A cluster baseline load prediction method and device, electronic equipment and storage medium

Abstract

The application provides a cluster baseline load prediction method and device, electronic equipment and storage medium, wherein the method comprises: obtaining sample data for cluster baseline load prediction; using a K-Means algorithm to divide users into photovoltaic users and non-photovoltaic users according to weather state characteristics of the sample data; using a Shapley Value method to calculate a cluster baseline load prediction value of the non-photovoltaic users according to a first baseline load prediction value and a second baseline load prediction value of the non-photovoltaic users; and summing the cluster baseline load prediction value of the non-photovoltaic users and a cluster baseline load prediction value of photovoltaic users determined based on an SVR model to obtain a cluster baseline load prediction result. Through the application, the problem of low accuracy of cluster baseline load prediction for a cluster containing distributed photovoltaics in the related art is solved.

Description

Technical Field

[0001] This invention relates to the field of power demand response analysis, and more particularly to a cluster baseline load forecasting method, apparatus, electronic device, and storage medium. Background Technology

[0002] With the large-scale grid connection of new energy sources, the penetration rate of distributed photovoltaic (PV) power in distribution networks is gradually increasing. Distributed PV systems are generally installed after the electricity meter, and their output is not visible. After the installation of distributed PV, the intermittent PV output power and the unstable actual load are coupled together, increasing the difficulty of baseline load forecasting. Generally, measuring instruments can be installed on each distributed PV system to monitor its output in real time. However, due to the large number of distributed PV systems installed, this method is costly. Furthermore, the unobservability of post-meter PV, coupled with the inherent random and intermittent characteristics of PV, greatly increases the difficulty of estimating aggregated baseline load (ABL), because the uncertain PV output and the equally uncertain user load are coupled together, amplifying the uncertainty. Therefore, existing technologies suffer from low accuracy in forecasting aggregated baseline load that includes distributed PV. Summary of the Invention

[0003] This invention provides a cluster baseline load forecasting method, apparatus, electronic device, and storage medium to at least solve the problem of low accuracy in forecasting cluster baseline loads that include distributed photovoltaics in related technologies.

[0004] According to a first aspect of the present invention, a cluster baseline load forecasting method is provided. The method includes: acquiring sample data for cluster baseline load forecasting; using the K-Means algorithm to classify users into photovoltaic users and non-photovoltaic users based on weather state characteristics of the sample data; using the Shapley Value method to calculate the cluster baseline load forecast value for non-photovoltaic users based on a first baseline load forecast value and a second baseline load forecast value for non-photovoltaic users; and summing the cluster baseline load forecast value for non-photovoltaic users with the cluster baseline load forecast value for photovoltaic users determined based on the SVR model to obtain a cluster baseline load forecasting result.

[0005] Optionally, after obtaining the sample data for cluster baseline load prediction, the method further includes: correcting the load change based on the distribution characteristics of the load change in the sample data; and using the corrected sample data to correct the temperature corresponding to the sample data that meets the temperature correction conditions.

[0006] Optionally, the step of correcting the load change based on the distribution characteristics of the load change in the sample data includes: obtaining the load change of the date to be identified and multiple adjacent days of the date to be identified within a preset time period based on the sample data; calculating a detection value based on the standard deviation of the per-unit values ​​of the load change of the date to be identified and multiple adjacent days of the date to be identified within the preset time period, the average of the per-unit values ​​of the load change of the multiple adjacent days of the date to be identified within the preset time period, and the load change of the date to be identified within the preset time period; comparing the detection value with a threshold determined by a target confidence interval; if the detection value is greater than the threshold, replacing the load change of the date to be identified within the preset time period with the average of the per-unit values ​​of the load change of the multiple adjacent days of the date to be identified within the preset time period.

[0007] Optionally, the step of correcting the temperature corresponding to the sample data that meets the temperature correction conditions using the corrected sample data includes: obtaining the temperature and relative humidity of the day to be corrected and multiple adjacent days before the day to be corrected based on the corrected sample data; determining whether the day to be corrected meets the temperature correction conditions, wherein the temperature correction conditions are that the temperature is greater than a temperature threshold and the relative humidity is greater than a relative humidity threshold; if the temperature correction conditions are met, calculating the corrected temperature of the day to be corrected based on the temperature and humidity indices of the day to be corrected and multiple adjacent days before the day to be corrected, and the correction coefficients corresponding to the temperature and humidity indices of the day to be corrected and multiple adjacent days before the day to be corrected.

[0008] Optionally, if the temperature correction condition is met, calculating the corrected temperature of the day to be corrected based on the temperature and humidity indices of the day to be corrected and multiple adjacent days before the day to be corrected, and the correction coefficients corresponding to the temperature and humidity indices of the day to be corrected and multiple adjacent days before the day to be corrected, includes: determining the minimum value of the identification objective function based on the weighted sum of the temperature and humidity indices and corresponding correction coefficients of the day to be corrected and multiple adjacent days before the day to be corrected, and the reciprocal of the absolute value of the Pearson correlation coefficient of the load of the day to be corrected; using a genetic algorithm to calculate the correction coefficients based on the minimum value of the identification objective function and the correction constraints; and calculating the corrected temperature of the day to be corrected based on the weighted sum of the temperature and humidity indices of the day to be corrected and multiple adjacent days before the day to be corrected, and the correction coefficients.

[0009] Optionally, the step of using the Shapley Value method to calculate the cluster baseline load forecast value for non-photovoltaic users based on the first and second baseline load forecast values ​​includes: using a neural network to calculate the first baseline load forecast value for non-photovoltaic users based on historical cluster baseline load data; using a corresponding model based on the user's electricity consumption behavior to calculate the second baseline load forecast value for non-photovoltaic users; and using the Shapley Value method to calculate the cluster baseline load forecast value for non-photovoltaic users based on the first baseline load forecast value, the first marginal contribution rate, the first absolute error, the second baseline load forecast value, the second marginal contribution rate, and the second absolute error.

[0010] Optionally, the step of using the Shapley Value method to calculate the cluster baseline load forecast value for non-PV users based on the first baseline load forecast value, the first marginal contribution rate, the first absolute error, the second baseline load forecast value, the second marginal contribution rate, and the second absolute error includes: constructing a system of linear equations for the cluster baseline load forecast value for non-PV users, the first baseline load forecast value, the first marginal contribution rate, the second baseline load forecast value, and the second marginal contribution rate; constructing an objective function and constraints based on the first marginal contribution rate, the first absolute error, the second marginal contribution rate, and the second absolute error; calculating the first marginal contribution rate and the second marginal contribution rate using Lagrange multipliers based on the objective function and constraints; and determining the cluster baseline load forecast value for non-PV users based on the first marginal contribution rate and the second marginal contribution rate.

[0011] According to a second aspect of the present invention, a cluster baseline load forecasting apparatus is also provided. The apparatus includes: an acquisition module for acquiring sample data of cluster baseline load forecasting; a clustering module for using the K-Means algorithm to classify users into photovoltaic users and non-photovoltaic users based on the weather state characteristics of the sample data; a calculation module for using the Shapley Value method to calculate the cluster baseline load forecast value of non-photovoltaic users based on a first baseline load forecast value and a second baseline load forecast value of non-photovoltaic users; and a result module for summing the cluster baseline load forecast value of the non-photovoltaic users with the cluster baseline load forecast value of photovoltaic users determined based on the SVR model to obtain the cluster baseline load forecast result.

