Rural power load interval prediction method considering interval structure, medium and processor
Through the timing convolution network and whaleization algorithm, the interval satisfaction index is optimized, and the accuracy and adaptability of rural power load interval prediction are solved, efficient and reliable interval prediction is achieved, and the stable operation of rural power systems and renewable energy are promoted. Disinfection.
Patent Information
- Application Number
- CN202411070026.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-12-27
AI Technical Summary
The prediction of rural power load intervals has problems such as difficult to control the interval width, complex principles, and high calculation costs, which affects the prediction accuracy and adaptability, and thus affects the stable operation of rural power systems.
The deterministic prediction of rural power loads is used to construct interval satisfaction indicators, and the proportional coefficients and hyperparameters are optimized through the whaleization algorithm, which is converted into a double-layer planning problem to achieve the optimal construction of the load prediction interval.
It improves the accuracy and adaptability of rural power load interval prediction, quantifies the uncertainty of electric use scenarios, promotes the consumption of renewable energy on the load side, and provides better power system scheduling decision support.
Smart Images
Figure CN119209463A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power load optimization management, and in particular to a rural power load interval prediction method, medium and processor considering interval structure. Background Art
[0002] With the development of rural power systems, the types of loads are increasing, and the energy behaviors of users are becoming more complex and changeable, which leads to a rapid increase in uncertainty on the supply and demand side, affecting the accuracy of the deterministic prediction method for rural power loads. Based on the deterministic prediction of rural power loads, interval prediction can quantify the uncertainty in the prediction, reflect the volatility of rural power loads to a certain extent, and provide upper and lower limits for the prediction results, thereby ensuring the stable dispatch of rural power systems. In view of the problems of difficult control of interval width, complex principles, and high calculation costs in the interval prediction of rural power loads, it is necessary to improve the load interval prediction technology, improve the prediction accuracy and adaptability, and ensure the stable operation of rural power systems.
[0003] In view of this, a rural power load interval prediction method, medium and processor considering interval structure are needed. Summary of the invention
[0004] In view of the problems in the prior art, the present invention provides a rural power load interval prediction method, medium and processor considering interval construction. On the basis of the deterministic prediction method of rural power load, the uncertainty in the prediction is quantified, and the probability upper and lower limits are provided for the prediction of the true value of rural power load, so as to realize more accurate rural power system dispatching. The specific technical scheme is as follows:
[0005] A rural power load interval prediction method considering interval construction includes:
[0006] The time series convolutional network is used to train and deterministically predict rural power load and extract time series features;
[0007] Construct the interval satisfaction index PISI, including interval width, interval coverage and interval center position, to achieve the optimal balance among the three;
[0008] The rural power load interval prediction model is transformed into a two-level planning problem, including using the whale algorithm to obtain the optimal proportional coefficient with the interval satisfaction index as the upper objective function, and using the root mean square error as the lower objective function to obtain the optimal hyperparameters and bring them into the time series convolutional network for deterministic prediction;
[0009] The load forecast interval is constructed using the load deterministic forecast results and the optimal proportional coefficient.
[0010] Preferably, the temporal convolutional network comprises an input layer, an output layer and a plurality of residual modules;
[0011] The residual module includes two convolution units and a 1×1 convolution module. The convolution unit first performs an extended causal convolution operation by adjusting the extended causal convolution factor d; then performs weight normalization processing to map the neural network weights to a certain range to eliminate the influence of bad data; and then uses ReLU as an activation function to enhance the nonlinear fitting ability of the neural network.
[0012] Preferably, the construction of interval satisfaction index PISI includes interval width, interval coverage and interval center position, achieving an optimal balance among the three, specifically including:
[0013] The reliability of the load forecast interval is evaluated by measuring the number of load forecast intervals covering the true value of PICP, where the larger the PICP, the higher the reliability of the load forecast interval;
[0014] The quality of uncertainty information contained in the load forecast interval is evaluated by the interval average width percentage PIMWP, where the smaller the PIMWP, the narrower the forecast interval width and the lower the uncertainty level;
[0015] The median value of the prediction interval and the true value are judged by the percentage of the mean central deviation (PIMCDP). The smaller the PIMCDP, the closer the median value of the prediction interval is to the true value, and the better the quality.
