Method, device and equipment for joint prediction of water, wind and light power considering spatio-temporal complementarity

Abstract

The present invention provides a method, device and equipment for combined forecasting of hydropower, wind power and photovoltaic power considering spatio-temporal complementarity, including collecting and sorting out data of a hydropower, wind power and photovoltaic power multi-energy complementary system, quantifying the complementarity of hydropower, wind power and photovoltaic power in the hydropower, wind power and photovoltaic power multi-energy complementary system, considering the combined influence of large-scale hydrometeorological factors on the hydropower, wind power and photovoltaic power conditions in the corresponding basins of the hydropower, wind power and photovoltaic power multi-energy complementary system, considering the autocorrelation of hydropower, wind power, photovoltaic power or total power in time, obtaining spatio-temporal complementary factors related to the power of the hydropower, wind power and photovoltaic power multi-energy complementary system, constructing a point forecasting model and an interval forecasting model based on a deep learning model, and combining the spatio-temporal complementary factors to independently forecast hydropower, wind power or photovoltaic power, and directly forecasting the medium- and long-term total power with the total power as the label. The present invention considers the complementarity within the hydropower, wind power and photovoltaic power multi-energy complementary system and large-scale hydrometeorological factors, and improves the forecasting accuracy of hydropower, wind power, photovoltaic power and total power.

Description

Technical Field

[0001] The present invention relates to the technical field of power prediction in power systems, and particularly to a method, device and equipment for joint prediction of water, wind and photovoltaic power considering spatio-temporal complementarity. Background Art

[0002] The complementary operation of water, wind and photovoltaic energy is an important way to solve the problem of new energy consumption and achieve the "dual carbon" goal. In this hybrid system, hydropower, wind power and photovoltaic power are bundled as the total power and transmitted to the power grid. The accurate prediction of the medium- and long-term total power plays an important role in the formulation of electricity prices, power market transactions, the arrangement of power generation plans and the management of operation modes of the power grid.

[0003] The water, wind and photovoltaic energy complementary system makes full use of the complementarity of runoff, wind energy and light energy resources in time, and meets the load demand through the complementary dispatching method of hydropower compensating for wind power and photovoltaic power. Therefore, there is a certain complementarity among wind power, photovoltaic power and hydropower in the water, wind and photovoltaic energy complementary system. For the power prediction of single-type power stations of hydropower, wind power and photovoltaic power, the methods can be divided into physical models and statistical models. If the medium- and long-term power prediction of the water, wind and photovoltaic energy complementary system is based on the physical model, a large amount of data such as local runoff, wind speed, photovoltaic and temperature need to be collected, and it is necessary to convert them into power based on physical relationships, which has the problems of high difficulty in data collection and unclear physical mechanisms; if the power prediction is based on the statistical model, due to the system being affected by more diverse and complex factors such as hydrology, meteorology and various power stations, the multiple uncertainties of its inputs bring difficulties to feature selection. In addition, for the total power prediction of the water, wind and photovoltaic energy complementary system, the current research on it is relatively less. Usually, the total power of the system is obtained by accumulating the separately predicted hydropower, wind power and photovoltaic power. However, due to the strong randomness and volatility of the single power itself, the prediction accuracy of the accumulated total power is not high.

[0004] However, in the past research, whether it is for the prediction of single power or total power, the complementarity of hydropower, wind power and photovoltaic power in the water, wind and photovoltaic energy complementary system in terms of resources and dispatching methods has not been fully utilized. Moreover, as a meteorological circulation anomaly in a distant area, large-scale hydrometeorological factors have a significant joint influence on the water, wind and photovoltaic conditions in the basin, which can solve the prediction difficulties brought by the multiple input uncertainties of the water, wind and photovoltaic energy complementary system. However, the related technologies have not fully considered the role of large-scale hydrometeorological factors. Summary of the Invention

[0005] In view of the deficiencies of the prior art, the object of the present invention is to provide a method, device and equipment for joint prediction of water, wind and photovoltaic power considering spatio-temporal complementarity, which fully considers the complementarity within the water, wind and photovoltaic energy complementary system and large-scale hydrometeorological factors, and improves the prediction accuracy.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0007] A combined power prediction method for water, wind and light considering spatio-temporal complementarity, comprising:

[0008] Step 1, collect and organize data of the water, wind and light multi-energy complementary system, including large-scale hydrometeorological factors, total power time series data, hydropower power time series data, wind power time series data and photovoltaic power time series data of the water, wind and light multi-energy complementary system;

[0009] Step 2, taking hydropower power, wind power, photovoltaic power or total power as the target power prediction object, and quantifying the complementarity of hydropower power, wind power and photovoltaic power in the water, wind and light multi-energy complementary system;

