actual operation scenario of the power grid. The fitting accuracy of the multi-factor nonlinear coupling relationship is significantly improved, and the prediction accuracy is significantly improved.

[0019] (3) This invention integrates the prediction results of LSTM neural network and generalized additive model to achieve dual coverage of time series trend and multi-factor coupling. It solves the defects of insufficient time series fitting and one-sided capture of multi-factor relationship of traditional model. It can accurately predict the overall trend of electricity sales over time, and accurately reflect the local influence of various factors such as grid operation status and external environment. The prediction results are more comprehensive and reliable, providing high-quality decision-making basis for multiple scenarios such as grid planning, new energy consumption and electricity market transactions. Attached Figure Description

[0020] Figure 1 This is a flowchart of the prediction method of the present invention; Figure 2 This is a flowchart of the fusion weight determination process of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0022] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0023] Example 1 This embodiment provides a combined electricity sales forecasting method, such as... Figure 1 As shown, the method includes the following steps: S1 collects power technical parameters, related parameters, and environmental adaptation parameters in real time and performs preprocessing.

[0024] S11, Power technical parameter acquisition.

[0025] This step collects power technical parameters from the power grid SCADA system and electricity consumption information collection terminals, including total electricity sales, time-of-use load, and voltage stability indicators within the region. The total electricity sales within the region are obtained and aggregated from the electricity consumption information collection terminals; the time-of-use load is calculated by arithmetically averaging the real-time load values ​​obtained from the power grid SCADA system within preset forecast time periods (e.g., 15 minutes, 1 hour); the voltage stability indicator is measured using the voltage deviation rate. ; in, express Within the time period The voltage value collected this time. Indicates the rated voltage. express Within the time period Voltage deviation rate of each acquisition.

[0026] S12, Related parameter acquisition.

[0027] The installed capacity of industrial electrical equipment, the duration of electricity consumption in the tertiary industry, and the number of residential electricity terminals are statistically analyzed through IoT terminals (deployed in industrial parks and residential community power distribution nodes) and used as correlation parameters.

[0028] S13, Environment Adaptation Parameter Acquisition.

[0029] Environmental adaptation parameters include seasonality and extreme weather indicators.

[0030] The seasonal coefficient is determined as follows: the seasonal coefficient for each of the four seasons is determined by the ratio of the average historical time-period electricity sales in the region for each season to the average historical time-period electricity sales in the region for the whole year.

[0031] The method for determining extreme weather indicators is as follows: This embodiment focuses on extreme weather that significantly affects electricity sales: high temperature, low temperature, rainstorm, snowstorm, and typhoon. It directly obtains relevant historical weather data from external meteorological data interfaces, judges extreme weather based on judgment conditions (such as high temperature ≥35℃, low temperature ≤-5℃), and matches each calculated weather with a corresponding extreme weather indicator. For example, the extreme weather indicator corresponding to high temperature is 1, the extreme weather indicator corresponding to low temperature is 2, and so on. If it is not extreme weather, the corresponding extreme weather indicator is 0.

[0032] S2 takes the preprocessed power technical parameters, correlation parameters and environmental adaptation parameters as input, uses the LSTM neural network to output the first time period power sales prediction value, and introduces the power error penalty factor during the training of the LSTM neural network to impose an additional penalty on the prediction error during peak and valley periods.

[0033] In this embodiment, the input layer of the LSTM neural network contains three types of parameters (three power technology parameters, three correlation parameters, and two environmental adaptation parameters), which are input in a time sequence (the time step is set to 24 hours, corresponding to daily time-segmented data); the hidden layer has 2-3 layers of LSTM units (64-128 units); the output layer is the predicted electricity sales value for each time period.

[0034] Introducing a power error penalty factor to construct the loss function: ; ; in, For loss function, To predict the total number of items in different time periods within the forecast period, express Power error penalty factor for the time period express Forecast error for a given period of time express Electricity consumption time period attributes, Indicates peak electricity consumption periods. Indicates off-peak electricity consumption periods. Indicates the time period of the level. express The relative deviation of electricity load during a given time period.

[0035] Subsequently, based on the loss function Train the LSTM neural network.