[0012] Optionally, the device further includes: a first correction module, configured to correct the load change based on the distribution characteristics of the load change in the sample data; and a second correction module, configured to correct the temperature corresponding to the sample data that meets the temperature correction conditions using the corrected sample data.

[0013] Optionally, the first correction module includes: a first acquisition unit, configured to acquire the load change amount of the date to be identified and multiple adjacent days of the date to be identified within a preset time period based on the sample data; a first calculation unit, configured to calculate a detection value based on the standard deviation of the per-unit values ​​of the load change amount of the date to be identified and multiple adjacent days of the date to be identified within the preset time period, the average value of the per-unit values ​​of the load change amount of the multiple adjacent days of the date to be identified within the preset time period, and the load change amount of the date to be identified within the preset time period; a comparison unit, configured to compare the detection value with a threshold determined by a target confidence interval; and a substitution unit, configured to substitute the load change amount of the date to be identified within the preset time period with the average value of the per-unit values ​​of the load change amount of the multiple adjacent days of the date to be identified within the preset time period if the detection value is greater than the threshold.

[0014] Optionally, the second correction module includes: a second acquisition unit, configured to acquire the temperature and relative humidity of the day to be corrected and multiple adjacent days before the day to be corrected based on the corrected sample data; a judgment unit, configured to determine whether the day to be corrected meets the temperature correction condition, wherein the temperature correction condition is that the temperature is greater than a temperature threshold and the relative humidity is greater than a relative humidity threshold; and a second calculation unit, configured to calculate the corrected temperature of the day to be corrected based on the temperature and humidity indices of the day to be corrected and multiple adjacent days before the day to be corrected, and the correction coefficients corresponding to the temperature and humidity indices of the day to be corrected and multiple adjacent days before the day to be corrected.

[0015] Optionally, the second calculation unit includes: a first determining submodule, configured to determine the minimum value of the identification objective function based on the weighted sum of the temperature and humidity indices and corresponding correction coefficients of the day to be corrected and multiple adjacent days before the day to be corrected, and the reciprocal of the absolute value of the Pearson correlation coefficient of the load on the day to be corrected; a first calculation submodule, configured to calculate the correction coefficients using a genetic algorithm based on the minimum value of the identification objective function and correction constraints; and a second calculation submodule, configured to calculate the corrected temperature of the day to be corrected based on the weighted sum of the temperature and humidity indices of the day to be corrected and multiple adjacent days before the day to be corrected, and the correction coefficients.

[0016] Optionally, the calculation module includes: a third calculation unit, used to calculate a first baseline load forecast value for non-photovoltaic users based on historical cluster baseline load data using a neural network; a fourth calculation unit, used to calculate a second baseline load forecast value for non-photovoltaic users using a corresponding model based on the user's electricity consumption behavior; and a fifth calculation unit, used to calculate the cluster baseline load forecast value for non-photovoltaic users based on the first baseline load forecast value, a first marginal contribution rate, a first absolute error, the second baseline load forecast value, a second marginal contribution rate, and a second absolute error using the Shapley Value method.

[0017] Optionally, the fifth calculation unit includes: a first construction submodule, used to construct a system of linear equations for the cluster baseline load forecast value, the first baseline load forecast value, the first marginal contribution rate, the second baseline load forecast value, and the second marginal contribution rate of non-photovoltaic users; a second construction submodule, used to construct an objective function and constraints based on the first marginal contribution rate, the first absolute error, the second marginal contribution rate, and the second absolute error; a third calculation submodule, used to calculate the first marginal contribution rate and the second marginal contribution rate using Lagrange multipliers based on the objective function and constraints; and a second determination submodule, used to determine the cluster baseline load forecast value of non-photovoltaic users based on the first marginal contribution rate and the second marginal contribution rate.

[0018] According to a third aspect of the present invention, an electronic device is also provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; wherein the memory is used to store a computer program; and the processor is used to execute the method steps of any of the above embodiments by running the computer program stored in the memory.

[0019] According to a fourth aspect of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored therein, wherein the computer program is configured to execute the method steps of any of the above embodiments when running.

[0020] In this embodiment of the invention, sample data for cluster baseline load forecasting is obtained; users are divided into photovoltaic (PV) users and non-PV users based on the weather state characteristics of the sample data using the K-Means algorithm; the Shapley Value method is used to calculate the cluster baseline load forecast value for non-PV users based on the first and second baseline load forecast values; the cluster baseline load forecast value for non-PV users is summed with the cluster baseline load forecast value for PV users determined based on the SVR model to obtain the cluster baseline load forecast result. Since PV users and non-PV users are decoupled based on the weather state characteristics of the sample data during the cluster baseline load forecasting process, two different models are used for the non-PV user cluster baseline load forecasting part. The Shapley Value method is used to calculate the marginal contribution rate of the single model forecast result to the combined model, obtaining the optimal combined forecast result, i.e., the cluster baseline load forecast value for non-PV users. The cluster baseline load forecast value for non-PV users is summed with the cluster baseline load forecast value for PV users calculated using the SVR model to obtain the final cluster baseline load forecast result. This achieves the effect of improving the accuracy of cluster baseline load forecasting and reducing forecast errors, solving the problem of low accuracy in cluster baseline load forecasting for distributed PV systems in related technologies.

[0021] In this embodiment of the invention, the load change is corrected based on the distribution characteristics of the load change in the sample data; the corrected sample data is then used to correct the temperature corresponding to the sample data that meets the temperature correction conditions. Because abnormal load data is identified and corrected using methods in mathematical statistics, and abnormal temperatures are corrected using the corrected load data, the accuracy of the sample data used for cluster baseline load prediction is improved, further achieving the goal of improving the accuracy of cluster baseline load prediction and reducing prediction errors. Attached Figure Description

[0022] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of the hardware environment for an optional cluster baseline load prediction method according to an embodiment of the present invention;

[0025] Figure 2 This is a flowchart illustrating an optional cluster baseline load prediction method according to an embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram of the overall process of an optional cluster baseline load prediction method according to an embodiment of the present invention;

[0027] Figure 4 This is a structural block diagram of an optional cluster baseline load forecasting device according to an embodiment of the present invention;

[0028] Figure 5 This is a structural block diagram of an optional electronic device according to an embodiment of the present invention. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0030] It should be noted that in the description of this invention, the terms "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0031] According to one aspect of the present invention, a cluster baseline load forecasting method is provided. Optionally, in this embodiment, the above-described cluster baseline load forecasting method can be applied to, for example... Figure 1 In the hardware environment shown. For example... Figure 1 As shown, terminal 102 may include memory 104, processor 106, and display 108 (optional component). Terminal 102 can communicate with server 112 via network 110. Server 112 can provide services (such as application services) to the terminal or clients installed on the terminal. Database 114 can be set up on or independently of server 112 to provide data storage services to server 112. In addition, server 112 may run a processing engine 116, which can be used to execute the steps performed by server 112.