[0010] Step 3, considering the combined influence of large-scale hydrometeorological factors on the water, wind and light conditions of the corresponding basin of the water, wind and light multi-energy complementary system, and screening the large-scale hydrometeorological factors of the corresponding basin;

[0011] Step 4, considering the autocorrelation of hydropower power, wind power, photovoltaic power or total power in time;

[0012] Step 5, constructing a mapping relationship of the power prediction model, the mapping relationship considering the complementarity of hydropower power, wind power and photovoltaic power in the water, wind and light multi-energy complementary system, the combined influence of large-scale hydrometeorological factors on the water, wind and light conditions of the corresponding basin of the water, wind and light multi-energy complementary system, and the autocorrelation of hydropower power, wind power, photovoltaic power or total power in time, to obtain spatio-temporal complementary factors related to the power of the water, wind and light multi-energy complementary system;

[0013] Step 6, constructing a point prediction model and an interval prediction model of power based on a deep learning model;

[0014] Step 7, based on the point prediction model and the interval prediction model, and according to the spatio-temporal complementary factors, perform independent predictions of hydropower power, wind power or photovoltaic power, and directly perform medium- and long-term predictions of the total power with the total power as the label.

[0015] Further, in step 2, the complementarity of hydropower power, wind power and photovoltaic power is quantified based on the Copula function.

[0016] Further, step 2 specifically includes:

[0017] Step 201, determine the marginal distributions of hydropower power, wind power and photovoltaic power based on kernel density estimation, and the kernel density estimation formula is as follows:

[0018]

[0019] where x iSample points with the same distribution as X; K(t) is the kernel function; h is the window width; N is the number of sample points; the kernel function K(t) needs to satisfy in the real number domain:

[0020]

[0021] Step 202: Solve the joint distribution of pairwise variables based on the Copula function, and the calculation formula is as follows:

[0022]

[0023] In the formula, is the n-dimensional joint distribution function; are the marginal distribution functions of hydropower power, wind power, and photovoltaic power; C is the Copula function of these variables;

[0024] Select the Euclidean distance d 2 to evaluate the fitting degree:

[0025]

[0026]

[0027] In the formula, C n (u, v) is the empirical Copula function, I is the indicator function. When F n (x ic ) ≤ u, I takes 1, otherwise it takes 0;

[0028] Based on the established joint distribution function, the conditional distribution when X ≤ x given that the variable Y = y can be established. The variables X and Y are any two of hydropower power, wind power, and photovoltaic power. The binary conditional probability distribution function is shown as follows:

[0029]

[0030] In the formula, f(x, y) is the joint probability density function of variables X and Y; f(x) and f(y) are the probability density functions of variables X and Y respectively; u and v are the cumulative distribution functions of variables X and Y; C(u, v) is the probability density function of the Copula function;

[0031] Step 203: Based on the binary conditional probability distribution function, establish the conditional expectation distribution function as shown in the following formula:

[0032]

[0033] Based on Equation (7), the average value of variable X can be obtained when the given variable Y = y, and the conditional expectation value of another related variable can be obtained when the distribution values of hydropower, wind power, or photovoltaic power are known. Thus, the complementarity between any two of them in the hydropower, wind power, and photovoltaic multi-energy complementary system can be quantified. Based on this, the power P of the power station complementary to hydropower, wind power, or photovoltaic power is determined as P = {P 1 , P 2 ,..., P n1}.

[0034] Furthermore, in Step 3, large-scale hydrometeorological factors corresponding to the basin are screened based on the maximum mutual information coefficient Step 3 specifically includes:

[0035] Step 301: There is an ordered pair data set D = {(x mi , y mi ), i = 1, 2,..., n}. By dividing the X-axis into x m parts and the Y-axis into y m parts, a grid division G of x m × y m is obtained, and the mutual information value of each grid is calculated:

[0036]

[0037] where p(x m ) and p(y m ) are the marginal probability density functions of variables X m , Y m ; p(x m , y m ) is the joint probability density function of the two variables, that is, the ratio of the number of points in the current grid to the total number of data points;

[0038] Step 302: Normalize the row with the maximum mutual information in the grid:

[0039]

[0040] Step 303: Perform grid divisions for different schemes and repeat Steps 301 and 302. The maximum normalized mutual information value among all schemes is the maximum mutual information coefficient:

[0041]

[0042] where x m y m <B is the constraint condition for the total number of grid divisions, and B is set to the 0.6th power of the total number of data.