[0036] S3 uses time-of-use electricity sales as the dependent variable, preprocessed correlation parameters as the core explanatory variables, and introduces grid transmission efficiency, transformer loss coefficient and extreme weather impact coefficient as power regulation factors to construct a generalized additive model and output the second time-of-use electricity sales forecast value.

[0037] The generalized additive model is represented as: ; in, To sell electricity in different time periods, Indicates the first A smooth function, Indicates the installed capacity of industrial electrical equipment. Indicates the duration of electricity consumption in the tertiary sector. Indicates the number of residential electricity terminals. Indicates the power grid transmission efficiency. This represents the transformer loss coefficient. The extreme weather impact coefficient is calculated based on seasonal coefficients and extreme weather indicators. This is the random error term.

[0038] In this embodiment, the power grid transmission efficiency characterizes the power transmission efficiency of the regional power grid within the predicted time period, defined as the ratio of the actual amount of electricity transmitted to users to the total amount of electricity transmitted by the power grid. The transformer loss coefficient characterizes the degree of power loss of distribution transformers within the region. Transformer losses are divided into no-load losses and load losses. Load losses are strongly correlated with the installed capacity of electrical equipment and the actual load. Specifically, the calculation method is as follows: the weighted sum of the installed capacity of industrial electrical equipment and the number of residential electrical terminals, multiplied by the corresponding time-of-use load and loss coefficient, and then divided by the total rated capacity of the regional transformers. The extreme weather impact coefficient is obtained by multiplying the seasonal coefficient by the extreme weather indicator.

[0039] S4. Calculate the fusion weight and fuse the electricity sales forecast values ​​of the first and second time periods to obtain the final electricity sales forecast result.

[0040] Methods for calculating fusion weights include, but are not limited to, empirical weighting, inverse error ratio, inverse variance, entropy weighting, cross-validation, analytic hierarchy process (AHP), and stacking weighting. This embodiment does not limit the specific implementation of these methods.

[0041] Example 2 This embodiment, based on Embodiment 1, provides a specific implementation method for determining fusion weights using the entropy method, such as... Figure 2 As shown, the specific steps include: Forecast of electricity sales for the first time period Forecast of electricity sales in the second time period Standardization was performed using the range method: ; ; in, Indicates the first i The prediction method obtained the first j Forecasted electricity sales for a given time period , This represents the minimum and maximum electricity sales predicted by the LSTM neural network. , These represent the minimum and maximum electricity sales predicted by the generalized additive model.

[0042] After standardization, the indicators may result in zero values. To ensure the data's reasonableness, the standardized data is shifted: .

[0043] The weight of the predicted electricity sales for the first and second time periods was calculated using the entropy method: ; And calculate the information entropy based on the stated proportion: ; in, These are constant coefficients.

[0044] Calculate the difference coefficient based on the information entropy: ; The final fusion weights are obtained by normalizing the difference coefficients: ; The future is obtained based on the weights. Forecasted electricity sales for the period: .

[0045] Example 3 The above is an introduction to the method embodiments. The following describes the solution of the present invention further through device embodiments.

[0046] A combined electricity sales forecasting device, the device comprising: Data acquisition module: Real-time collection and preprocessing of power technology parameters, related parameters, and environmental adaptation parameters; The first prediction module takes the preprocessed power technical parameters, related parameters and environmental adaptation parameters as input, uses the LSTM neural network to output the first time-segmented electricity sales prediction value, and introduces a power error penalty factor during the training process of the LSTM neural network to impose an additional penalty on the prediction error during peak and valley periods. The second prediction module uses the time-segmented electricity sales volume as the explained variable, the preprocessed correlation parameters as the core explanatory variables, and introduces the power grid transmission efficiency, transformer loss coefficient and extreme weather impact coefficient as power regulation factors to construct a generalized additive model and output the predicted value of the second time-segmented electricity sales volume. Output module: Calculates the fusion weight, merges the electricity sales forecast values ​​of the first and second time periods, and obtains the final electricity sales forecast result.

[0047] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0048] Example 4 The electronic device of this invention includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in read-only memory (ROM) or loaded from a storage unit into random access memory (RAM). The RAM may also store various programs and data required for device operation. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0049] Multiple components in the device are connected to the I / O interface, including: input units such as keyboards and mice; output units such as various types of displays and speakers; storage units such as disks and optical discs; and communication units such as network interface cards (NICs), modems, and wireless transceivers. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0050] The processing unit executes the various methods and processes described above, such as methods S1 to S4. For example, in some embodiments, methods S1 to S4 may be implemented as computer software programs tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via ROM and / or a communication unit. When the computer program is loaded into RAM and executed by the CPU, one or more steps of methods S1 to S4 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S4 by any other suitable means (e.g., by means of firmware).