[0032] Optionally, terminal 102 may be, but is not limited to, a terminal capable of computing data, such as a mobile terminal (e.g., mobile phone, tablet computer), laptop computer, PC (Personal Computer), etc. The aforementioned network may include, but is not limited to, a wireless network or a wired network. The wireless network includes Bluetooth, Wi-Fi (Wireless Fidelity), and other networks that enable wireless communication. The aforementioned wired network may include, but is not limited to, a wide area network (WAN), a metropolitan area network (MAN), and a local area network (LAN). The aforementioned server 112 may include, but is not limited to, any hardware device capable of computing.

[0033] Furthermore, in this embodiment, the aforementioned cluster baseline load forecasting method can also be applied to, but is not limited to, independent processing devices with powerful processing capabilities, without the need for data interaction. For example, the processing device can be, but is not limited to, a terminal device with powerful processing capabilities; that is, the various operations in the aforementioned cluster baseline load forecasting method can be integrated into a single independent processing device. The above is merely an example, and no limitation is made in this embodiment.

[0034] Optionally, in this embodiment, the above-described cluster baseline load prediction method can be executed by server 112, by terminal 102, or by both server 112 and terminal 102. Alternatively, the cluster baseline load prediction method of this embodiment can be executed by a client installed on terminal 102.

[0035] Taking the application of the cluster baseline load prediction method to the central processing unit as an example, Figure 2 This is a flowchart illustrating an optional cluster baseline load forecasting method according to an embodiment of the present invention, as shown below. Figure 2 As shown, the process of this method may include the following steps:

[0036] Step S201: Obtain sample data for cluster baseline load forecasting. Optionally, the sample data used for cluster baseline load forecasting includes load data and weather data from multiple consecutive sample days, wherein the weather data may include temperature, relative humidity, and temperature-humidity index, etc. In this embodiment of the invention, load data and weather data from 20 consecutive days are selected as sample data for cluster baseline load forecasting.

[0037] Step S202: The K-Means algorithm is used to classify users into photovoltaic (PV) users and non-PV users based on the weather conditions of the sample data. Optionally, since PV users have distinct characteristics and significant differences in PV output under different weather conditions, a PV output vector for each sample day is constructed based on the weather conditions. The K-Means algorithm is then used to cluster these PV output vectors, thereby achieving the goal of classifying users into PV users and non-PV users.

[0038] Step S203: The Shapley Value method is used to calculate the cluster baseline load forecast value for non-PV users based on the first and second baseline load forecast values. Optionally, two different models, such as a backpropagation neural network model and a load forecasting model established based on user electricity consumption behavior, are used to predict the baseline load of non-PV users, obtaining the first and second baseline load forecast values. Then, the Shapley Value method is used to optimize and combine the forecasting models to obtain the optimal cluster baseline load forecast value for non-PV users.

[0039] Step S204: Summing the cluster baseline load prediction value of non-PV users with the cluster baseline load prediction value of PV users determined based on the SVR model yields the cluster baseline load prediction result. Optionally, since PV users are significantly affected by weather, the SVR regression model is used to predict the cluster baseline load of PV users according to different weather types. SVR is a regression model mainly used for fitting numerical values. It is generally applied to scenarios with sparse features and a small number of features, and has the advantages of computational complexity not depending on the input space dimension, strong generalization ability, and high prediction accuracy. This embodiment of the invention uses three different weather types—sunny, cloudy, and rainy—for calculations. Since these three weather types cover most weather conditions, occur frequently, and have significant differences, corresponding SVR models are established for the three typical weather types of sunny, cloudy, and rainy days to calculate the cluster baseline load prediction value of PV users. The cluster baseline load prediction value of the PV users is then summed with the cluster baseline load prediction value of non-PV users calculated in step S203 to obtain the cluster baseline load prediction result.

[0040] In this embodiment of the invention, sample data for cluster baseline load forecasting is obtained; users are divided into photovoltaic (PV) users and non-PV users based on the weather state characteristics of the sample data using the K-Means algorithm; the Shapley Value method is used to calculate the cluster baseline load forecast value for non-PV users based on the first and second baseline load forecast values; the cluster baseline load forecast value for non-PV users is summed with the cluster baseline load forecast value for PV users determined based on the SVR model to obtain the cluster baseline load forecast result. Since PV users and non-PV users are decoupled based on the weather state characteristics of the sample data during the cluster baseline load forecasting process, two different models are used for the non-PV user cluster baseline load forecasting part. The Shapley Value method is used to calculate the marginal contribution rate of the single model forecast result to the combined model, obtaining the optimal combined forecast result, i.e., the cluster baseline load forecast value for non-PV users. The cluster baseline load forecast value for non-PV users is summed with the cluster baseline load forecast value for PV users calculated using the SVR model to obtain the final cluster baseline load forecast result. This achieves the effect of improving the accuracy of cluster baseline load forecasting and reducing forecast errors, solving the problem of low accuracy in cluster baseline load forecasting for distributed PV systems in related technologies.

[0041] As an optional embodiment, after obtaining sample data for cluster baseline load forecasting, the method further includes: correcting the load change based on the distribution characteristics of the load change in the sample data; and using the corrected sample data to correct the temperature corresponding to the sample data that meets the temperature correction conditions. Optionally, the power load, i.e., the load data, in the sample data has the following properties:

[0042] 1) Electricity load has both regularity and randomness. The overall distribution of its data conforms to a normal distribution. This property can be verified by skewness and kurtosis tests.

[0043] 2) The load between adjacent time periods is sticky, meaning that there will be no sudden changes. This is the basis for verifying abnormal load data.

[0044] 3) The load characteristics are the same or similar under the same or similar related factors (such as weather).

[0045] The nature of the power load determines the testing methods and effectiveness of abnormal load data. According to the definition of the t-distribution, when the sample size is ≥20, the t-distribution curve is very close to the normal curve. Considering the nature of the power load, the t-test can be used to identify and correct abnormal data in the power load. On the other hand, relative humidity and temperature influence each other. Relative humidity refers to the water vapor content in the atmosphere, expressed as the percentage of the maximum amount of water vapor that air can contain. Relative humidity changes with temperature and thus affects temperature. Therefore, corrected sample data is used to correct the temperature and the temperature of sample days with high relative humidity. In this embodiment of the invention, abnormal load data is identified and corrected using methods from mathematical statistics, and abnormal temperatures are corrected using the corrected load data. This improves the accuracy of the sample data used for cluster baseline load prediction, further achieving the goal of improving the accuracy of cluster baseline load prediction and reducing prediction errors.