[0016] The formula for constructing the PISI including multiple evaluation indicators is as follows:
[0017] PISI=1-PIMWP(1+λPIMCDP)(1+e- η(PICP-μ) )
[0018] Where λ and η are the penalty coefficients of PIMCDP and PICP evaluation indicators respectively; μ is the confidence level (1-α). If the PISI calculation result is less than 0, then PISI is set to 0.
[0019] Preferably, the optimization method of the whale algorithm comprises the following steps:
[0020] S1, randomly initialize the whale population;
[0021] S2, calculate the fitness value of the whale;
[0022] S3, based on the comparison, the optimal fitness values of individuals and groups are obtained;
[0023] S4, the whale group updates its position, increasing the iteration once;
[0024] S5. The optimal fitness value introduced in the above step S3 is the root mean square error or the interval satisfaction index; if the optimal fitness value introduced in the above step S3 is the root mean square error, then determine whether the neural network parameters of TCN meet the optimal hyperparameter value, that is, the root mean square error is relatively minimum. If it is satisfied, the optimal hyperparameter is obtained and brought into the time series convolutional network for load deterministic prediction. If it is not satisfied, enter the step S3 of proportional coefficient optimization; if the optimal fitness value introduced in the above step S3 is the interval satisfaction index, then determine whether the proportional coefficient meets the optimal condition, that is, the interval satisfaction index is relatively maximum. If it is satisfied, the optimal proportional coefficient is obtained to construct the distribution network load interval. If it is not satisfied, enter the step S3 of hyperparameter optimization;
[0025] S6. Construct a load forecast interval using the load deterministic forecast results and the optimal proportionality coefficient.
[0026] Preferably, the PICP expression formula of the interval coverage is:
[0027]
[0028] Among them, c i The expression is as follows:
[0029]
[0030] Where N is the total number of prediction results; x i and i are the input value and true value of the neural network in the i-th sampling interval; U(x i ) and L(x i ) are the upper and lower bounds of the neural network prediction interval respectively.
[0031] Preferably, the expression of the interval average width percentage PIMWP is:
[0032]
[0033] Preferably, the interval mean center deviation percentage expression is:
[0034]
[0035] A computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method described.
[0036] A processor is used to run a program, wherein the method described is executed when the program is run.
[0037] Compared with the prior art, the present invention has the following beneficial effects:
[0038] The present invention combines the deterministic prediction results of rural power load based on the optimal hyperparameters and the optimal proportional coefficient, and uses the proportional coefficient to expand and scale the relatively optimal deterministic prediction results of rural power load, so as to obtain efficient and reliable interval prediction results, quantify the uncertainty of rural users' electricity consumption scenarios, effectively promote the consumption of renewable energy on the load side, and provide a good reference for the operation and scheduling decisions of rural power systems. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.
[0040] Figure 1 It is the overall flow chart of the method of the present invention;
[0041] Figure 2 It is the block diagram of TCN structure;
[0042] Figure 3 This is the result of using the method of the present invention to predict the rural power load interval. DETAILED DESCRIPTION
[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0044] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.
[0045] It should also be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an" and "the" are intended to include plural forms unless the context clearly indicates otherwise.
[0046] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.
[0047] Example
[0048] Please refer to Figure 1-3 The present invention discloses a rural power load interval prediction method considering interval construction, comprising the following steps:
[0049] Step 1: Use the TCN neural network model (temporal convolutional network) to perform deterministic forecasting of rural power load.
[0050] The TCN prediction structure diagram is as follows Figure 2 As shown in Figure 1. TCN consists of an input layer, an output layer, and a stack of several TCN residual modules. One branch of the TCN residual module performs a transformation operation F on the input X, and the other branch performs a simple 1×1 conversion to maintain the consistency of the number of feature maps in parallel with the existing branches. The output X of the hth residual module (h) It can be expressed as:
[0051] X (h) =ReLU(F(X (h-1) )+X (h-1) )
[0052] Where ReLU represents the activation function as shown below:
[0053]
[0054] A TCN residual module contains two convolution units and a 1×1 convolution module. In the convolution unit, the dilated causal convolution operation is first performed. By adjusting the dilated causal convolution factor d (the value of d increases with an integer power of 2), the sampling interval is changed, and a larger receptive field is obtained with fewer convolution layers, that is, the area range where the features on the convolution layer can be mapped, so that longer historical data can be received, ensuring the integrity of historical data, and only performing convolution operations on the input data before time t to obtain the output at time t, ensuring that the output is not affected by future information; then weight normalization is performed to map the neural network weights to a certain range to eliminate the influence of bad data; then ReLU is used as the activation function to enhance the nonlinear fitting ability of the neural network; finally, regularization is performed to stop the activation of neurons according to a certain probability, ensuring the training speed of the neural network and preventing overfitting.