[0043] Further, in step 5, the spatio-temporal complementary factors include the power P of the i-th power station complementary to the target power prediction object in the t-th period i (t), the j-th large-scale hydrometeorological factor LC in the t-th period j (t), and the historical time series N(t) of the target power prediction object in the t-th period. The mapping relationship is obtained as follows:

[0044] N(t) = f(P 1 (t - 1),..., P n1 (t - m), LC 1 (t - 1),..., LC n2 (t - m), N(t - 1),..., N(t - m))( 11)

[0046] In the formula, f is the simulation function of the deep learning method; n1 is the number of power stations complementary to hydropower, wind power or photovoltaic power; n2 is the number of large-scale hydrometeorological forecast factors; m is the lag period of the spatio-temporal complementary factors, and the power of the target power prediction object in the next period is predicted by using the spatio-temporal complementary factors in the previous m periods.

[0047] Further, in step 6, an interval prediction model is constructed by using the upper and lower limit estimation method, and the deep learning is trained in the direction of better prediction interval performance by constructing a loss function. Considering that the prediction interval needs to meet the criterion of smaller interval width on the basis of covering as many true values as possible, the loss function used is shown in the following formula:

[0048] Loss = f1 + f2 (12)

[0049]

[0050] f2 = k2u q (t) - l q (t) (14)

[0051] In the formula, f1 represents the distance between the median of the interval and the actual value. The closer the median of the interval is to the actual value, the more accurate the result of the prediction interval is. At the same time, a penalty factor λ is also set in f1 q to penalize the situation where the actual value is not within the prediction interval, so as to improve the interval coverage rate during the network training process; f2 is used to calculate the interval width. Under the same interval coverage rate, the narrower the interval width, the higher the quality of the corresponding prediction interval; by adjusting the proportional coefficients k1 and k2, the network can be trained to be targeted at a certain index. When the ratio of k1 and k2 is appropriate, by adjusting the penalty factor λ q the actual value coverage rate and width of the interval can be made both appropriate; N(t) represents the actual power value in the t-th period, with the unit of MW; uq (t), l q (t) is the forecast result of the model at the t-th time period, representing the upper and lower bounds of the forecast respectively; λ q is the penalty coefficient; k1 and k2 are both proportionality coefficients.

[0052] Furthermore, it also includes step 8. For point forecasts, various indicators such as the Nash efficiency coefficient NSE, bias coefficient Bias, root mean square error RMSE, and mean absolute error MAE are used for evaluation;

[0053] For interval forecasts, an improved coverage width composite criterion CWC, interval coverage probability PICP, and interval normalized average width PINAW indicators are used for evaluation. The specific calculation formulas are as follows;

[0054]

[0055]

[0056]

[0057] In the formula, N(t) represents the actual power value at the t-th time period, in MW; T is the total number of sample time periods; represents the mean value of power, in MW; β and η are the penalty coefficients of CWC; VR is the difference between the maximum and minimum values of the forecast samples; μ is the confidence level;

[0058] Compare the medium- and long-term results of the total power obtained in step 7 with the results of the sum of hydropower, wind power, and photovoltaic power.

[0059] A water-wind-solar power combined forecasting device considering spatio-temporal complementarity includes:

[0060] A data collection module for collecting and organizing data of the water-wind-solar multi-energy complementary system, including large-scale hydrometeorological factors, time series data of the total power of the water-wind-solar multi-energy complementary system, time series data of hydropower, time series data of wind power, and time series data of photovoltaic power;

[0061] A power complementarity quantification module for quantifying the complementarity of hydropower, wind power, and photovoltaic power in the water-wind-solar multi-energy complementary system with hydropower, wind power, photovoltaic power, or total power as the target power forecasting object;

[0062] A large-scale hydrometeorological factor determination module for considering the combined influence of large-scale hydrometeorological factors on the water-wind-solar conditions in the corresponding basin of the water-wind-solar multi-energy complementary system and screening the large-scale hydrometeorological factors of the corresponding basin;

[0063] An autocorrelation determination module, configured to consider the autocorrelation of hydropower, wind power, photovoltaic power or total power over time;

[0064] A spatio-temporal complementary factor determination module, configured to construct a mapping relationship of a power prediction model, where the mapping relationship considers the complementarity of hydropower, wind power and photovoltaic power in a water-wind-solar multi-energy complementary system, the combined influence of large-scale hydrometeorological factors on the water-wind-solar conditions of the corresponding basin of the water-wind-solar multi-energy complementary system, and the autocorrelation of hydropower, wind power, photovoltaic power or total power over time, and obtain spatio-temporal complementary factors related to the power of the water-wind-solar multi-energy complementary system;

[0065] A deep learning module, configured to construct a point prediction model and an interval prediction model of power based on a deep learning model;

[0066] A power prediction module, configured to perform independent predictions of hydropower, wind power or photovoltaic power based on the point prediction model and the interval prediction model, and according to the spatio-temporal complementary factors, and directly perform medium- and long-term predictions of the total power with the total power as a label.