[0046] As an optional embodiment, the load change is corrected based on the distribution characteristics of the load change in the sample data, including: obtaining the load change of the date to be identified and multiple adjacent days before the date to be identified within a preset time period based on the sample data; calculating a detection value based on the standard deviation of the per-unit values ​​of the load change of the date to be identified and multiple adjacent days before the date to be identified within the preset time period, the average of the per-unit values ​​of the load change of the multiple adjacent days before the date to be identified within the preset time period, and the load change of the date to be identified within the preset time period; comparing the detection value with a threshold determined by a target confidence interval; if the detection value is greater than the threshold, replacing the load change of the date to be identified within the preset time period with the average of the per-unit values ​​of the load change of the multiple adjacent days before the date to be identified within the preset time period.

[0047] Optionally, the load data for each sample day in the sample data can be segmented. Specifically, each segment is 15 minutes long, with the start time of the day being 0:00 and the end time being 23:45. A sample day may include 96 segments, corresponding to 96 load data points. Each segment / point of load data represents the load change within 15 minutes. The load change Δl d,t The calculation method is shown in formula (1):

[0048] Δld,t =l d,t -l d,t-1 (d=1,2,…,20; t=1,2,…,96) (1)

[0049] In the formula: d represents the sample day, t represents the time period, and Δl d,t Let l be the load change during time period t on the d-th sample day, in MW. d,t This represents the load value at the end of time period t on the d-th sample day, in MW.

[0050] To further increase the recognition sensitivity, Δl d,t Standardization is performed, transforming the data into dimensionless data. Specifically, the load change Δl with the largest absolute value among the 20 load changes within the same time period t (i.e., the preset time period) is used as the dividend, and each Δl... d,t As a divisor, the per-unit value of the load change is calculated according to formula (2):

[0051]

[0052] The day to be identified is the sample day on which the load data needs to be judged to be abnormal. Whether the load data on the day to be identified is abnormal can be determined by checking each of the 96 time periods of the day to be identified. Specifically, check whether there are any missing load data points of the 96 days to be identified. If there are any missing data points, replace them with a value of 0. Assume that the data before the day to be identified has been identified and corrected and is normal data. Select the data of the 19 days adjacent to the day to be identified as the identification reference data. The day to be identified is the 20th day. The identification reference data and the data of the day to be identified are the per-unit values ​​of the load change calculated according to formula (2).

[0053] Calculate the per-unit value Δl' of the load change during the t-period of the 19-day identification reference day according to formula (3). d,t average

[0054]

[0055] The average value is calculated using formula (3). The per-unit value Δl' of 20 load changes during time period t is calculated using formula (4). d,t Standard deviation S:

[0056]

[0057] The average value is calculated using formula (3). The standard deviation S obtained by formula (4) is used to calculate the load change Δl for the day t to be identified according to formula (5). 20,t The detection value k:

[0058]

[0059] Find k based on the target confidence interval, i.e., the 95% confidence interval probability. (95％,20) =2.16, thus obtaining the threshold k (p,n) Compare k (p,n) The magnitude of the detected value k, if k <k (p,n) Then Δl 20,t The value is normal data and does not need to be corrected; if k≥k (p,n) The load change Δl during time period t 20,t This is abnormal data and needs to be corrected. The average per-unit value is calculated using formula (3). Replace abnormal load change Δl 20,t , Δl 20,t-1 Keep the original value, Δl 20,t It can be calculated according to formula (1). After checking each of the 96 time periods of the day to be identified according to the above process, the abnormal load data in the sample data can be found and corrected. This achieves the purpose of correcting the abnormal load data, improves the reliability of the sample data, and thus improves the accuracy of cluster baseline load prediction.

[0060] As an optional embodiment, the temperature corresponding to the sample data that meets the temperature correction conditions is corrected using the corrected sample data, including: obtaining the temperature and relative humidity of the day to be corrected and several adjacent days before the day to be corrected based on the corrected sample data; determining whether the day to be corrected meets the temperature correction conditions, wherein the temperature correction conditions are that the temperature is greater than a temperature threshold and the relative humidity is greater than a relative humidity threshold; if the temperature correction conditions are met, calculating the corrected temperature of the day to be corrected based on the temperature and humidity indices of the day to be corrected and several adjacent days before the day to be corrected, and the correction coefficients corresponding to the temperature and humidity indices of the day to be corrected and several adjacent days before the day to be corrected.

[0061] Optionally, the temperature can be corrected based on the cumulative effect. Specifically, the corrected temperature T' for the day to be corrected can be calculated using the following formula:

[0062] T'=k0H0+k1H1+k2H2+k3H3 (6)

[0063] In the formula, H0 is the temperature and humidity index for that day, H i Let T be the temperature and humidity index on the i-th day before the correction date, and k0, k1, k2, and k3 be the corresponding correction coefficients. It should be noted that the day to be corrected must meet the temperature correction conditions, i.e., the temperature must be greater than the temperature threshold and the relative humidity must be greater than the relative humidity threshold. Temperature T, temperature and humidity index H, and relative humidity R can be obtained from the corrected sample data. In this embodiment, the temperature T and relative humidity R are taken as the average values ​​for the day, with a temperature threshold of 27°C and a relative humidity threshold of 40%, meaning the day to be corrected should satisfy T > 27°C and R > 40%.

[0064] As an optional embodiment, if the temperature correction condition is met, the corrected temperature of the day to be corrected is calculated based on the temperature and humidity indices of the day to be corrected and several adjacent days before the day to be corrected, as well as the correction coefficients corresponding to the temperature and humidity indices of the day to be corrected and several adjacent days before the day to be corrected. This includes: determining the minimum value of the objective function based on the weighted sum of the temperature and humidity indices and corresponding correction coefficients of the day to be corrected and several adjacent days before the day to be corrected, and the reciprocal of the absolute value of the Pearson correlation coefficient of the load of the day to be corrected; using a genetic algorithm to calculate the correction coefficients based on the minimum value of the objective function and the correction constraints; and calculating the corrected temperature of the day to be corrected based on the weighted sum of the temperature and humidity indices and correction coefficients of the day to be corrected and several adjacent days before the day to be corrected.

[0065] Optionally, the correction coefficients k0, k1, k2, and k3 are calculated using a genetic algorithm based on the minimum value of the objective function and the correction constraints, whereby the objective function and the correction constraints are shown in formulas (7) and (8), respectively:

[0066]

[0067]

[0068] In formula (7), r(T',P) represents the Pearson correlation coefficient between the corrected temperature T' and the load power, i.e., the load data of the day to be identified. The corrected temperature T' can be represented by the weighted sum of the temperature and humidity indices and corresponding correction coefficients of multiple adjacent days before the correction day and the day to be corrected in formula (6). To ensure that the temperature correction result is within a reasonable range and meets the principle of "nearer larger, farther smaller", that is, the temperature data of the day to be corrected depends more on the recent development pattern in the historical period and has a weaker correlation with the distant period, formula (8) is used as the correction constraint. For example, the single-objective minimum optimization problem with inequality constraints and equality constraints in formulas (7) and (8) can be solved using the genetic algorithm method in the prior art to obtain the correction coefficients k0, k1, k2, and k3. Then, the obtained correction coefficients are substituted into formula (6) to obtain the corrected temperature T' of the day to be corrected. In this embodiment, the corrected load data is used to correct the abnormal temperature, which improves the accuracy of the temperature data in the sample data used for cluster baseline load prediction, and further achieves the purpose of improving the accuracy of cluster baseline load prediction and reducing prediction error.

