Load prediction method and device, electronic equipment and storage medium

Abstract

One or more embodiments of the invention provide a load prediction method and device, electronic equipment and a storage medium. The method comprises the steps that to-be-predicted time period data is matched with historical time period data, multiple pieces of target historical time period data with the highest similarity are screened out, and the historical time period data are divided according to power utilization characteristics; according to the similarity between the target historical time period data and the to-be-predicted time period data, distributing a weight for each piece of target historical time period data; and obtaining predicted load data of the to-be-predicted time period data according to the load corresponding to the target historical time period data, and evaluating contribution values of features in the target historical time period data to the predicted load. Through the technical scheme of the invention, the precision and credibility of short-term load prediction can be improved, and a reliable basis can be provided for power dispatching and other scenes.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and in particular to a load forecasting method, apparatus, electronic device, and storage medium. Background Technology

[0002] It should be noted that the above description of the technical background is only for the purpose of providing a clear and complete explanation of the technical solutions of the present invention and facilitating understanding by those skilled in the art. It should not be assumed that these technical solutions are known to those skilled in the art simply because they have been described in the background section of this invention.

[0003] The current power system is facing a trend of high-proportion renewable energy integration and diversified electricity demand, which makes intraday load exhibit high-frequency fluctuations. Against this backdrop, traditional load forecasting methods encounter significant challenges in dealing with such high-frequency fluctuations, especially when processing the 96-point intraday load curve, and are unable to meet the needs of power dispatching, demand-side management, and renewable energy consumption scenarios for short-term, refined forecasting.

[0004] Existing load forecasting technologies suffer from two main limitations: first, statistical methods based on time series models struggle to effectively capture the nonlinear characteristics of loads; second, machine learning models, while capable of improving forecast accuracy to some extent, are sensitive to the quality of historical data and lack generalization ability. More critically, existing methods generally ignore the temporal similarities in load changes, leading to a significant increase in forecast errors during periods of electricity consumption mode transition (such as the morning-evening peak transition period). Summary of the Invention

[0005] In view of the above, the purpose of one or more embodiments of this disclosure is to provide a load forecasting method, apparatus, electronic device and storage medium to solve the problems raised in the background art.

[0006] To achieve the above objectives, one or more embodiments of this disclosure provide a load forecasting method, including:

[0007] The data for the period to be predicted is matched with historical period data, and multiple target historical period data with the highest similarity are selected. The historical period data is divided according to electricity consumption characteristics.

[0008] Based on the similarity between the target historical time period data and the time period data to be predicted, a weight is assigned to each target historical time period data;

[0009] Based on the load corresponding to the target historical time period data, the predicted load data of the time period to be predicted is obtained, and the contribution value of the features in the target historical time period data to the predicted load is evaluated.

[0010] Optionally, the data for the period to be predicted is matched with historical period data, and multiple target historical period data with the highest similarity are selected, including:

[0011] Based on the time period type and date attribute of the time period data to be predicted, the historical time period data is filtered to obtain filtered historical time period data;

[0012] Calculate the dynamic time-normalized distance between the filtered historical time period data and the data to be predicted, whereby the dynamic time-normalized distance represents the similarity between the filtered historical time period data and the data to be predicted;

[0013] Select a preset number of target historical time periods from the filtered historical time period data that have the largest dynamic time regularization distance.

[0014] Optionally, calculating the dynamic time-normalized distance between the filtered historical time period data and the data to be predicted includes:

[0015] Extract the first feature data from the filtered historical time period data and the second feature data from the time period data to be predicted;

[0016] Calculate the dynamic time warp distance between the first feature data and the second feature data for each feature dimension;

[0017] Based on the weight coefficient of each feature dimension, the dynamic time warping distance between the first feature data and the second feature data is weighted and summed to obtain the dynamic time warping distance between the filtered historical time period data and the data to be predicted. The feature dimensions are obtained through a grid search optimization algorithm.

[0018] Optionally, the predicted load of the time period to be predicted is obtained based on the load data corresponding to the target historical time period data, including:

[0019] The weight of the load data corresponding to the target historical time period data in the fusion calculation is determined based on the dynamic time warping distance between the target historical time period data and the data to be predicted.

[0020] Based on the weight of the load data in the fusion calculation and the load data itself, the predicted load of the data for the period to be predicted is obtained through weighted fusion.