[0067] A water-wind-solar power joint prediction device considering spatio-temporal complementarity, including a processor and a memory for storing a computer program that can run on the processor, and when the processor is used to run the computer program, it executes the steps of the water-wind-solar power joint prediction method considering spatio-temporal complementarity described in any one of the above.

[0068] A computer storage medium, in which a computer program is stored, characterized in that when the computer program is executed by a processor, it realizes the steps of the water-wind-solar power joint prediction method considering spatio-temporal complementarity described in any one of the above.

[0069] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0070] (1) The water-wind-solar power joint prediction method, device and equipment considering spatio-temporal complementarity provided by the present invention fully consider the complementarity of hydropower, wind power and photovoltaic power in the water-wind-solar multi-energy complementary system in terms of resources and scheduling methods, and improve the prediction accuracy.

[0071] (2) The water-wind-solar power joint prediction method, device and equipment considering spatio-temporal complementarity provided by the present invention add large-scale hydrometeorological factors as spatio-temporal complementary factors, utilize the combined influence and long-term prediction ability of large-scale hydrometeorological factors on the water-wind-solar conditions of the basin, and improve the effect of long-term power prediction of the water-wind-solar multi-energy complementary system.

[0072] (3) The method, device and equipment for joint prediction of hydropower, wind power and photovoltaic power considering spatio-temporal complementarity provided by the present invention consider spatio-temporal complementarity during the independent prediction of hydropower, wind power and photovoltaic power in a multi-energy complementary system of water, wind and light. This not only improves the prediction accuracy of the power of a single type of power station, but also improves the prediction accuracy of the total power obtained by the traditional summation method.

[0073] (4) The method, device and equipment for joint prediction of hydropower, wind power and photovoltaic power considering spatio-temporal complementarity provided by the present invention obtain the total power through a joint prediction method with the total power as the label based on spatio-temporal complementary factors during the total power prediction of a multi-energy complementary system of water, wind and light. This avoids separately predicting hydropower, wind power and photovoltaic power with significant volatility and randomness, and realizes more accurate point prediction and interval prediction of the total power compared with the traditional summation method while the method is more direct and simple. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The drawings described herein are used to provide a further understanding of the present invention and form a part of this application. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0075] Figure 1 is a flowchart of the method for joint prediction of hydropower, wind power and photovoltaic power considering spatio-temporal complementarity of the present invention;

[0076] FIG. 2(a) is a schematic diagram of the expected value of hydropower under the probability distribution of photovoltaic power in a multi-energy complementary system of water, wind and light of the present invention;

[0077] FIG. 2(b) is a schematic diagram of the expected value of hydropower under the probability distribution of wind power in a multi-energy complementary system of water, wind and light of the present invention;

[0078] FIG. 2(c) is a schematic diagram of the expected value of photovoltaic power under the probability distribution of wind power in a multi-energy complementary system of water, wind and light of the present invention;

[0079] FIG. 3(a) is the point prediction result of hydropower in a multi-energy complementary system of water, wind and light of the present invention;

[0080] FIG. 3(b) is the point prediction result of wind power in a multi-energy complementary system of water, wind and light of the present invention;

[0081] FIG. 3(c) is the point prediction result of photovoltaic power in a multi-energy complementary system of water, wind and light of the present invention;

[0082] FIG. 4(a) is the interval prediction result of hydropower in a multi-energy complementary system of water, wind and light of the present invention;

[0083] FIG. 4(b) is the interval prediction result of wind power in a multi-energy complementary system of water, wind and light of the present invention;

[0084] Figure 4(c) shows the interval prediction results of the photovoltaic power of the water-wind-solar multi-energy complementary system of the present invention;

[0085] Figure 5 It is a comparison of the point prediction results of the total power of the water-wind-solar system obtained by the combined prediction method and the cumulative method;

[0086] Figure 6 It is a comparison of the interval prediction results of the total power of the water-wind-solar system obtained by the combined prediction method and the cumulative method. Specific embodiments

[0087] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention.

[0088] The present invention provides a combined power prediction method for water-wind-solar considering spatio-temporal complementarity, as Figure 1 shown, including:

[0089] Step 1: Collect and organize the data of the water-wind-solar multi-energy complementary system, including large-scale hydrological and meteorological factors, the total power time series data, hydropower power time series data, wind power power time series data, and photovoltaic power time series data of the water-wind-solar multi-energy complementary system;

[0090] Step 2: Quantify the complementarity of hydropower power, wind power, and photovoltaic power in the water-wind-solar multi-energy complementary system with the hydropower power, wind power, or photovoltaic power or the total power as the target power prediction object;

[0091] Step 3: Consider the combined influence of large-scale hydrological and meteorological factors on the water-wind-solar conditions in the corresponding basin of the water-wind-solar multi-energy complementary system, and screen the large-scale hydrological and meteorological factors in the corresponding basin;