[0069] As an optional embodiment, the Shapley Value method is used to calculate the cluster baseline load forecast value for non-PV users based on the first and second baseline load forecast values. This includes: using a neural network to calculate the first baseline load forecast value for non-PV users based on historical cluster baseline load data; using a corresponding model based on the user's electricity consumption behavior to calculate the second baseline load forecast value for non-PV users; and using the Shapley Value method to calculate the cluster baseline load forecast value for non-PV users based on the first baseline load forecast value, the first marginal contribution rate, the first absolute error, the second baseline load forecast value, the second marginal contribution rate, and the second absolute error.

[0070] Optionally, since historical load data and weather data are known, the historical days that are most similar to the day on which cluster baseline load forecasting is required can be selected from the historical days. A set of similar days can be generated based on one or more historical days, and a cluster baseline load forecasting model can be established using a backpropagation neural network. Based on the set of similar days, the load data corresponding to the similar days, and the cluster baseline load forecasting model, the first baseline load forecast value for non-PV users can be obtained.

[0071] On the other hand, the load during periods of higher system load within a given time period (e.g., a day) is called peak load. Peak load is a key focus of current domestic demand response and is of great significance for the application and development of demand response under energy transition. This embodiment divides the load data of non-PV users into peak load and non-peak load. Peak load is further subdivided into three types: temperature-sensitive electricity consumption mode, holiday-sensitive electricity consumption mode, and electricity consumption mode that is insensitive to both holidays and temperature. Based on this, a cluster baseline load forecasting method adapted to each type is used. Finally, the cluster baseline load under each type is summed to obtain the second baseline load forecast value for non-PV users. Specifically, for the temperature-sensitive electricity consumption mode, after calculating the initial cluster baseline load, the initial cluster baseline load is adjusted using a linear regression method based on the actual load and temperature on the day of demand response. For the holiday-sensitive electricity consumption mode, after determining whether the day of demand response is a working day, the existing ABL estimation method is used to calculate the cluster baseline load. If the day of demand response is a working day, only working days are selected when choosing historical data; otherwise, if the day of demand response is a holiday, only historical data for holidays must be selected. For electricity usage patterns that are insensitive to holidays and temperature, the historical data calculation window is directly selected during the data selection phase, without considering whether the calculation window is a holiday or temperature adjustment calculation. That is, for this type of user, a simple cluster baseline load calculation using the existing ABL estimation method is sufficient. Similarly, for non-peak loads, the commonly used baseline load forecasting method, i.e., the existing ABL estimation method, is used.

[0072] The above steps utilize a neural network, specifically a backpropagation neural network, to establish a cluster baseline load forecasting model, yielding the first baseline load forecast for non-PV users. After segmenting users according to their electricity consumption, the corresponding model is used to calculate the second baseline load forecast for non-PV users. The Shapley Value method is employed to combine the two forecasting models. By calculating the marginal contribution rate of the individual model's forecast to the combined model, the optimal combined forecasting result is obtained. It is assumed that the two models are independent entities, with their forecasts (first baseline load forecast x1 and second baseline load forecast x2) independent of each other. The marginal contribution rates of the two models to the system are k1 and k2, respectively. The absolute errors between the predicted and actual values ​​are the first and second absolute errors, respectively. k1 and k2 are solved by minimizing the sum of absolute squares. Based on k1 and k2, the optimal result, the cluster baseline load forecast for non-PV users, is calculated. In this embodiment, the combined forecasting model, without altering the calculation method of the individual models, ensures optimal forecasting results while avoiding issues caused by missing data.

[0073] As an optional embodiment, the Shapley Value method is used to calculate the cluster baseline load forecast value for non-PV users based on a first baseline load forecast value, a first marginal contribution rate, a first absolute error, a second baseline load forecast value, a second marginal contribution rate, and a second absolute error. This includes: constructing a system of linear equations for the cluster baseline load forecast value for non-PV users, the first baseline load forecast value, the first marginal contribution rate, the second baseline load forecast value, and the second marginal contribution rate; constructing an objective function and constraints based on the first marginal contribution rate, the first absolute error, the second marginal contribution rate, and the second absolute error; calculating the first marginal contribution rate and the second marginal contribution rate using Lagrange multipliers based on the objective function and constraints; and determining the cluster baseline load forecast value for non-PV users based on the first marginal contribution rate and the second marginal contribution rate.

[0074] Optionally, the linear equations relating the first baseline load forecast x1, the second baseline load forecast x2, the first marginal contribution rate k1, the second marginal contribution rate k2, and the cluster baseline load forecast X of non-PV users can be expressed by formula (9):

[0075]

[0076] In the formula, x1 and x2 are 1×n dimensional vectors. The predicted value X of the i-th model... t Compared with the actual value Y t The absolute error e between it As shown in formula (10):

[0077]

[0078] The objective function is to minimize the absolute sum of squares error, where the absolute sum of squares error and the constraints are shown in equations (11) and (12):

[0079]

[0080] U T K = 1 (12)

[0081] To satisfy k1+k2=1 in formula (9), U in the formula T =[1 1], using the Lagrange multipliers to solve the objective function, we get:

[0082]

[0083] In the formula, K = [k1 k2] T The predicted value of the cluster baseline load for non-photovoltaic users can be obtained by using k1, k2 and formula (9).

[0084] As an optional embodiment, Figure 3 This is a schematic diagram of the overall process of an optional cluster baseline load forecasting method according to an embodiment of the present invention, as shown below. Figure 3 As shown, cluster user identification and decoupling are performed based on cluster user data. That is, users are classified according to weather characteristics using methods such as clustering or SVM to obtain photovoltaic users and non-photovoltaic users. The cluster user data is divided according to three weather types: sunny, cloudy, and rainy. The K-Means algorithm is used to cluster the data according to formula (14):

[0085]

[0086] In the formula, This represents the photovoltaic power output vector on day d under the k-th weather type, where k includes three types: sunny, cloudy, and rainy. D represents the mean vector of photovoltaic power output under the k-th weather type. k This represents the set of all days included in the k-th weather type.

[0087] To reduce the adverse impact of random fluctuations in residential load on the identification results, typical net load curves under different generalized weather types are extracted for each user, as shown in Equation (15):

[0088]

[0089] In the formula, |D k | indicates the number of days included in the k-th weather type; Represents the d∈D weather type under the k-th weather category. kNet load power during period t of day t.

[0090] Choose a time window δ = [t] for non-zero photovoltaic output. s ,t e ], t s and t e These represent the start and end times, respectively. Using the typical net load curve of this time window, features describing the differences in net load patterns between photovoltaic (PV) and non-PV users are extracted. The specific steps are as follows:

[0091] 1) The amplitude of the typical net load curve for photovoltaic users varies under different weather types, with this difference being particularly pronounced on sunny and rainy days. Therefore, the ratio of the absolute values ​​of the typical net load power under sunny and rainy weather types is taken as the first characteristic, denoted as F1:

[0092]

[0093] In the formula, the denominator and numerator represent the absolute values ​​of typical net load power on sunny and rainy days, respectively. For photovoltaic users, this characteristic value should be greater than 1, and for non-photovoltaic users, this characteristic value should be close to 1.