[0021] Optionally, based on the weight of the load data in the fusion calculation and the load data itself, a weighted fusion is performed to obtain the predicted load of the time period data to be predicted, including:

[0022] Obtain adjacent load data for each adjacent time period of the load data;

[0023] The load data is smoothed between segments based on the adjacent load data.

[0024] Based on the weights of the load data in the fusion calculation and the load data after inter-segment smoothing, the predicted load of the time period to be predicted is obtained through weighted fusion.

[0025] Optionally, evaluating the contribution of features in the target historical time period data to the predicted load includes:

[0026] Obtain the characteristics of the target historical time period data;

[0027] Determine the SHAP value for each of the said features in relation to the predicted load, which reflects the contribution of the feature to the predicted load.

[0028] Optionally, it also includes:

[0029] Based on the time period to be predicted, determine the corresponding time period rules;

[0030] Based on the data for the period to be predicted and the period rules, a human judgment is made as to whether to adjust the medical workload.

[0031] The time period rules are generated using CART decision trees or rule fitting algorithms based on the characteristics of the historical time period data and the load data corresponding to the historical time period data.

[0032] Based on the same inventive concept, one or more embodiments of this disclosure also provide a load forecasting device, including:

[0033] The matching module is configured to match the data of the time period to be predicted with historical time period data, and filter out multiple target historical time period data with the highest similarity, wherein the historical time period data is divided according to electricity consumption characteristics;

[0034] The similarity calculation module is configured to assign weights to each of the target historical time periods based on the similarity between the target historical time period data and the time period data to be predicted;

[0035] The prediction module is configured to obtain the predicted load data of the time period to be predicted based on the load corresponding to the target historical time period data, and to evaluate the contribution value of the features in the target historical time period data to the predicted load.

[0036] Based on the same inventive concept, one or more embodiments of this disclosure also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the load forecasting method as described in any of the foregoing.

[0037] Based on the same inventive concept, one or more embodiments of this disclosure also provide a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform any of the load forecasting methods described above.

[0038] As can be seen from the above, the load forecasting method provided in one or more embodiments of this disclosure matches the data of the period to be predicted with historical period data divided according to electricity consumption characteristics, filters out multiple target historical period data with the highest similarity, assigns weights to them according to the similarity, and combines the load corresponding to the target historical period data to obtain the predicted load data of the period to be predicted. It also evaluates the contribution value of features in the target historical period data to the predicted load, thereby achieving accurate load forecasting and interpretability analysis of the forecast results. The technical solution of this disclosure can improve the accuracy and reliability of short-term load forecasting and provide a reliable basis for scenarios such as power dispatching.

[0039] The load forecasting apparatus, electronic device, and computer-readable storage medium disclosed herein can all implement the steps of the above-described load forecasting method, and therefore also possess the beneficial effects of the above-described load forecasting method. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in one or more embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only one or more embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart illustrating one or more embodiments of the load forecasting method disclosed herein;

[0042] Figure 2 This is a schematic diagram of the structure of a load forecasting device according to one or more embodiments of the present disclosure;

[0043] Figure 3 This is a schematic diagram of the hardware structure of an electronic device according to one or more embodiments of this disclosure. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0045] It should be noted that, unless otherwise defined, the technical or scientific terms used in one or more embodiments of this disclosure should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar words used in one or more embodiments of this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0046] As described in the background section, existing technologies have many shortcomings in the operational application of 96-point load forecasting: although machine learning and deep learning methods can automatically extract features, the forecast results lack interpretability and it is difficult to trace key influencing factors, resulting in insufficient trust in the forecast results by dispatchers; related technologies treat the 96 time periods as independent time points, ignoring the temporal correlation of load changes (such as the similarity between weekday morning peak and historical electricity consumption patterns), resulting in poor forecasting performance in scenarios with strong electricity consumption regularity; existing similar day selection methods are mostly based on matching the load curve throughout the day, which cannot accurately capture the similar characteristics of local time periods within the day (such as the load response pattern during the afternoon photovoltaic output drop period).

[0047] Therefore, for reference Figure 1 The load forecasting method of one or more embodiments disclosed herein includes the following steps:

[0048] Step S101: Match the data of the period to be predicted with the historical period data, and filter out the target historical period data with the highest similarity. The above historical period data is divided according to electricity consumption characteristics.