[0055] During the training process, the parallel 1×1 convolution module not only enables the neural network to have the function of transmitting information across layers, but also ensures the consistency of input and output, which can effectively alleviate the gradient explosion or disappearance phenomenon caused by too many training layers of the neural network.
[0056] Step 2: Construct PISI (interval satisfaction index) to take into account the interval width, interval coverage and interval center position to achieve the optimal balance among the three.
[0057] (1) Prediction interval coverage probability (PICP)
[0058] PICP evaluates the load forecast interval by measuring the amount of coverage of the actual value [L(x i ),U(x i )] The larger the PICP is, the higher the reliability of the load forecast interval is. The formula is as follows:
[0059]
[0060] where c i The expression is as follows:
[0061]
[0062] Where N is the total number of prediction results; x i and i are the input value and true value of the neural network in the i-th sampling interval; U(x i ) and L(x i ) are the upper and lower bounds of the neural network prediction interval respectively.
[0063] (2) Prediction interval mean width percentage (PIMWP)
[0064] If the interval width is too large, although it can cover all the real values and meet the ideal interval coverage rate of PICP = 100%, it is impossible to analyze the impact of uncertain information on the trend of power load changes, making the interval prediction results less practical, so it is necessary to evaluate the interval width. PIMWP is used to evaluate the quality of the load forecast interval containing uncertainty information. The smaller the PIMWP, the narrower the prediction interval width and the lower the uncertainty level. The formula is as follows:
[0065]
[0066] (3) Prediction interval mean centre deviation percentage (PIMCDP)
[0067] In general, the prediction interval is required to be as centrally symmetric as possible with the true value. If only PICP and PIMWP are used as evaluation indicators in the prediction, the degree of central deviation of the prediction interval cannot be determined, and the interval quality evaluation is not comprehensive enough. Therefore, the present invention proposes the PIMCDP indicator, which is defined as the percentage of deviation between the midline value of the prediction interval and the true value when the prediction interval covers the true value. The smaller the PIMCDP, the closer the midline value of the prediction interval is to the true value, and the better its quality. The formula is as follows:
[0068]
[0069] (4) Interval satisfaction index
[0070] PICP, PIMWP and PIMCDP are three conflicting indicators, namely: the higher the PICP of the rural power load forecast interval, the higher the corresponding PIMWP will be, and the value of PIMCDP will also fluctuate due to changes in the upper and lower limits of the load interval. All three only consider the properties of a certain aspect of the load forecast interval and cannot comprehensively evaluate the quality of the forecast interval. Therefore, the forecast interval should have a width and center deviation as small as possible while meeting the PICP requirements. The present invention uses linear weighting and other methods to construct a PISI containing multiple evaluation indicators, so that it takes into account the three aspects of interval width, interval coverage capability and interval center position. The PISI formula is as follows:
[0071]
[0072] Where λ and η are the penalty coefficients of PIMCDP and PICP evaluation indicators respectively; μ is the confidence level (1-α). If the PISI calculation result is less than 0, then PISI is set to 0.
[0073] Step 3: The rural power load interval prediction model is transformed into a two-level programming problem. The WOA algorithm is used to optimize the PISI as the upper objective function and the root mean square error (RMSE) as the lower objective function to obtain the optimal proportional coefficient. The optimal rural power load prediction interval is constructed based on the deterministic prediction results.
[0074] (1) Two-level programming model
[0075] The predicted output values of the TCN neural network model for historical rural power load data are magnified and reduced by the proportional coefficients a and b respectively as the upper and lower bounds of the prediction interval, and the mathematical expressions are as follows:
[0076]
[0077] Where x i and are the input value of the neural network at time i and the predicted value of the test set; U(x i ) and L(x i ) are the upper and lower limits of the prediction interval of the neural network at time i; a i and b i is the proportional coefficient at time i.