[0094] 2) After a user installs photovoltaics, their net load curve within the time window δ=[t s ,t e The interior will be concave downwards, a phenomenon that is more pronounced on sunny days because photovoltaic power generation is much greater on sunny days compared to other weather types. Based on this, a second feature, denoted as F2, is extracted from the typical net load curve under the generalized sunny weather type. After the time window is determined, there exists a straight line connecting the start and end points, represented by formula (17):

[0095]

[0096] It should be noted that A, B, and C represent sunny days, cloudy days, and rainy days, respectively. In the formula, y(t) represents the sampling points on the straight line in time period t. All sampling points form a set, denoted as S = {y(t) | t ∈ [t]}. s ,t e The set of sampling points located below the straight line is... It can be expressed by equation (18):

[0097]

[0098] Therefore, the second feature F2 is as shown in formula (19):

[0099]

[0100] In the formula, card(·) represents the number of elements in the set. This feature describes the proportion of sampling points along this line and can be used to reflect the convexity and concavity of the net load curve within the time window. The above two features F1 and F2 are extracted for all users, forming the feature vector F = [F1, F2] for each user. T The normalized feature vector is used as the input vector for clustering. The K-Means algorithm is then used to divide all users into two classes: the class with the larger cluster center feature value corresponds to photovoltaic users, and the rest are non-photovoltaic users.

[0101] For data related to non-PV users, the data is acquired using the first method, normalized, and all meteorological factors are weighted and sorted. The data for each weather category is then ranked based on similarity using meteorological and time factors. This data is then input into a BP neural network load forecasting model to obtain the forecast results. The specific process is as follows: Historical daily data and forecast daily data related to preset meteorological factors (such as temperature, humidity, air pressure, precipitation, irradiance, wind speed, etc.) and corresponding historical load data are acquired. Historical daily data includes meteorological factor data from previous years, and forecast daily data includes known or predicted meteorological factor data for the date of load forecasting. After data normalization, the historical daily data is represented in matrix form. A principal component analysis method based on entropy is used to determine the principal components in the meteorological factors, and the corresponding principal component comprehensive score is calculated using this method. Specifically, the principal component analysis method based on entropy uses the entropy value corresponding to each principal component to weight the principal component scores when calculating the principal component comprehensive score, thus improving the accuracy of the principal component weights. The meteorological similarity between the predicted day and the historical day is calculated based on the meteorological factors corresponding to the meteorological factors after being divided by the principal component comprehensive score. Based on the meteorological similarity and the temporal similarity between the predicted day and the historical day, a set of historical days similar to the predicted day is determined. Based on the obtained set of historical days, corresponding data are selected from the historical load data and historical day data, and a backpropagation (BP) neural network is trained. The backpropagation neural network can fit the relationship curve between meteorology and load through training. The first prediction result is obtained using the prediction day data.

[0102] The second method divides the load data of non-PV users into peak-season loads and non-peak-season loads. Peak-season loads are further subdivided into three types: temperature-sensitive electricity consumption patterns, holiday-sensitive electricity consumption patterns, and electricity consumption patterns insensitive to both holidays and temperature. Based on this, the cluster baseline load forecasting method corresponding to each electricity consumption pattern is used. The cluster baseline load results under various electricity consumption patterns and non-peak-season loads are summed to obtain the second baseline load forecast value for non-PV users. The Shapley value method is used to calculate the marginal contribution rates k1 and k2 (corresponding to the two methods) for the first and second baseline load forecast values. The combined forecast results yield the ABL for non-PV users.

[0103] For photovoltaic (PV) user-related data, based on K-Means weather type classification, the data is divided into non-response day data and response day data. Response refers to electricity demand response, which means that when wholesale electricity prices rise or system reliability is threatened, electricity users, upon receiving a direct compensation notice from the power supplier inducing load reduction or a signal of rising electricity prices, change their inherent habitual electricity consumption patterns to reduce or postpone electricity load during a certain period, thereby ensuring grid stability and suppressing short-term behavior related to rising electricity prices. Three prediction models are established for PV users under typical weather types A (sunny), B (cloudy), and C (rainy), namely Prediction Model A, Prediction Model B, and Prediction Model C. The SVR estimation model is used here. Each prediction model is trained on non-response day data, and the cluster baseline load prediction effect is tested on response day data. The trained models are used to calculate the ABL (Advanced Load Balance) for PV users under each typical weather type, and the sum of this ABL and the ABL for non-PV users yields the user ABL.

[0104] According to another aspect of the present invention, a cluster baseline load prediction apparatus is also provided. Figure 4 This is a structural block diagram of an optional cluster baseline load forecasting device according to an embodiment of the present invention, such as... Figure 4 As shown, the device may include: an acquisition module 401 for acquiring sample data of cluster baseline load forecast; a clustering module 402 for using the K-Means algorithm to classify users into photovoltaic users and non-photovoltaic users based on the weather state characteristics of the sample data; a calculation module 403 for using the Shapley Value method to calculate the cluster baseline load forecast value of non-photovoltaic users based on the first and second baseline load forecast values ​​of non-photovoltaic users; and a result acquisition module 404 for summing the cluster baseline load forecast value of non-photovoltaic users with the cluster baseline load forecast value of photovoltaic users determined based on the SVR model to obtain the cluster baseline load forecast result.

[0105] It should be noted that the acquisition module 401 in this embodiment can be used to execute the above step S201, the clustering module 402 in this embodiment can be used to execute the above step S202, the calculation module 403 in this embodiment can be used to execute the above step S203, and the obtaining module 404 in this embodiment can be used to execute the above step S204.

[0106] Through the above modules, the cluster baseline load forecasting process decouples photovoltaic (PV) users from non-PV users based on the weather condition characteristics of the sample data. The non-PV user cluster baseline load forecasting portion uses two different models. The Shapley Value method is used to calculate the marginal contribution rate of the single model's forecast result to the combined model, obtaining the optimal combined forecast result, which is the non-PV user cluster baseline load forecast value. The non-PV user cluster baseline load forecast value is then summed with the PV user cluster baseline load forecast value calculated using the SVR model to obtain the final cluster baseline load forecast result. This achieves the effect of improving the accuracy of cluster baseline load forecasting and reducing forecast errors, solving the problem of low accuracy in cluster baseline load forecasting for distributed PV systems in related technologies.

[0107] As an optional embodiment, the device further includes: a first correction module, used to correct the load change based on the distribution characteristics of the load change in the sample data; and a second correction module, used to correct the temperature corresponding to the sample data that meets the temperature correction conditions using the corrected sample data.

