A method and system for extreme high temperature weather load forecasting based on tensor completion

Abstract

The present invention discloses a method and system for extreme high temperature weather load prediction based on tensor completion, including: calculating a comprehensive high temperature index based on a heat index and temperature considering air humidity constraints, screening load data of historical extreme high temperature weather; completing the screened load data based on a tensor completion algorithm to obtain completed load data; calculating the correlation between the completed load data and factors related to extreme high temperature weather load, screening input feature quantities for load prediction; using the input feature quantities, constructing an LSTM-RST short-term load prediction model based on a long short-term memory network and rough set theory RST, and predicting extreme high temperature weather load. The present invention can effectively improve the accuracy of load prediction under extreme high temperature weather.

Description

Technical Field

[0001] The present invention belongs to the technical field of power load forecasting, and relates to a method and system for forecasting load in extreme high temperature weather based on tensor completion. Background Art

[0002] With the continuous advancement of the construction of new power systems in my country, the scale of wind and solar power installations has continued to expand. However, new energy power generation forms such as wind and solar are greatly affected by weather factors such as sunshine, wind speed, and temperature. Therefore, while promoting the construction of new power systems, there is also a large demand for flexible response resources such as energy storage. In this scenario, accurate power load forecasting under extreme weather scenarios will contribute to the safe and stable operation of the power grid and the planning and design of the power system.

[0003] At present, domestic and foreign scholars have conducted extensive research on short-term load forecasting and proposed various methods to improve the prediction accuracy. For example, based on the gray-gray system model, the first, second and piecewise regression models between the daily cooling electricity load and temperature in summer are established respectively, and the daily cooling load in summer is predicted based on this model; based on the threshold autoregressive moving average model considering the influence of temperature, the threshold autoregressive moving average model with temperature as the covariate is established to improve the accuracy of load prediction; based on the grid power grid load prediction algorithm of support vector machine, the particle swarm algorithm is used to optimize the penalty function, projection distance, kernel function and other parameters of the improved support vector machine model to improve the accuracy of the prediction algorithm; based on the feature correlation, the weighted dimensional processing method of the influencing factors of power load. The above load prediction research is limited to objects such as power grid load, microgrid load, and electric vehicle charging load under normal conditions, and no research has been conducted on load prediction in unconventional scenarios such as special events and extreme weather. However, existing research on load forecasting under extreme weather conditions often ignores the problem of small amount of load forecasting data and high randomness under extreme high temperature scenarios, fails to deal with possible missing data, and also lacks a specific definition and screening method for extreme weather, resulting in low accuracy of load forecasting under extreme weather conditions. Summary of the invention

[0004] In order to solve the deficiencies in the prior art, the present invention provides a method and system for extreme high temperature weather load prediction based on tensor completion, which can realize load prediction in extreme weather scenarios and improve the safe operation level of a high-proportion new energy power grid in extreme high temperature weather.

[0005] The present invention adopts the following technical solution. A first aspect of the present invention provides an extreme high temperature weather load prediction method based on tensor completion, comprising:

[0006] Step 1: Calculate the comprehensive high temperature index based on the heat index and temperature considering the air humidity constraint, and filter the load data of historical extreme high temperature weather;

[0007] Step 2: Based on the tensor completion algorithm, the load data obtained by screening in step 1 is completed to obtain the completed load data;

[0008] Step 3: Calculate the correlation between the completed load data and the factors related to the extreme high temperature weather load, and select the input characteristic quantities for load forecasting;

[0009] Step 4: Using the input feature quantity, construct an LSTM-RST short-term load forecasting model based on long short-term memory network and rough set theory RST;

[0010] Step 5: Use the LSTM-RST short-term load forecasting model to predict the extreme high temperature weather load.

[0011] Preferably, in step 1, the heat index considering the air humidity constraint is:

[0012]

[0013] Where: H t Represents the air humidity at time period t;

[0014] W t represents the temperature at time period t;

[0015] E t represents the heat index in period t;

[0016] γ represents the weather persistence coefficient;

[0017] t∈T, T represents the total number of time periods in a day.

[0018] Preferably, in step 1, the calculation formula of the comprehensive high temperature index is:

[0019]

[0020] Where: R d Represents the comprehensive high temperature index on the dth date;

[0021] W t represents the temperature at time period t;

[0022] E t represents the heat index in period t;

[0023] κ w and κ e Represent the weights of temperature and heat index respectively;

[0024] T represents the total number of time periods in a day.