[0078] In solving a i and b i In the process, in order to ensure the optimal prediction interval and avoid the impact of data leakage, the validation set is constructed using appropriate recent rural power load historical data, and the prediction interval is evaluated based on the validation set interval satisfaction index PISI. At the same time, the prediction interval optimization also involves the problem of deterministic prediction of rural power load. It is necessary to optimize the TCN hyperparameters based on the validation set data to ensure the relatively optimal deterministic prediction results.
[0079] Assuming that the proportional coefficient changes smoothly over a continuous period of time, the optimal proportional coefficient of the test set is and Approximately the optimal scaling factor of the validation set and (The corresponding test set post-prediction interval is optimal). At the same time, RMSE is introduced in the TCN hyperparameter optimization process to evaluate the deterministic prediction results. Therefore, the construction of the prediction interval requires solving the following two-level programming model:
[0080]
[0081] Where PISI Valid is the PISI of the validation set; RMSE Valid is the RMSE of the validation set; y i,Valid and are the true value and predicted value of the validation set respectively.
[0082] The present invention transforms the interval prediction model into a two-level programming problem, and obtains the optimal proportional coefficient of the verification set through multiple iterations of the group optimization algorithm. With appropriate TCN model hyperparameters, the optimal prediction interval construction formula for the test set is as follows:
[0083]
[0084] Where y i,Valid and They are the optimal upper bound and optimal lower bound of the prediction interval of the test set at time i.
[0085] (2) Whale Optimization Algorithm
[0086] The WOA algorithm simulates the social behavior of humpback whales, and optimizes the model's hyperparameters and ensures the accuracy of predictions through the process of whale group searching, encircling, chasing and attacking prey. The WOA algorithm has a simple principle, is easy to program, and requires few parameter settings to be adjusted. It is widely used in water resource demand forecasting, power load scheduling, fault diagnosis, image segmentation, feature selection, load forecasting and other fields.
[0087] Since the optimal position of the prey is unknown a priori, the WOA algorithm assumes the prey position as the optimization target or approximate optimal solution. Individuals in other populations use this to update their own positions and iteratively surround the prey position. The mathematical model of this process is shown in the following formula:
[0088] D=|C·X * (txt)
[0089] X(t+1)=X * (t)-A·D
[0090] Where D is the distance between individuals in other populations and the location of the prey; X * (t) is the optimal whale position at the tth iteration; X(t) is the current whale position; t is the number of iterations. The random coefficient vectors A and C can be determined using the following formula:
[0091] A=2ar-a
[0092] C=2r
[0093] Where a is a decreasing variable with values from 2 to 0; r is a random vector in [0,1].
[0094] From the process of surrounding the prey, we know that the range of A is [-a, a]. When -1≤|A|≤1, after the position is updated, the whale will shrink and surround the prey, and may approach the prey in the form of spiral steps. The probability of the whale executing the shrinking and surrounding strategy and the spiral update position strategy is 50%, and its mathematical model is determined by the following formula:
[0095] D'=|X * (txt)
[0096]
[0097] Where D' is the distance between the current position of the whale and the current optimal individual; b is a constant that controls the shape of the spiral; l is a random number in the range of [-1,1]; and p is a probability variable.
[0098] In addition, when |A|>1, whales hunt in a random search form, and its mathematical model can be expressed by the following formula:
[0099] D rand =|C·X rand -X
[0100] X(t+1)=X rand -A.D rand
[0101] Where X rand is the position of a random individual in the whale group; D rand is the distance between the current position of a random individual of the whale and the current optimal individual.
[0102] In the construction of rural power load forecast intervals, the optimization goal can be converted into the relative optimality of RMSE and PISI, where RMSE is minimized and PISI is maximized, that is, the hyperparameter optimization of the TCN neural network model and the proportional coefficient optimization of the interval construction model are simultaneously performed for the rural power load deterministic forecast model and the interval construction model, so as to achieve the relative optimality of the forecast interval. Therefore, the individual spatial dimension in the whale corresponds to the number of parameters to be optimized. Assuming that the hyperparameter set of the TCN neural network model is s, the number of hyperparameters is m, the proportional coefficient set is a, b, and the number of proportional coefficients is n, the position of the i-th whale individual can be expressed as:
[0103] X i =[x i,1 ,x i,2 ,…,x i,m+2n ]
[0104] Where x i,j is the coordinate of the individual whale position, and its corresponding relationship with the parameters to be optimized is shown below.