[0051] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0052] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0053] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0054] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A combined method for predicting electricity sales, characterized in that, The method includes the following steps: Real-time acquisition and preprocessing of power technical parameters, related parameters, and environmental adaptation parameters; The preprocessed power technical parameters, correlation parameters, and environmental adaptation parameters are used as inputs. The first time-segmented electricity sales prediction value is output using an LSTM neural network. During the training process of the LSTM neural network, an electricity error penalty factor is introduced to impose an additional penalty on the prediction error during peak and valley periods. Using time-of-use electricity sales as the dependent variable, preprocessed correlation parameters as the core explanatory variables, and introducing grid transmission efficiency, transformer loss coefficient and extreme weather impact coefficient as power regulation factors, a generalized additive model is constructed to output the second time-of-use electricity sales forecast value. Calculate the fusion weight and fuse the electricity sales forecast values ​​of the first and second time periods to obtain the final electricity sales forecast result.

2. The combined electricity sales forecasting method according to claim 1, characterized in that, The aforementioned power technical parameters include total electricity sales within the region, time-of-use load, and voltage stability indicators.

3. The combined electricity sales forecasting method according to claim 1, characterized in that, The associated parameters include the installed capacity of industrial electrical equipment, the duration of electricity consumption in the tertiary sector, and the number of residential electricity terminals.

4. The combined electricity sales forecasting method according to claim 1, characterized in that, The environmental adaptation parameters include seasonality and extreme weather indicators.

5. The combined electricity sales forecasting method according to claim 1, characterized in that, The specific details of introducing an electricity error penalty factor during LSTM neural network training to impose additional penalties on peak and valley prediction errors are as follows: Introducing a power error penalty factor to construct the loss function: ； ； in, For loss function, To predict the total number of items in different time periods within the forecast period, express Power error penalty factor for the time period express Forecast error for a given period of time express Electricity consumption time period attributes, Indicates peak electricity consumption periods. Indicates off-peak electricity consumption periods. Indicates the time period of the level. express The relative deviation of electricity load during a given time period.

6. The combined electricity sales forecasting method according to claim 1, characterized in that, The generalized additive model is expressed as follows: ； in, To sell electricity in different time periods, Indicates the first A smooth function, Indicates the installed capacity of industrial electrical equipment. Indicates the duration of electricity consumption in the tertiary sector. Indicates the number of residential electricity terminals. Indicates the power grid transmission efficiency. This represents the transformer loss coefficient. The extreme weather impact coefficient is calculated based on seasonal coefficients and extreme weather indicators. This is the random error term.

7. The combined electricity sales forecasting method according to claim 1, characterized in that, The fusion weights are determined using the entropy method, specifically including the following steps: The electricity sales forecasts for the first and second time periods were standardized and shifted respectively. The weight of the predicted electricity sales for the first and second time periods after processing is calculated using the entropy method, and the information entropy is calculated based on the weight. The difference coefficient is calculated based on the information entropy, and the difference coefficient is normalized to obtain the final fusion weight.

8. A combined electricity sales forecasting device, characterized in that, The device includes: Data acquisition module: Real-time collection and preprocessing of power technology parameters, related parameters, and environmental adaptation parameters; The first prediction module takes the preprocessed power technical parameters, related parameters and environmental adaptation parameters as input, uses the LSTM neural network to output the first time-segmented electricity sales prediction value, and introduces a power error penalty factor during the training process of the LSTM neural network to impose an additional penalty on the prediction error during peak and valley periods. The second prediction module uses the time-segmented electricity sales volume as the explained variable, the preprocessed correlation parameters as the core explanatory variables, and introduces the power grid transmission efficiency, transformer loss coefficient and extreme weather impact coefficient as power regulation factors to construct a generalized additive model and output the predicted value of the second time-segmented electricity sales volume. Output module: Calculates the fusion weight, merges the electricity sales forecast values ​​of the first and second time periods, and obtains the final electricity sales forecast result.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 7.

Images