[0025] Preferably, in step 2, based on the tensor completion algorithm, the function for completing the load data screened in step 1 is:

[0026]

[0027] Where: || || * represents the nuclear norm;

[0028] α i represents the weighting coefficient;

[0029] To solve the tensor, that is, the data to be completed;

[0030] is the original missing tensor, i.e., the load data obtained by screening in step 1;

[0031] Ω represents the set of observable data in the tensor;

[0032] Among them, l1, l2, and l3 represent the data type order, data date day order, and data daily sampling number respectively. The data type order includes load, temperature, and heat index;

[0033] j represents the rank of matrix X;

[0034] σ k (X) represents the kth singular value of the matrix X in descending order of size;

[0035] X (i) Specifically:

[0036]

[0037] Where: X (1) , X (2) , X (3) are the modular expansion matrices of the matrix X sampled alternately by data type order, data date day order, and data daily sampling order.

[0038] Preferably, in step 3, the Pearson correlation coefficient between the completed load data and the factors related to the extreme high temperature weather load is calculated, and the six factors with the largest correlation coefficients are selected as input feature quantities for load prediction.

[0039] Preferably, the extreme high temperature weather load-related factors include maximum temperature, heat index, average temperature, average humidity, month, day, hour, minute, working day and time-of-use electricity price.

[0040] Preferably, in step 4, the input feature quantity and the load data completed in step 2 are substituted into LSTM for training, and the load forecast value obtained after training is corrected by RST algorithm, and the correction result is evaluated. When the requirements are met, the model construction is completed, otherwise the LSTM or RST algorithm parameters are adjusted and retrained.

[0041] Preferably, the mean absolute percentage error and the maximum absolute error are used to evaluate the correction results, respectively, wherein the mean absolute percentage error measures the overall error, and the maximum absolute error index measures the local maximum prediction deviation.

[0042] Preferably, when the average percentage error does not exceed 8% and the maximum absolute error does not exceed 10%, the requirements are met and the model construction is completed;

[0043] When the average percentage error is greater than 8%, adjust the LSTM parameters and retrain;

[0044] When the maximum absolute error is greater than 10%, adjust the RST parameters and perform corrections.

[0045] A second aspect of the present invention provides an extreme high temperature weather load forecasting system based on tensor completion, comprising:

[0046] A data screening module is used to calculate a comprehensive high temperature index based on the heat index and temperature considering air humidity constraints, and to screen the load data of historical extreme high temperature weather;

[0047] A data completion module is used to complete the filtered load data based on a tensor completion algorithm to obtain completed load data;

[0048] The feature screening module is used to calculate the correlation between the completed load data and the factors related to the extreme high temperature weather load, and screen the input feature quantities for load forecasting;

[0049] A model building module, used to build a LSTM-RST short-term load forecasting model based on a long short-term memory network and a rough set theory RST by using the input feature quantity;

[0050] The load forecasting module is used to predict the load in extreme high temperature weather using the LSTM-RST short-term load forecasting model.

[0051] A terminal comprises a processor and a storage medium; the storage medium is used to store instructions;

[0052] The processor is configured to operate according to the instructions to execute the steps of the method.

[0053] A computer-readable storage medium stores a computer program, which implements the steps of the method when executed by a processor.

[0054] Compared with the prior art, the beneficial effects of the present invention include at least:

[0055] In view of the small amount of load prediction data and strong randomness in extreme high temperature scenarios, the present invention first screens extreme high temperature scenario data based on the two indicators of improved heat index and temperature, and defines the extreme high temperature scenario of extreme weather in a formula, then completes the load data based on tensors, and then screens the input feature quantity through correlation analysis to build an LSTM-RST short-term load prediction model. The prediction results of a certain link of the present invention using different methods and the prediction results of the method of the present invention are evaluated and analyzed, and it can be seen that the present invention can effectively improve the accuracy of load prediction in extreme high temperature weather and improve the research on load prediction in extreme weather scenarios.