[0092] Step 4: Consider the autocorrelation of hydropower power, wind power, photovoltaic power, or total power in time;

[0093] Step 5: Construct the mapping relationship of the power prediction model. The mapping relationship considers the complementarity of hydropower power, wind power, and photovoltaic power in the water-wind-solar multi-energy complementary system, the combined influence of large-scale hydrological and meteorological factors on the water-wind-solar conditions in the corresponding basin of the water-wind-solar multi-energy complementary system, and the autocorrelation of hydropower power, wind power, photovoltaic power, or total power in time, and obtain the spatio-temporal complementary factors related to the power of the water-wind-solar multi-energy complementary system;

[0094] Step 6: Build a point prediction model and an interval prediction model of power based on a deep learning model;

[0095] Step 7: Based on the point prediction model and the interval prediction model, and according to the spatio-temporal complementary factors, perform independent predictions of hydropower power, wind power, or photovoltaic power, and directly perform medium- and long-term predictions of the total power with the total power as the label.

[0096] The combined forecasting method for hydropower, wind power, and photovoltaic power considering spatio-temporal complementarity provided by the present invention fully considers the complementarity of hydropower, wind power, and photovoltaic power within the multi-energy complementary system of hydropower, wind power, and photovoltaic power in terms of resources and scheduling methods, thereby improving the forecasting accuracy.

[0097] The combined forecasting method for hydropower, wind power, and photovoltaic power considering spatio-temporal complementarity provided by the present invention incorporates large-scale hydrometeorological factors as spatio-temporal complementary factors, utilizes the combined influence and long-term forecasting ability of large-scale hydrometeorological factors on the hydropower, wind power, and photovoltaic power conditions in the basin, and enhances the long-term power forecasting effect of the multi-energy complementary system of hydropower, wind power, and photovoltaic power.

[0098] When performing the total power forecasting of the multi-energy complementary system of hydropower, wind power, and photovoltaic power, the combined forecasting method for hydropower, wind power, and photovoltaic power considering spatio-temporal complementarity provided by the present invention obtains the total power through the combined forecasting method with the total power as the label based on spatio-temporal complementary factors, avoiding the separate forecasting of hydropower power, wind power, and photovoltaic power with significant volatility and randomness. While the method is more direct and simple, it realizes more accurate total power point forecasting and interval forecasting compared with the traditional cumulative method.

[0099] In an embodiment of the present invention, taking the Ertan multi-energy complementary system of hydropower, wind power, and photovoltaic power as an example, due to the lack of long-time series actual system power data, first, the runoff, wind speed, photovoltaic, etc. are converted into the time series data of hydropower power, wind power, and photovoltaic power of the multi-energy complementary system of hydropower, wind power, and photovoltaic power through medium- and long-term optimal scheduling, and the time series data of the total power of the multi-energy complementary system of hydropower, wind power, and photovoltaic power are obtained through accumulation, and the following forecasting research is carried out based on this.

[0100] Through the combined forecasting method for hydropower, wind power, and photovoltaic power considering spatio-temporal complementarity provided by the present invention, in step 1, a long short-term memory network (LSTM) model is selected as the deep learning model for medium- and long-term power combined forecasting, and the required data are collected and sorted out, including 96 items of large-scale hydrometeorological factor data and the time series data of the total power, hydropower power, wind power, and photovoltaic power of the multi-energy complementary system of hydropower, wind power, and photovoltaic power. The data used in the embodiment are the monthly data from January 1959 to December 2010.

[0101] In the present invention, in step 2, the complementarity of hydropower power, wind power, and photovoltaic power is quantified based on the Copula function.

[0102] Specifically, step 2 specifically includes:

[0103] Step 201: Determine the marginal distributions of hydropower power, wind power, and photovoltaic power based on kernel density estimation. The kernel density estimation formula is as follows:

[0104]

[0105] where x iSample points with the same distribution as X; K(t) is the kernel function; h is the window width; N is the number of sample points; the kernel function K(t) needs to satisfy within the real number domain:

[0106]

[0107] Step 202: After determining the marginal distributions of the single variables of hydropower power, wind power, and photovoltaic power, solve the joint distributions of pairwise variables based on the Copula function. The calculation formula is as follows:

[0108]

[0109] In the formula, is the n-dimensional joint distribution function; are the marginal distribution functions of hydropower power, wind power, and photovoltaic power; C is the Copula function connecting the variables;

[0110] Select the Euclidean distance d 2 to evaluate the fitting degree:

[0111]

[0112]

[0113] In the formula, C n (u, v) is the empirical Copula function, I is the indicator function. When F n (x ic ) ≤ u, I takes 1, otherwise it takes 0;

[0114] Based on the established joint distribution function, a conditional distribution when X ≤ x given that the variable Y = y is known can be established. The variables X and Y are any two of hydropower power, wind power, and photovoltaic power. The binary conditional probability distribution function is shown as follows.