[0105]
[0106] Through the above method, the optimization of various parameters can be converted into the optimization of the individual position of the whale. For a known spatial coordinate position X of an individual i According to the above formula, the corresponding model parameters can be obtained, the interval prediction model is imported, and the upper and lower limits of the prediction interval and the values of various evaluation indicators are calculated. Finally, PISI is used as the upper optimization target and RMSE is used as the lower optimization target, that is, the fitness function of the WOA algorithm is:
[0107]
[0108] Where a i,j 、b i,j are the elements in a and b; f is the activation function of the TCN neural network; y is the true value.
[0109] In summary, if Figure 1 As shown, the optimization method of the whale algorithm includes the following steps:
[0110] S1, randomly initialize the whale population;
[0111] S2, calculate the fitness value of the whale;
[0112] S3, based on the comparison, the optimal fitness values of individuals and groups are obtained;
[0113] S4, the whale group updates its position, increasing the iteration once;
[0114] S5. The optimal fitness value introduced in the above step S3 is the root mean square error or the interval satisfaction index; if the optimal fitness value introduced in the above step S3 is the root mean square error, then determine whether the neural network parameters of TCN meet the optimal hyperparameter value, that is, the root mean square error is relatively minimum. If it is satisfied, the optimal hyperparameter is obtained and brought into the time series convolutional network for load deterministic prediction. If it is not satisfied, enter the step S3 of proportional coefficient optimization; if the optimal fitness value introduced in the above step S3 is the interval satisfaction index, then determine whether the proportional coefficient meets the optimal condition, that is, the interval satisfaction index is relatively maximum. If it is satisfied, the optimal proportional coefficient is obtained to construct the distribution network load interval. If it is not satisfied, enter the step S3 of hyperparameter optimization;
[0115] S6. Use the load deterministic forecast results and the optimal proportionality coefficient to construct the load forecast interval
[0116] The present invention simultaneously optimizes hyperparameters and proportional coefficients based on the WOA algorithm. In the WOA hyperparameter optimization, first, the whale population representing each hyperparameter to be optimized is randomly initialized to ensure the rationality of the optimization result; then the fitness value of the whale, that is, RMSE, is calculated; then the current position of the whale is updated according to the optimal fitness value of the individual and the population, and it is used as the reference position in the proportional coefficient optimization. In the WOA proportional coefficient optimization process, the objective function PISI is solved, and the whale position is updated, which is used as the reference position for hyperparameter optimization. WOA maintains an iterative manner for interactive optimization until the optimal hyperparameters and the optimal proportional coefficient are solved.
[0117] The present invention combines the deterministic prediction results of rural power load based on the optimal hyperparameters and the optimal proportional coefficient, and uses the proportional coefficient to expand and scale the relatively optimal deterministic prediction results of rural power load, so as to obtain efficient and reliable interval prediction results, quantify the uncertainty of rural users' electricity consumption scenarios, effectively promote the consumption of renewable energy on the load side, and provide a good reference for the operation and scheduling decisions of rural power systems.
[0118] The power load data of a rural area is taken as an example to verify the effectiveness of the rural power load interval prediction method described in the present invention.
[0119] In this embodiment, the power load data set of a rural area in southern China from January 1, 2017 to January 10, 2020 is used, and 96 points are collected in one day with a time interval of 15 minutes. In addition, meteorological factors have an important influence on load interval prediction. Accurately considering and predicting the changes in meteorological factors can improve the accuracy and generalization ability of load interval prediction, thereby providing more reliable data support for the dispatching and operation of rural power systems. Therefore, the present invention takes into account the influence of meteorological factors and collects climate condition data from January 1, 2017 to January 10, 2020 in the region, including maximum temperature, minimum temperature, average temperature, relative humidity and rainfall.
[0120] The present invention takes climate conditions, week types and historical rural power load data as input, and the model outputs the upper and lower limits of the rural power load prediction interval; classic scenarios are selected as test sets, 80% of the remaining days are training sets, and 20% are verification sets; a sliding window prediction method is adopted to roll-up the training set and test set on a daily basis.
[0121] The final prediction result is as follows: Figure 3 As shown, at a confidence level of 95%, PICP = 100%, PIMWP = 5.637%, PIMCDP = 0.932%, and PISI = 92.763%, indicating that the prediction method used in the present invention is relatively reliable and has strong adaptability.
[0122] The present invention is applicable to complex and changeable electricity consumption scenarios of rural users and can perform reliable rural power load interval prediction.