[0056] The improved heat index of the present invention takes air humidity as a constraint condition and takes into account the impact of weather continuity on human comfort, making up for the defect that the value is low during hot and dry weather in summer. Combined with the temperature index, it can adaptively screen out extreme high temperature weather, improve the accuracy of load forecasting under extreme high temperature weather, and can be applied to other extreme weather through adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a flow chart of the load forecasting method of the present invention;

[0058] Figure 2 This is a schematic diagram of the load forecasting method of the present invention;

[0059] Figure 3 This is an untreated extreme high temperature daily load curve diagram in an embodiment of the present invention;

[0060] Figure 4 This is an extreme high temperature load curve diagram after data completion in an embodiment of the present invention;

[0061] Figure 5 This is a graph of correlation coefficients of LSTM input feature quantities in an embodiment of the present invention;

[0062] Figure 6 It is a load curve diagram of the simulation prediction results of the example in the embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only embodiments of a part of the present invention, rather than all embodiments. Based on the spirit of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the protection scope of the present invention.

[0064] like Figure 1-2 As shown, embodiment 1 of the present invention provides a method for extreme high temperature weather load prediction based on tensor completion, comprising the following steps:

[0065] Step 1: Calculate the comprehensive high temperature index based on the heat index and temperature considering the air humidity constraint, and filter the load data of historical extreme high temperature weather;

[0066] Further preferably, an evaluation model is established based on the improved heat index and temperature to screen extreme high temperature scene data;

[0067] For hot weather, the most intuitive measurement indicator is temperature. Considering that the way electricity is used is affected to a certain extent by human factors and thus affects the load, the present invention introduces a heat index for comprehensive judgment. The heat index is an evaluation index of human comfort, taking into account influencing factors such as temperature and humidity.

[0068] The heat index of the present invention takes air humidity as a constraint condition and takes into account the impact of weather persistence on human comfort, making up for the defect of a lower value in hot and dry weather in summer, and is expressed as follows:

[0069]

[0070] Where: H t Represents the air humidity at time period t;

[0071] W t Represents the temperature in the tth period / ℃;

[0072] E t represents the heat index in period t;

[0073] γ represents the weather continuous influence coefficient, which is taken as 0.02 in the present invention;

[0074] t∈T, T represents the total number of time periods in a day. The present invention takes 24 for calculation, and each time period represents one hour. The supplementary item The hot weather conditions of the day before the current period also have an impact on the current period.

[0075] In summary, the present invention establishes a comprehensive evaluation system based on the two indicators of temperature and heat index.

[0076]

[0077] Where: κ w and κ e Represent the weights of temperature and heat index respectively, and their values ​​are all greater than 0 and the sum is 1;

[0078] R d Represents the comprehensive high temperature index on the dth date.

[0079] The above extreme high temperature scenario formulas (1) and (2) take into account the impact of weather persistence on human comfort, make up for the defect of low value in hot and dry weather in summer, and comprehensively consider the impact of temperature. Through formula (2), extreme high temperature weather within a certain time range can be adaptively screened out according to a certain proportion, thereby improving the accuracy of load forecasting under extreme high temperature weather.

[0080] In this embodiment, a total of 306 days of data from June 1, 2021 to October 31, 2021, and from June 1, 2022 to October 31, 2022 in a certain area of ​​Suzhou are selected for analysis, and load forecasting is performed on the high temperature date of July 20, 2023. Using formula (2) as the evaluation index, the 30-day data with the highest comprehensive high temperature index are selected from the 306-day data according to the comprehensive high temperature index value as the research data of the present invention. The 30-day load curve is as follows: Figure 3 shown.

[0081] from Figure 3 It can be seen that there are some missing data due to meter measurement or other reasons, which are shown as curve interruptions in the graph, and two places have been marked in the figure. In order to improve the accuracy of load forecasting, it is necessary to complete the missing data.

[0082] Step 2: Based on the tensor completion algorithm, the load data obtained by screening in step 1 is completed to obtain the completed load data;

[0083] Further preferably, a tensor-based load data completion model is proposed, which is as follows:

[0084] The present invention constructs a load forecasting completion data third-order tensor Among them, l1, l2, and l3 represent the data type order, data date day order, and data daily sampling number respectively, that is, the data type order includes the specific values ​​of the three items of load, temperature, and heat index. The tensor is called "order" here;.

[0085] Defining Tensors The rank of The optimization function of tensor completion is expressed as follows:

[0086]

[0087] Where: and The solution tensor and the original missing tensor, respectively.

[0088] Ω represents the set of observable data in the tensor.