[0049] Step S102: Assign weights to each of the above target historical time periods based on the similarity between the above target historical time period data and the above predicted time period data;

[0050] Step S103: Based on the load corresponding to the target historical time period data, obtain the predicted load data of the time period to be predicted, and evaluate the contribution of the features in the target historical time period data to the predicted load.

[0051] In some implementations, both the data for the period to be predicted and the historical data for the period are obtained through monitoring systems in the target area.

[0052] In some embodiments, data acquisition can be targeted at a specific area and collected through power monitoring systems, meteorological monitoring stations, and time information databases, where historical time period data may include power load, meteorological, and time-related data from the past two years.

[0053] The aforementioned monitoring system can acquire data according to a preset frequency or preset collection time. For example, power load data can be sampled at 15-minute intervals, recording 96 load data points daily. Meteorological data can be collected simultaneously for the corresponding time period, including various meteorological indicators such as temperature, humidity, wind speed, precipitation, and sunshine duration, forming a meteorological data vector.

[0054] In some implementations, the data for the period to be predicted is preprocessed before being matched with historical data. For example, this preprocessing may include handling data anomalies.

[0055] In some embodiments, data anomaly handling may include: calculating the mean and standard deviation of the load data at each time point, and identifying data points exceeding or falling outside the mean by ±3 standard deviations as outliers. For outliers, deletion or replacement with the median may be selected. For missing values, cubic spline interpolation may be used to fill in the missing values. For example, if data for a 15-minute interval is missing, spline interpolation can be performed using data from adjacent time points.

[0056] In some implementations, the process of matching the data for the forecast period with historical data may include: quickly filtering irrelevant candidate periods based on period type (peak / off-peak / valley) and date attribute (weekday / weekend / holiday); calculating the DTW distance for the remaining candidate set, sorting them by similarity, and selecting the top k similar periods as the basis for load forecasting.

[0057] In some embodiments, the specific steps for calculating the DTW distance may be as follows: based on historical time period data, the daily data to be predicted is divided into corresponding time periods, the DTW distance is calculated using a dynamic time warping algorithm, and time period clusters with similar electricity consumption characteristics are identified (such as "weekday morning peak hours" and "morning peak hours with similar historical temperatures and weather").

[0058] In some embodiments, the formula for calculating the DTW distance can be:

[0059]

[0060] Where X = (x1,...,x m ) and Y = (y1,...,y n ) represent the characteristics of historical data and data to be predicted, respectively, and π represents the regular path, satisfying the constraints of boundary, monotonicity and continuity.

[0061] In some implementations, multidimensional features are introduced to calculate the DTW distance in order to improve the accuracy of the calculation results.

[0062] In some embodiments, the multidimensional features may include one or more of the following: temperature, electricity price, photovoltaic output, and holiday markers.

[0063] In some embodiments, the above features can be calculated based on the power load characteristics to calculate statistical features such as average load, maximum load, minimum load, and standard deviation within time windows such as the past hour, past day, and past week, and to extract the load change trends of the daily cycle (24 hours) and weekly cycle (7 days), and to extract periodic features through Fourier transform.

[0064] In some embodiments, for meteorological features, in addition to the meteorological indicators at the current moment, meteorological lag values ​​are introduced, such as the temperature and humidity of the previous day, to capture the delayed impact of meteorological conditions on power load. Combined features of meteorological indicators are calculated, including perceived temperature (considering temperature, humidity, and wind speed), comfort index, and cumulative temperature. For time features, the year is mapped to a relative value (e.g., calculating the difference starting from a base year), and months and days of the week are represented using trigonometric functions to reflect their periodicity. Holiday identifiers are set to 0 (non-holiday) or k (number of holiday days), forming a time feature vector. Then, the top m features with high correlation coefficients are selected based on the correlation coefficients, and the features are normalized.

[0065] In some implementations, the specific algorithms for calculating DTW distance by introducing multi-dimensional features include: multi-feature weighted dynamic time warping distance model, the formula of which can be expressed as:

[0066]

[0067] Where L, T, and P represent characteristic sequences, such as load, temperature, and electricity price sequences, respectively. S(t1,t2) represents the DTW distance between the historical data and the data to be predicted, and S(t1,t2) represents the DTW distance between the features of the historical data and the data to be predicted. α, β, and γ represent the weighting coefficients, and α+β+γ=1 (e.g., α=0.6, β=0.2, γ=0.2). These values ​​can be obtained through a grid search optimization algorithm.