[0108] As an optional embodiment, the first correction module includes: a first acquisition unit, configured to acquire, based on sample data, the load change amount within a preset time period for the date to be identified and multiple adjacent days prior to the date to be identified; a first calculation unit, configured to calculate a detection value based on the standard deviation of the per-unit values ​​of the load change amount within the preset time period for the date to be identified and multiple adjacent days prior to the date to be identified, the average of the per-unit values ​​of the load change amount within the preset time period for multiple adjacent days prior to the date to be identified, and the load change amount within the preset time period for the date to be identified; a comparison unit, configured to compare the detection value with a threshold determined by a target confidence interval; and a substitution unit, configured to, if the detection value is greater than the threshold, substitute the load change amount within the preset time period for the date to be identified using the average of the per-unit values ​​of the load change amount within the preset time period for multiple adjacent days prior to the date to be identified.

[0109] As an optional embodiment, the second correction module includes: a second acquisition unit, used to acquire the temperature and relative humidity of the day to be corrected and multiple adjacent days before the day to be corrected based on the corrected sample data; a judgment unit, used to determine whether the day to be corrected meets the temperature correction conditions, wherein the temperature correction conditions are that the temperature is greater than a temperature threshold and the relative humidity is greater than a relative humidity threshold; and a second calculation unit, used to calculate the corrected temperature of the day to be corrected based on the temperature and humidity indices of the day to be corrected and multiple adjacent days before the day to be corrected, and the correction coefficients corresponding to the temperature and humidity indices of the day to be corrected and multiple adjacent days before the day to be corrected.

[0110] As an optional embodiment, the second calculation unit includes: a first determining submodule, used to determine the minimum value of the identification objective function based on the weighted sum of the temperature and humidity indices and corresponding correction coefficients of the day to be corrected and multiple adjacent days before the day to be corrected, and the reciprocal of the absolute value of the Pearson correlation coefficient of the load on the day to be corrected; a first calculation submodule, used to calculate the correction coefficients based on the minimum value of the identification objective function and correction constraints using a genetic algorithm; and a second calculation submodule, used to calculate the corrected temperature of the day to be corrected based on the weighted sum of the temperature and humidity indices and correction coefficients of the day to be corrected and multiple adjacent days before the day to be corrected.

[0111] As an optional embodiment, the calculation module includes: a third calculation unit for calculating a first baseline load forecast value for non-PV users using a neural network based on historical cluster baseline load data; a fourth calculation unit for calculating a second baseline load forecast value for non-PV users using a corresponding model based on the user's electricity consumption behavior; and a fifth calculation unit for calculating the cluster baseline load forecast value for non-PV users using the Shapley Value method based on the first baseline load forecast value, a first marginal contribution rate, a first absolute error, the second baseline load forecast value, the second marginal contribution rate, and the second absolute error.

[0112] As an optional embodiment, the fifth calculation unit includes: a first construction submodule, used to construct a system of linear equations for the cluster baseline load forecast value, the first baseline load forecast value, the first marginal contribution rate, the second baseline load forecast value, and the second marginal contribution rate of non-photovoltaic users; a second construction submodule, used to construct an objective function and constraints based on the first marginal contribution rate, the first absolute error, the second marginal contribution rate, and the second absolute error; a third calculation submodule, used to calculate the first marginal contribution rate and the second marginal contribution rate using Lagrange multipliers based on the objective function and constraints; and a second determination submodule, used to determine the cluster baseline load forecast value of non-photovoltaic users based on the first marginal contribution rate and the second marginal contribution rate.

[0113] It should be noted that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should also be noted that the above modules, as part of a device, can operate in environments such as... Figure 1 The hardware environment shown can be implemented through software or hardware, and the hardware environment includes the network environment.

[0114] According to another aspect of the present invention, an electronic device for implementing the above-described cluster baseline load prediction method is also provided. The electronic device may be a server, a terminal, or a combination thereof.

[0115] Figure 5 This is a structural block diagram of an optional electronic device according to an embodiment of the present invention, such as... Figure 5 As shown, the system includes a processor 501, a communication interface 502, a memory 503, and a communication bus 504. The processor 501, communication interface 502, and memory 503 communicate with each other via the communication bus 504. The memory 503 stores computer programs. When the processor 501 executes the computer program stored in the memory 503, it performs the following steps:

[0116] Obtain sample data for cluster baseline load forecasting; use the K-Means algorithm to classify users into photovoltaic users and non-photovoltaic users based on the weather state characteristics of the sample data; use the Shapley Value method to calculate the cluster baseline load forecast value for non-photovoltaic users based on the first and second baseline load forecast values; sum the cluster baseline load forecast value for non-photovoltaic users with the cluster baseline load forecast value for photovoltaic users determined based on the SVR model to obtain the cluster baseline load forecast result.

[0117] Optionally, in this embodiment, the communication bus can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. This communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0118] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0119] The memory may include RAM, or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0120] As an example, such as Figure 5 As shown, the memory 503 may include, but is not limited to, the acquisition module 401, clustering module 402, calculation module 403, and obtaining module 404 from the cluster baseline load prediction device. Furthermore, it may include, but is not limited to, other module units from the cluster baseline load prediction device, which will not be elaborated upon in this example.

[0121] The processors mentioned above can be general-purpose processors, including but not limited to: CPU (Central Processing Unit), NP (Network Processor), etc.; they can also be DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0122] In addition, the aforementioned electronic equipment also includes a display for showing the cluster baseline load forecast results.

[0123] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments, and will not be repeated here.

[0124] Those skilled in the art will understand that Figure 5 The structure shown is for illustrative purposes only. The device implementing the above cluster baseline load prediction method can be a terminal device, such as a smartphone (e.g., Android phone, iOS phone), tablet computer, PDA, mobile Internet Devices (MID), PAD, etc. Figure 5 This does not limit the structure of the aforementioned electronic devices. For example, the terminal device may also include components that are more... Figure 5 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 5 The different configurations shown.

[0125] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, ROM, RAM, disk or optical disk, etc.

[0126] According to another aspect of the present invention, a storage medium is also provided. Optionally, in this embodiment, the storage medium can be used to execute program code for a cluster baseline load prediction method.

[0127] Optionally, in this embodiment, the storage medium may be located on at least one of the network devices in the network shown in the above embodiment.

[0128] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps:

[0129] Obtain sample data for cluster baseline load forecasting; use the K-Means algorithm to classify users into photovoltaic users and non-photovoltaic users based on the weather state characteristics of the sample data; use the Shapley Value method to calculate the cluster baseline load forecast value for non-photovoltaic users based on the first and second baseline load forecast values; sum the cluster baseline load forecast value for non-photovoltaic users with the cluster baseline load forecast value for photovoltaic users determined based on the SVR model to obtain the cluster baseline load forecast result.

[0130] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments, and will not be repeated in this embodiment.

[0131] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, ROMs, RAMs, portable hard drives, magnetic disks, or optical disks.

[0132] According to another aspect of the present invention, a computer program product or computer program is also provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to perform the cluster baseline load forecasting method steps in any of the above embodiments.

[0133] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0134] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause one or more computer devices (which may be personal computers, servers, or network devices, etc.) to execute all or part of the steps of the cluster baseline load prediction method of the various embodiments of the present invention.

[0135] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0136] In the several embodiments provided by this invention, it should be understood that the disclosed client can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0137] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the solution provided in this embodiment, depending on actual needs.