[0089] The objective function is non-convex and discrete, so it needs to be convexly relaxed to transform it into a convex optimization problem. The minimum nuclear norm is used instead of the minimum rank. The tensor nuclear norm is defined as the sum of the nuclear norms of all matrices along the module expansion of the tensor, expressed as follows:

[0090]

[0091] Where: || || * represents the nuclear norm;

[0092] α i represents the weighting coefficient, which is a positive number and is generally equal (i.e., 1 / 3 in this case).

[0093] Load forecasting completion data tensor The three mode matrices of are expanded as follows:

[0094]

[0095] Where: X (1) , X (2) , X (3) It represents the modular expansion matrix of the matrix X by staggered sampling according to the data type order, the data date day order and the data daily sampling order.

[0096] Therefore, formula (3) can be transformed as follows:

[0097]

[0098] Where: j represents the rank of matrix X;

[0099] To solve the tensor, the tensor obtained by solving formulas (6)-(7) is the data used for completion;

[0100] is the original missing tensor, i.e., the load data obtained by screening in step 1;

[0101] σ k Represents the kth singular value of the matrix X in descending order of size.

[0102] In this embodiment, the third-order tensor is selected as the data type order (load, temperature, heat index), the data date day order (30 days) and the data daily sampling order (96 points sampling) for data completion. The completed load curve is as follows Figure 4 shown.

[0103] Step 3: Calculate the correlation between the completed load data and the factors related to the extreme high temperature weather load, and select the input characteristic quantities for load forecasting;

[0104] Further preferably, the input feature quantity is screened by Pearson correlation analysis, as follows.

[0105] The present invention uses the Pearson correlation coefficient to calculate the correlation between the load output power of the existing data set and various meteorological factors. The Pearson correlation coefficient is calculated as shown in the following formula:

[0106]

[0107] Where: λ represents the correlation coefficient between the completed load data vector A and the extreme high temperature weather load related factor vector B; A is fixed, and B is the largest six factors obtained by taking turns among various factors;

[0108] cov(A,B) represents the covariance of vectors A and B;

[0109] Represent the mean square error of vectors A and B respectively.

[0110] After calculation by the above formula, the top six factors are taken as the input features of the LSTM network for extreme high temperature day load prediction.

[0111] In this embodiment, a feature set affecting load forecasting is constructed based on meteorological factors, date factors, etc., as shown in Table 1.

[0112] Table 1 Characteristic quantities affecting load forecasting

[0113] category Feature quantity Meteorological factors Maximum temperature, heat index, average temperature, average humidity Date Factor Month, day, hour, minute, weekday Other factors Time-of-use electricity price

[0114] The present invention uses the Pearson correlation coefficient to calculate the correlation between the load output power of the existing data set and various meteorological factors. The results are shown in Figure 5 .

[0115] Depend on Figure 5 It can be seen that the factors with a high correlation with the load on extreme high temperature days are the maximum temperature, heat index, average temperature, average humidity, hours, and working days. The correlation of other factors is relatively low, so they are regarded as irrelevant meteorological factors. Therefore, the maximum temperature, heat index, average temperature, average humidity, hours, and working days are used as the input features of the LSTM network for extreme high temperature day load prediction.

[0116] Step 4: Using the input feature quantity, construct an LSTM-RST short-term load forecasting model based on long short-term memory network and rough set theory RST;

[0117] Preferably, the input features and historical data (load data after completion in step 2) are substituted into LSTM for training. After training, the prediction results are obtained and the RST algorithm is corrected. The correction results are evaluated by the mean absolute percentage error and the maximum absolute error. When both the mean absolute percentage error and the maximum absolute error meet the requirements, the model construction is completed. Otherwise, the LSTM or RST algorithm parameters are adjusted and retrained.

[0118] The RST algorithm can effectively correct inconsistent defective information, information that requires error correction, or information that has data loss, as follows:

[0119]

[0120] Where: y t+1 and t ' +1 They represent the predicted value and the revised value at time t+1 respectively, where the load prediction value is the load prediction value output by LSTM, and the revised value is the load prediction value output by the LSTM-RST short-term load prediction model, that is, one is not revised and the other is revised;

[0121] y t ,y t+2 are the load forecast values ​​at time t and t+2 respectively;

[0122] y t ′ is the load forecast value at time t;

[0123] k t+1 , k t , k t-1 They represent the differences on both sides of the prediction function at the corresponding moments;

[0124] s t Represents the scale factor.