[0068] To ensure the interpretability of the prediction results, some implementations obtain the prediction results through dynamic weighted fusion of loads.

[0069] In some embodiments, the dynamic weighted fusion process may include: for each segmented time period, the top k similar time periods are selected, and weights are calculated based on their multidimensional feature DTW distance. The weight calculation formula may be:

[0070]

[0071] Where t0 represents the time period to be predicted, S(t0,t i ) represents the time period to be predicted and the historical time period t. i The multidimensional feature DTW distance. Predicted value calculation formula: In the formula, The actual load values ​​for similar historical periods are used to achieve dynamic fitting of similar patterns through weighted fusion.

[0072] In some implementations, to avoid abrupt changes in predicted values ​​between adjacent time periods, a first-order difference smoothing constraint can be introduced:

[0073]

[0074] in, and This represents the predicted values ​​of adjacent time periods to be predicted, and a smooth transition between time periods can be achieved through mean filtering.

[0075] In some implementations, based on the results of screening similar time periods, it is possible to trace back to the similar time periods corresponding to each time period, intuitively observe the characteristics and power load curves of similar time periods, identify the differences in curves and characteristics between similar time periods and actual loads, and have strong interpretability.

[0076] In some implementations, the contribution of features in historical time period data to the predicted load can be evaluated by using the SHAP value of the features in the historical time period data.

[0077] In some embodiments, time-based SHAP value feature attribution focuses on breaking down the prediction process into each specific time period. Through the quantitative calculation of SHAP values, the contribution of input features such as temperature, humidity, precipitation, and preceding time period load to the prediction results in different time periods can be accurately analyzed.

[0078] Compared to holistic attribution analysis, this approach can capture the temporal differences in the impact of features more subtly—for example, the weight of preceding loads during the morning rush hour is significantly higher than at night, and the impact of extreme precipitation on load during commuting hours is drastically different from that during inactive periods. At critical junctures where loads experience abrupt changes (such as sudden increases or decreases), the dominant factors driving these changes can be identified by the sign and magnitude of the SHAP values ​​for each feature, providing a micro-level attribution basis for interpreting the temporal patterns of load fluctuations and for precise regulation.

[0079] In some implementations, the method also includes generating explicit judgment logic (e.g., "if the temperature is >30℃ and the electricity price is during peak hours, then the load increase is ≥5%) through CART decision trees or rule fitting algorithms, for dispatchers to manually verify or intervene in the adjustment.

[0080] It is understandable that this method can be executed by any device, equipment, platform, or cluster of devices with computing and processing capabilities.

[0081] It should be noted that the methods of one or more embodiments of this disclosure can be executed by a single device, such as a computer or server. The methods of this embodiment can also be applied in a distributed scenario, where multiple devices cooperate to complete the process. In such a distributed scenario, one of these devices may execute only one or more steps of the methods of one or more embodiments of this disclosure, and the multiple devices will interact with each other to complete the method described.

[0082] It should be noted that the above description pertains to specific embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than those shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0083] Based on the same inventive concept, corresponding to any of the methods in the above embodiments, this disclosure also provides a load forecasting device. For example... Figure 2 As shown, the above-mentioned device includes:

[0084] Matching module 11 is configured to match the data of the time period to be predicted with historical time period data, and filter out multiple target historical time period data with the highest similarity, wherein the historical time period data is divided according to electricity consumption characteristics;

[0085] The similarity calculation module 12 is configured to assign weights to each target historical time period data based on the similarity between the target historical time period data and the time period data to be predicted;

[0086] The prediction module 13 is configured to obtain the predicted load data of the time period to be predicted based on the load corresponding to the target historical time period data, and to evaluate the contribution value of the features in the target historical time period data to the predicted load.

[0087] For ease of description, the above apparatus is described in terms of its functions, divided into various modules. Of course, when implementing one or more embodiments of this disclosure, the functions of each module can be implemented in one or more software and / or hardware.

[0088] The apparatus described above is used to implement the corresponding methods in the foregoing embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0089] Figure 3This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0090] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this disclosure.

[0091] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this disclosure are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0092] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0093] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0094] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0095] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this disclosure, and not necessarily all the components shown in the figures.