[0123] Those of ordinary skill in the art will appreciate that the units of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition of each example has been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0124] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored, etc.
[0125] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0126] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a 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, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-0nlyMemory), random access memory (RAM, RandomAccessMemory), mobile hard disk, magnetic disk or optical disk, etc., which can store program code.
[0127] 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 aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.
Claims
1. A rural power load interval prediction method considering interval structure, characterized in that: include: The time series convolutional network model is used to train and deterministically predict rural power load and extract time series features; Construct the interval satisfaction index PISI, including interval width, interval coverage and interval center position, to achieve the optimal balance among the three; The rural power load interval prediction model is transformed into a two-level planning problem, including using the whale algorithm to obtain the optimal proportional coefficient with the interval satisfaction index as the upper objective function, and using the root mean square error as the lower objective function to obtain the optimal hyperparameters and bring them into the time series convolutional network for deterministic prediction; The load forecast interval is constructed using the load deterministic forecast results and the optimal proportional coefficient.
2. A rural power load interval prediction method considering interval structure according to claim 1, characterized in that: The temporal convolutional network includes an input layer, an output layer and a plurality of residual modules; The residual module includes two convolution units and a 1×1 convolution module. The convolution unit first performs an extended causal convolution operation by adjusting the extended causal convolution factor d; then performs weight normalization processing to map the neural network weights to a certain range to eliminate the influence of bad data; and then uses ReLU as an activation function to enhance the nonlinear fitting ability of the neural network.
3. A rural power load interval prediction method considering interval structure according to claim 1, characterized in that: The interval satisfaction index PISI is constructed, including interval width, interval coverage and interval center position, to achieve the optimal balance among the three, specifically including: The reliability of the load forecast interval is evaluated by measuring the number of load forecast intervals covering the true value of PICP, where the larger the PICP, the higher the reliability of the load forecast interval; The quality of uncertainty information contained in the load forecast interval is evaluated by the interval average width percentage PIMWP, where the smaller the PIMWP, the narrower the forecast interval width and the lower the uncertainty level; The median value of the prediction interval and the true value are judged by the percentage of the mean central deviation (PIMCDP). The smaller the PIMCDP, the closer the median value of the prediction interval is to the true value, and the better the quality. The formula for constructing the PISI including multiple evaluation indicators is as follows: PISI=1-PIMWP(1+λPIMCDP)(1+e- η(PICP-μ) ) Where λ and η are the penalty coefficients of PIMCDP and PICP evaluation indicators respectively; μ is the confidence level (1-α). If the PISI calculation result is less than 0, then PISI is set to 0.
4. A rural power load interval prediction method considering interval structure according to claim 1, characterized in that: The optimization method of the whale algorithm includes the following steps: S1, randomly initialize the whale population; S2, calculate the fitness value of the whale; S3, based on the comparison, the optimal fitness values of individuals and groups are obtained; S4, the whale group updates its position and adds one iteration; S5. The optimal fitness value introduced in the above step S3 is the root mean square error or the interval satisfaction index; if the optimal fitness value in the above step S3 introduces the root mean square error, then determine whether the optimal hyperparameter value is satisfied. If so, the optimal hyperparameter is obtained and brought into the time series convolutional network for load deterministic prediction. If not, enter the step S3 of proportional coefficient optimization; if the optimal fitness value in the above step S3 introduces the interval satisfaction index, then determine whether the optimal proportional coefficient condition is satisfied. If so, the optimal proportional coefficient is obtained to construct the distribution network load interval. If not, enter the step S3 of hyperparameter optimization; S6. Construct a load forecast interval using the load deterministic forecast results and the optimal proportionality coefficient.
5. A rural power load interval prediction method considering interval structure according to claim 3, characterized in that: The PICP expression formula of the interval coverage is: Among them, c i The expression is as follows: Where N is the total number of prediction results; x i and i are the input value and true value of the neural network in the i-th sampling interval; U(x i ) and L(x i ) are the upper and lower bounds of the neural network prediction interval respectively.
6. A rural power load interval prediction method considering interval structure according to claim 4, characterized in that: The expression of the interval average width percentage PIMWP is:
7. A rural power load interval prediction method considering interval structure according to claim 4, characterized in that: The interval mean center deviation percentage expression is:
8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.
9. A processor, characterized in that: The processor is used to run a program, wherein the program executes the method according to any one of claims 1 to 7 when running.