[0138] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0139] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A cluster baseline load forecasting method, characterized in that, The method includes: Obtain sample data for cluster baseline load forecasting; The K-Means algorithm is used to classify users into photovoltaic users and non-photovoltaic users based on the weather conditions of the sample data; The Shapley Value method is used to calculate the cluster baseline load forecast for non-PV users based on the first and second baseline load forecasts for non-PV users. The cluster baseline load prediction result is obtained by summing the cluster baseline load prediction value of the non-photovoltaic users with the cluster baseline load prediction value of the photovoltaic users determined based on the SVR model. After obtaining sample data for cluster baseline load prediction, the method further includes: The load change is corrected based on the distribution characteristics of the load change in the sample data; The temperature corresponding to the sample data that meets the temperature correction conditions is corrected using the corrected sample data. The step of correcting the temperature of sample data that meets the temperature correction conditions using the corrected sample data includes: Based on the corrected sample data, obtain the temperature and relative humidity of the date to be corrected and several adjacent days before the date to be corrected; Determine whether the date to be corrected meets the temperature correction condition, wherein the temperature correction condition is that the temperature is greater than a temperature threshold and the relative humidity is greater than a relative humidity threshold. If the temperature correction condition is met, the corrected temperature of the day to be corrected is calculated based on the temperature and humidity index of the day to be corrected and several adjacent days before the day to be corrected, as well as the correction coefficients corresponding to the temperature and humidity index of the day to be corrected and several adjacent days before the day to be corrected. If the temperature correction condition is met, the corrected temperature for the day to be corrected is calculated based on the temperature and humidity indices of the day to be corrected and several adjacent days prior to the day to be corrected, as well as the correction coefficients corresponding to the temperature and humidity indices of the day to be corrected and several adjacent days prior to the day to be corrected. This includes: The minimum value of the identification objective function is determined by the weighted sum of the temperature and humidity indices and corresponding correction coefficients of the date to be corrected and multiple adjacent days before the date to be corrected, and the reciprocal of the absolute value of the Pearson correlation coefficient of the load on the date to be corrected. A genetic algorithm is used to calculate the correction coefficients based on the minimum value of the identification objective function and the correction constraints. The corrected temperature for the date to be corrected is calculated by weighting the temperature and humidity indices of the date to be corrected and several adjacent days before the date to be corrected, as well as the correction coefficient.

2. The cluster baseline load forecasting method according to claim 1, characterized in that, The correction of the load change based on the distribution characteristics of the load change in the sample data includes: Based on the sample data, obtain the load change amount of the day to be identified and multiple adjacent days before the day to be identified within a preset time period; The detection value is calculated based on the standard deviation of the per-unit values ​​of the load change on the date to be identified and multiple adjacent days before the date to be identified within a preset time period, the average value of the per-unit values ​​of the load change on multiple adjacent days before the date to be identified within a preset time period, and the load change on the date to be identified within a preset time period; The detected value is compared with a threshold determined by the target confidence interval; If the detected value is greater than the threshold, the average per-unit value of the load change during the preset time period on the day to be identified is used to replace the load change during the preset time period on the day to be identified.

3. The cluster baseline load prediction method according to claim 1, characterized in that, The calculation of the cluster baseline load forecast value for non-PV users using the ShapleyValue method based on the first and second baseline load forecast values ​​includes: The neural network is used to calculate the first baseline load forecast for non-PV users based on historical cluster baseline load data. The second baseline load forecast value for non-photovoltaic users is calculated using the corresponding model based on the user's electricity consumption behavior. The Shapley Value method is used to calculate the cluster baseline load forecast for non-PV users based on the first baseline load forecast, the first marginal contribution rate, the first absolute error, the second baseline load forecast, the second marginal contribution rate, and the second absolute error.

4. The cluster baseline load forecasting method according to claim 1, characterized in that, The calculation of the cluster baseline load forecast value for non-PV users using the ShapleyValue method based on the first baseline load forecast value, the first marginal contribution rate, the first absolute error, the second baseline load forecast value, the second marginal contribution rate, and the second absolute error includes: Construct a system of linear equations for the cluster baseline load forecast, the first baseline load forecast, the first marginal contribution rate, the second baseline load forecast, and the second marginal contribution rate for non-PV users; Construct the objective function and constraints based on the first marginal contribution rate, the first absolute error, the second marginal contribution rate, and the second absolute error; The first marginal contribution rate and the second marginal contribution rate are calculated using Lagrange multipliers based on the objective function and constraints. The cluster baseline load forecast value for non-PV users is determined based on the first marginal contribution rate and the second marginal contribution rate.

5. A cluster baseline load forecasting device, characterized in that, The device includes: The acquisition module is used to acquire sample data for cluster baseline load forecasting; The clustering module is used to divide users into photovoltaic users and non-photovoltaic users based on the weather status characteristics of the sample data using the K-Means algorithm. The calculation module is used to calculate the cluster baseline load forecast for non-PV users based on the first and second baseline load forecasts using the Shapley Value method. The module is used to sum the cluster baseline load prediction value of the non-photovoltaic user cluster with the cluster baseline load prediction value of the photovoltaic user cluster determined based on the SVR model to obtain the cluster baseline load prediction result; The device further includes: The first correction module is used to correct the load change based on the distribution characteristics of the load change in the sample data; The second correction module is used to correct the temperature of the sample data that meets the temperature correction conditions using the corrected sample data. The first correction module includes: The second acquisition unit is used to acquire the temperature and relative humidity of the date to be corrected and multiple adjacent days before the date to be corrected, based on the corrected sample data. The judgment unit is used to determine whether the day to be corrected meets the temperature correction condition, wherein the temperature correction condition is that the temperature is greater than a temperature threshold and the relative humidity is greater than a relative humidity threshold. The second calculation unit is used to calculate the corrected temperature of the day to be corrected based on the temperature and humidity index of the day to be corrected and a plurality of adjacent days before the day to be corrected, and the correction coefficient corresponding to the temperature and humidity index of the day to be corrected and a plurality of adjacent days before the day to be corrected, if the temperature correction condition is met. The second computing unit includes: The first determining submodule is used to determine the minimum value of the identification objective function based on the weighted sum of the temperature and humidity indices and corresponding correction coefficients of the day to be corrected and multiple adjacent days before the day to be corrected, as well as the reciprocal of the absolute value of the Pearson correlation coefficient of the load of the day to be corrected. The first calculation submodule is used to calculate the correction coefficients using a genetic algorithm based on the minimum value of the identification objective function and the correction constraints. The second calculation submodule is used to calculate the corrected temperature of the day to be corrected based on the weighted sum of the temperature and humidity indices of the day to be corrected and multiple adjacent days before the day to be corrected, as well as the correction coefficient.

6. An electronic device comprising a processor, a communication interface, a memory, and a communication bus, wherein, The processor, the communication interface, and the memory communicate with each other via the communication bus, characterized in that... The memory is used to store computer programs; The processor is configured to perform the method steps of any one of claims 1 to 4 by running the computer program stored in the memory.

7. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of the method described in any one of claims 1 to 4.

Images