[0096] The electronic devices described above are used to implement the corresponding methods in the foregoing embodiments and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0097] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0098] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure (including the claims) is limited to these examples; within the framework of this disclosure, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of this disclosure as described above, which are not provided in detail for the sake of brevity.

[0099] Additionally, to simplify the description and discussion, and to avoid obscuring one or more embodiments of this disclosure, the provided drawings may or may not show well-known power / ground connections to integrated circuit (IC) chips and other components. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring one or more embodiments of this disclosure, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which one or more embodiments of this disclosure will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuitry) have been set forth to describe exemplary embodiments of this disclosure, it will be apparent to those skilled in the art that one or more embodiments of this disclosure may be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0100] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0101] This disclosure includes one or more embodiments intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A load forecasting method, characterized in that, include: The data for the period to be predicted is matched with historical period data, and multiple target historical period data with the highest similarity are selected. The historical period data is divided according to electricity consumption characteristics. Based on the similarity between the target historical time period data and the time period data to be predicted, a weight is assigned to each target historical time period data; Based on the load corresponding to the target historical time period data, the predicted load data of the time period to be predicted is obtained, and the contribution value of the features in the target historical time period data to the predicted load is evaluated.

2. The method according to claim 1, characterized in that, The data for the period to be predicted is matched with historical data for different periods, and the historical data for the target periods with the highest similarity are selected, including: Based on the time period type and date attribute of the time period data to be predicted, the historical time period data is filtered to obtain filtered historical time period data; Calculate the dynamic time-normalized distance between the filtered historical time period data and the data to be predicted, whereby the dynamic time-normalized distance represents the similarity between the filtered historical time period data and the data to be predicted; Select a preset number of target historical time periods from the filtered historical time period data that have the largest dynamic time regularization distance.

3. The method according to claim 2, characterized in that, Calculating the dynamic time-normalized distance between the filtered historical time period data and the data to be predicted includes: Extract the first feature data from the filtered historical time period data and the second feature data from the time period data to be predicted; Calculate the dynamic time warp distance between the first feature data and the second feature data for each feature dimension; Based on the weight coefficient of each feature dimension, the dynamic time regularization distance between the first feature data and the second feature data is weighted and summed to obtain the dynamic time regularization distance between the filtered historical time period data and the time period data to be predicted. The feature dimensions are obtained through a grid search optimization algorithm.

4. The method according to claim 1, characterized in that, Based on the load data corresponding to the target historical time period data, the predicted load of the time period to be predicted is obtained, including: The weight of the load data corresponding to the target historical time period data in the fusion calculation is determined based on the dynamic time warping distance between the target historical time period data and the data to be predicted. Based on the weight of the load data in the fusion calculation and the load data itself, the predicted load of the data for the period to be predicted is obtained through weighted fusion.

5. The method according to claim 4, characterized in that, Based on the weights of the load data in the fusion calculation and the load data itself, the predicted load of the time period to be predicted is obtained through weighted fusion, including: Obtain adjacent load data for each adjacent time period of the load data; The load data is smoothed between segments based on the adjacent load data. Based on the weights of the load data in the fusion calculation and the load data after inter-segment smoothing, the predicted load of the time period to be predicted is obtained through weighted fusion.

6. The method according to claim 1, characterized in that, Evaluating the contribution of features in the target historical time period data to the predicted load includes: Obtain the characteristics of the target historical time period data; Determine the SHAP value for each of the said features in relation to the predicted load, which reflects the contribution of the feature to the predicted load.

7. The method according to claim 1, characterized in that, Also includes: Based on the time period to be predicted, determine the corresponding time period rules; Based on the data for the period to be predicted and the period rules, a human judgment is made as to whether to adjust the medical workload. The time period rules are generated using CART decision trees or rule fitting algorithms based on the characteristics of the historical time period data and the load data corresponding to the historical time period data.

8. A load forecasting device, characterized in that, include: The matching module is configured to match the data of the time period to be predicted with historical time period data, and filter out multiple target historical time period data with the highest similarity, wherein the historical time period data is divided according to electricity consumption characteristics; The similarity calculation module is configured to assign weights to each of the target historical time periods based on the similarity between the target historical time period data and the time period data to be predicted; The prediction module is configured to obtain the predicted load data of the time period to be predicted based on the load corresponding to the target historical time period data, and to evaluate the contribution value of the features in the target historical time period data to the predicted load.

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

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing the computer to perform the method of any one of claims 1 to 7.

Images