[0115]

[0116] In the formula, f(x, y) is the joint probability density function of the variables X and Y; f(x) and f(y) are the probability density functions of the variables X and Y respectively; u and v are the cumulative distribution functions of the variables X and Y; C(u, v) is the probability density function of the Copula function;

[0117] Step 203: Based on the binary conditional probability distribution function, establish the conditional expectation distribution function as shown in the following formula.

[0118]

[0119] Based on Equation (7), when the given variable Y = y, the average value of variable X can be obtained, and when the distribution values of hydropower, wind power or photovoltaic power are known, the conditional expectation value of another related variable can be obtained, so as to quantify the complementarity between any two of them in the hydropower-wind-solar multi-energy complementary system. Based on this, the power of the power station complementary to hydropower, wind power or photovoltaic power is determined as P = {P 1 , P 2 ,..., P n1}.

[0120] Based on the Copula function, when independently predicting the power of each single type of power station in the hydropower-wind-solar multi-energy complementary system, the complementarity between hydropower, wind power and photovoltaic power is considered, and the power of other significantly related power stations in the hydropower-wind-solar multi-energy complementary system is added to the prediction factors, which not only improves the prediction accuracy of the power of a single type of power station, but also improves the prediction accuracy of the total power obtained by the traditional cumulative method.

[0121] In the embodiment of the present invention, in step 2, three commonly used Archimedean Copula functions are used to fit the joint distributions of hydropower-photovoltaic power, hydropower-wind power and photovoltaic-wind power, and the parameter values of each Copula function are estimated based on the maximum likelihood method MLE. The calculation results are shown in Table 1.

[0122] Table 1 Copula function parameter estimation values

[0123]

[0124]

[0125] In order to select a suitable Copula function for fitting the corresponding three-way relationship, the Euclidean distance between the Copula function and the actual empirical distribution is calculated, and the results are shown in Table 2.

[0126] Table 2 Euclidean distances of different Copula functions for the output of wind, solar and hydropower

[0127]

[0128] As can be seen from Table 2, for hydropower-photovoltaic power and hydropower-wind power, the d 2 of the Frank Copula function is the smallest. For photovoltaic-wind power, the d 2The minimum. Therefore, the Frank Copula function is selected to fit the joint distributions of hydropower - photovoltaic power and hydropower - wind power, and the Gumble - Hougaard Copula function is selected to fit the joint distribution of photovoltaic power - wind power.

[0129] Based on the selected Copula functions and Equation (7), the conditional expectation distribution functions between different outputs can be established, and the images are plotted as Figure 2(a) - Figure 2(c) shown.

[0130] From Figure 2(a) - Figure 2(c) it can be seen that as the probability distribution values of photovoltaic power and wind power become larger and larger, that is, the corresponding output values of photovoltaic power and wind power become larger and larger, the corresponding conditional expectation value of hydropower becomes smaller and smaller, quantifying the complementarity between hydropower and wind power and photovoltaic power.

[0131] In the present invention, in step 3, based on the maximum information coefficient (MIC), the large - scale hydrometeorological factors LC = {LC 1 , LC 2 ,..., LC n2} corresponding to the basin are screened out.

[0132] Step 3 specifically includes:

[0133] Step 301: There exists an ordered - pair data set D = {(x mi , y mi ), i = 1, 2,..., n}. By dividing the X - axis into x m parts and the Y - axis into y m parts, a grid division G of x m ×y m is obtained, and the mutual - information value of each grid is calculated:

[0134]

[0135] where p(x m ) and p(y m ) are the marginal probability density functions of variables X m , Y m ; p(x m , y m ) is the joint probability density function of the two variables, that is, the ratio of the number of points in the current grid to the total number of data points;

[0136] Step 302: Normalize the row with the maximum mutual information in the grid:

[0137]

[0138] Step 303: Conduct grid division for different schemes, and repeat Step 301 and Step 302. The maximum normalized mutual information value among all schemes is the maximum mutual information coefficient:

[0139]

[0140] where x m y m <B is the constraint condition of the total number of grid divisions, and B is set to the 0.6th power of the total number of data.

[0141] In the related art, only linear relationships can be characterized by the linear correlation coefficient. In the present invention, the maximum mutual information coefficient (MIC) is an index for measuring the correlation coefficient between two variables based on mutual information. MIC can measure the linear and non-linear relationships between two variables, and is not easily affected by outliers in the data, featuring universality, robustness, and fairness.

[0142] In the embodiment of the present invention, in Step 3, through correlation analysis, the large-scale hydrometeorological forecasting factors of the power of each single type of power station in the system are shown in Table 3.