[0125]

[0126] Among them, q is the adjustment coefficient;

[0127] So far, the load forecast value data can be corrected through equations (9)-(10). Further preferably, the forecast result evaluation is performed as follows:

[0128] It is observed that the evaluation criteria for load forecasting generally use indicators such as root mean square error, mean absolute error, and mean absolute percentage error, all of which are used to measure the overall error level and lack indicators for evaluating local errors.

[0129] After comprehensive consideration, the present invention selects the mean absolute percentage error to measure the overall error, and selects the maximum absolute error index to measure the local maximum prediction deviation. The calculation formula of the index is expressed as follows:

[0130]

[0131] Where: τ MAE , τ MPE They represent the mean absolute percentage error and the maximum absolute percentage error respectively;

[0132] y i 'Represents the load forecast value output by the model during period t;

[0133] y t represents the actual load value during period t;

[0134] m represents the number of forecast periods.

[0135] When the average percentage error does not exceed 8% and the maximum absolute error does not exceed 10%, the requirements are met and the model construction is completed;

[0136] When the average percentage error is greater than 8%, adjust the LSTM parameters and retrain;

[0137] When the maximum absolute error is greater than 10%, adjust the RST parameter q and correct it again.

[0138] Step 5: Use the LSTM-RST short-term load forecasting model to predict the extreme high temperature weather load.

[0139] In order to verify the effectiveness of the method of the present invention, basic examples and comparative examples are set, as shown in Table 2. The difference between comparative examples 1-3 and the basic examples is that one step in the basic example is missing, and the difference between comparative examples 4-5 and the basic example is that a different method is used in a certain link of the model, while the rest of the settings are the same. The prediction results of each example are shown in Figure 6 and Table 3.

[0140] Table 2 Example settings

[0141]

[0142]

[0143] Table 3 Simulation results evaluation

[0144] category MAE / % MPE / % Basic Example 3.75％ 6.15％ Comparative Example 1 4.25％ 7.67％ Comparative Example 2 3.99％ 6.67％ Comparative Example 3 4.50％ 10.58％ Comparative Example 4 4.66％ 9.79％ Comparative Example 5 5.56％ 11.96％

[0145] analyze Figure 6 From Table 3, we can draw the following conclusions:

[0146] The prediction accuracy of the basic example of the present invention is significantly better than that of other examples, with a maximum absolute error of 6.15% and an average percentage error of 3.75%. In terms of MAE and MPE, the two indicators are improved by 13.33%, 6.40%, 20.00%, 24.27%, 48.27% and 24.72%, 8.46%, 72.03%, 59.19%, 94.47% respectively compared with other methods. The extreme high temperature weather load prediction method proposed by the present invention has the highest prediction accuracy and the best prediction effect.

[0147] Embodiment 2 of the present invention provides an extreme high temperature weather load prediction system based on tensor completion, comprising:

[0148] A data screening module is used to calculate a comprehensive high temperature index based on the heat index and temperature considering air humidity constraints, and to screen the load data of historical extreme high temperature weather;

[0149] A data completion module is used to complete the filtered load data based on a tensor completion algorithm to obtain completed load data;

[0150] The feature screening module is used to calculate the correlation between the completed load data and the factors related to the extreme high temperature weather load, and screen the input feature quantities for load forecasting;

[0151] A model building module, used to build a LSTM-RST short-term load forecasting model based on a long short-term memory network and a rough set theory RST by using the input feature quantity;

[0152] The load forecasting module is used to predict the load in extreme high temperature weather using the LSTM-RST short-term load forecasting model.

[0153] Embodiment 3 of the present invention provides a terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps according to the method.

[0154] Embodiment 4 of the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method when executed by a processor.

[0155] Compared with the prior art, the beneficial effects of the present invention include at least:

[0156] Aiming at the situation that the load prediction data volume is small and the randomness is strong in extreme high temperature scenes, the present invention firstly screens the extreme high temperature scene data based on the two indicators of improved heat index and temperature, and defines the extreme high temperature scene of extreme weather in a formula, and then completes the load data based on tensors, and then screens the input feature quantity through correlation analysis, constructs the LSTM-RST short-term load prediction model, and evaluates the RST correction results by the mean absolute percentage error and the maximum absolute error to achieve model optimization and adjustment. The prediction results of different methods in a certain link of the present invention and the prediction results of the method of the present invention are evaluated and analyzed, and it can be seen that the present invention can effectively improve the accuracy of load prediction in extreme high temperature weather, and improve the research on load prediction in extreme weather scenes.