[0143] Table 3 Large-scale hydrometeorological forecasting factors of the power of each single type of power station in the system

[0144]

[0145]

[0146] In the present invention, in Step 5, the spatio-temporal complementary factors include the power P i (t) of the i-th power station complementary to the target power forecasting object in the t-th period, the j-th large-scale hydrometeorological factor LC j (t) in the t-th period, and the historical time series N(t) of the target power forecasting object in the t-th period, obtaining the following mapping relationship:

[0147] N(t) = f(P 1 (t - 1),..., P n1 (t - m), LC 1 (t - 1),..., LC n2 (t - m), N(t - 1),..., N(t - m))( 11)

[0149] where f is the simulation function of the deep learning method; n1 is the number of power stations whose power is complementary to the target power forecasting object; n2 is the number of large-scale hydrometeorological forecasting factors; m is the lag period of the spatio-temporal complementary factors, and the power of the target power forecasting object in the next period is forecasted using the spatio-temporal complementary factors in the previous m periods.

[0150] In step 6, for interval forecasting, an upper and lower bound estimation method is adopted to construct an interval forecasting model. By constructing a loss function, deep learning is trained in the direction of better forecasting interval performance. Considering that the forecasting interval needs to meet the criterion of having a smaller interval width while covering as many true values as possible, the loss function used is as follows:

[0151] Loss=f 1 +f2 (12)

[0152]

[0153] f2=k2u q (t)-l q (t) (14)

[0154] In the formula, f1 represents the distance between the interval median and the actual value. The closer the interval median is to the actual value, the more accurate the result of the forecasting interval. At the same time, a penalty factor λ is also set within f1 q to penalize the situation where the actual value is not within the forecasting interval, so as to improve the interval coverage rate during the network training process; f2 is used to calculate the interval width. Under the same interval coverage rate, the narrower the interval width, the higher the quality of the corresponding prediction interval; by adjusting the proportional coefficients k1 and k2, the network can be trained to be targeted towards a certain indicator. When the proportions of k1 and k2 are appropriate, by adjusting the penalty factor λ q the actual value coverage rate and width of the interval can both be made appropriate; N(t) represents the actual power value at the t-th time period, with the unit of MW; u q (t), l q (t) are the forecasting results of the model at the t-th time period, representing the upper and lower bounds of the forecast respectively; λ q is the penalty coefficient; k1 and k2 are both proportional coefficients.

[0155] The present invention also includes step 8. For point forecasting, evaluation is carried out using various indicators such as the Nash-Sutcliffe efficiency coefficient NSE, bias coefficient Bias, root mean square error RMSE, and mean absolute error MAE;

[0156] For interval forecasting, evaluation is carried out using the improved coverage width composite criterion CWC, interval coverage rate PICP, and interval normalized average width PINAW indicators;

[0157] The medium- and long-term results of the total power obtained in step 7 are compared with the results of the sum of the hydropower power, wind power, and photovoltaic power.

[0158]

[0159]

[0160]

[0161]

[0162]

[0163]

[0164]

[0165] Wherein, N(t) represents the actual power value in the t-th period, in MW; represents the predicted power value in the t-th period, in MW; T is the total number of sample periods; represents the mean value of the power, in MW; β and η are the penalty coefficients of CWC; VR is the difference between the maximum and minimum values of the prediction samples; μ is the confidence level.

[0166] In interval prediction, the higher the PICP interval coverage rate and the narrower the PINAW interval normalized average width, the better the interval prediction effect. As an index that comprehensively considers the interval coverage rate and the interval normalized average width, CWC gives an exponential penalty when the PICP is lower than the confidence level μ. When the PICP exceeds the confidence level μ, the evaluation target mainly focuses on the interval width PINAW index.

[0167] However, there is a problem with the original CWC index, that is, when the PICP is lower than the confidence level, the influence of the interval width index PINAW on the CWC index is almost ignored, and the advantages and disadvantages of the model cannot be scientifically evaluated. Here, an improved CWC comprehensive evaluation index is adopted. By adding the β parameter, the proportion of the PINAW index is linearly amplified, and the α parameter is used to avoid the problem that the influence of PICP caused by too small PINAW will be ignored, so as to more reasonably evaluate the prediction interval.

[0168] In the embodiment of the present invention, May 1959 - January 2003 is set as the calibration period, and February 2003 - April 2010 is set as the verification period. In order to verify the influence of considering spatio-temporal complementary factors on the power prediction effect of single-type power stations, different prediction schemes are set as shown in Table 4. After adjusting the parameters in the LSTM model, the point prediction results of the power of each single-type power station in the system under different schemes are shown in Table 5.