[0157] The improved heat index of the present invention takes air humidity as a constraint condition and takes into account the impact of weather continuity on human comfort, making up for the defect that the value is low during hot and dry weather in summer. Combined with the temperature index, it can adaptively screen out extreme high temperature weather, improve the accuracy of load forecasting under extreme high temperature weather, and can be applied to other extreme weather through adjustment.

[0158] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0159] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0160] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0161] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be customized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0162] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for extreme high temperature weather load forecasting based on tensor completion, characterized in that: include: Step 1: Calculate the comprehensive high temperature index based on the heat index and temperature considering the air humidity constraint, and filter the load data of historical extreme high temperature weather; the heat index considering the air humidity constraint is: Where: H t Represents the air humidity in the tth period; W t represents the temperature at time period t; E t represents the heat index of the tth period; γ represents the weather duration coefficient; t∈T, T represents the total number of periods in a day; The calculation formula of the comprehensive high temperature index is: Where: R d represents the comprehensive high temperature index on the dth date; W t represents the temperature at time period t; E t represents the heat index in period t; κ w and κ e Represent the weights of temperature and heat index respectively; T represents the total number of time periods in a day; Step 2: Based on the tensor completion algorithm, the load data obtained by screening in step 1 is completed to obtain the completed load data; Among them, based on the tensor completion algorithm, the function for completing the load data obtained by screening in step 1 is: Where: || || * represents the nuclear norm; α i represents the weighting coefficient; To solve the tensor, that is, the data to be completed; is the original missing tensor, i.e., the load data obtained by screening in step 1; Ω represents the set of observable data in the tensor; where l1, l2, and l3 represent the data type order, data date day order, and data daily sampling number, respectively. The data type order includes load, temperature, and heat index; j represents the rank of the matrix X; σ k (X) represents the kth singular value of the matrix X in descending order of size; X (i) Specifically: Where: X (1) , X (2) , X (3) are the modular expansion matrices of the matrix X sampled alternately by data type order, data date day order, and data daily sampling order; Step 3: Calculate the correlation between the completed load data and the factors related to the extreme high temperature weather load, and select the input characteristic quantities for load forecasting; Step 4: Using the input feature quantity, construct an LSTM-RST short-term load forecasting model based on long short-term memory network and rough set theory RST; Substitute the input feature quantity and the load data completed in step 2 into LSTM for training, and use the RST algorithm to correct the load forecast value obtained after training. Evaluate the correction result. When it meets the requirements, complete the model construction. Otherwise, adjust the LSTM or RST algorithm parameters and retrain; The mean absolute percentage error and the maximum absolute error are used to evaluate the correction results. The mean absolute percentage error measures the overall error, and the maximum absolute error measures the local maximum prediction deviation. When the average percentage error does not exceed 8% and the maximum absolute error does not exceed 10%, the requirements are met and the model construction is completed; When the average percentage error is greater than 8%, adjust the LSTM parameters and retrain; When the maximum absolute error is greater than 10%, adjust the RST parameters and make corrections; Step 5: Use the LSTM-RST short-term load forecasting model to predict the extreme high temperature weather load.

2. The method for extreme high temperature weather load forecasting based on tensor completion according to claim 1 is characterized in that: In step 3, the Pearson correlation coefficient between the completed load data and the factors related to the extreme high temperature weather load is calculated, and the six factors with the largest correlation coefficients are selected as input feature quantities for load forecasting.

3. The method for extreme high temperature weather load forecasting based on tensor completion according to claim 1 is characterized in that: The extreme high temperature weather load-related factors include maximum temperature, heat index, average temperature, average humidity, month, day, hour, minute, working day and time-of-use electricity price.

4. An extreme high temperature weather load forecasting system based on tensor completion, using the method described in any one of claims 1 to 3, characterized in that: The system comprises: A data screening module is used to calculate a comprehensive high temperature index based on the heat index and temperature considering air humidity constraints, and to screen the load data of historical extreme high temperature weather; A data completion module is used to complete the filtered load data based on a tensor completion algorithm to obtain completed load data; The feature screening module is used to calculate the correlation between the completed load data and the factors related to the extreme high temperature weather load, and screen the input feature quantities for load forecasting; A model building module, used to build a LSTM-RST short-term load forecasting model based on a long short-term memory network and a rough set theory RST by using the input feature quantity; The load forecasting module is used to predict the load in extreme high temperature weather using the LSTM-RST short-term load forecasting model.

5. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images