[0169] Table 4 Setting of prediction factor schemes for the power of single-type power stations

[0170]

[0171] Table 5 Comparison of point prediction effects of the power of single-type power stations in the system under different prediction schemes

[0172]

[0173]

[0174] As can be seen from Table 5, for the spatio-temporal complementary factors obtained by the present invention and by using the combined water-wind-solar power forecasting method considering spatio-temporal complementarity provided by the present invention, the forecasting effects of the power of each single type of power station in the water-wind-solar multi-energy complementary system are generally improved.

[0175] Table 6 shows the error indexes of the point forecasting results of the power of each single type of power station in the water-wind-solar multi-energy complementary system. Figure 3(a) - Figure 3(c) The point forecasting results of the hydropower, wind power and photovoltaic power in the water-wind-solar multi-energy complementary system are shown. Table 7 shows the error indexes of the interval forecasting results of the power of each single type of power station in the water-wind-solar multi-energy complementary system at a 90% confidence level. Figure 4(a) - Figure 4(c) The interval forecasting results of the hydropower, wind power and photovoltaic power in the water-wind-solar multi-energy complementary system are shown.

[0176] Table 6 Error indexes of the point forecasting results of the power of each single type of power station in the water-wind-solar multi-energy complementary system

[0177]

[0178] As can be seen from Fig. 3 and Table 6, since the correlation between the hydropower power and the corresponding forecasting factors is strong, the forecasting effects of both are good; while the forecasting results of the wind power and photovoltaic power can only reflect the overall change trend of the actual power, and the forecasting effects are relatively poor.

[0179] Table 7 Error indexes of the interval forecasting results of the power of each single type of power station in the water-wind-solar multi-energy complementary system

[0180]

[0181]

[0182] As can be seen from Fig. 4 and Table 7, at a 90% confidence level, the hydropower, wind power and photovoltaic power have all achieved good interval forecasting effects. The forecasting intervals can better cover the actual values and have a relatively narrow interval width. In the interval forecasting, the forecasting effect of the wind power is the best, followed by the hydropower power, and the photovoltaic power is the worst.

[0183] Step 8: Compare the result obtained by adding the hydropower power, wind power and photovoltaic power with the forecasting result of the total power directly obtained by the combined water-wind-solar power forecasting method considering spatio-temporal complementarity provided by the present invention, so as to verify the superiority of the present invention in the total power forecasting of the water-wind-solar multi-energy complementary system.

[0184] Table 8 shows the error indicators of the total power prediction results obtained by the present invention and the point prediction results of the total power obtained by the cumulative prediction method. Figure 5 It shows the comparison of the point prediction results of the total power obtained by the two methods. Table 9 shows the error indicators of the interval prediction results of the total power obtained by the two methods. Figure 6 It shows the comparison of the interval prediction results of the total power obtained by the two methods. It can be seen from the table that compared with the cumulative prediction method, the accuracy of the total power obtained by the prediction method provided by the present invention has been improved in both point prediction and interval prediction, verifying the effectiveness and superiority of the combined prediction method in total power prediction.

[0185] Table 8 Comparison of the point prediction results of the total power obtained by the combined prediction method and the cumulative prediction method

[0186]

[0187] Table 9 Comparison of the interval prediction results of the total power obtained by the combined prediction method and the cumulative prediction method

[0188]

[0189]

[0190] The present invention also provides a combined water, wind and solar power prediction device considering spatio-temporal complementarity, including:

[0191] A data collection module for collecting and collating data of the water, wind and solar multi-energy complementary system, including large-scale hydrometeorological factors, total power time series data, hydropower power time series data, wind power time series data and photovoltaic power time series data of the water, wind and solar multi-energy complementary system;

[0192] A power complementarity quantification module for quantifying the complementarity of hydropower power, wind power and photovoltaic power in the water, wind and solar multi-energy complementary system with hydropower power, wind power, photovoltaic power or total power as the target power prediction object;

[0193] A large-scale hydrometeorological factor determination module for considering the combined influence of large-scale hydrometeorological factors on the water, wind and solar conditions of the corresponding basin of the water, wind and solar multi-energy complementary system and screening the large-scale hydrometeorological factors of the corresponding basin;

[0194] An autocorrelation determination module for considering the autocorrelation of hydropower power, wind power, photovoltaic power or total power in time and determining the historical time series of hydropower power, wind power, photovoltaic power or total power;

[0195] A spatio-temporal complementary factor determination module, which is used to construct the mapping relationship of the power prediction model. The mapping relationship takes into account the complementarity of hydropower, wind power, and photovoltaic power in the water-wind-solar multi-energy complementary system, the combined influence of large-scale hydrometeorological factors on the water-wind-solar conditions in the corresponding basins of the water-wind-solar multi-energy complementary system, and the autocorrelation of hydropower, wind power, photovoltaic power, or total power in time, so as to obtain spatio-temporal complementary factors related to the power of the water-wind-solar multi-energy complementary system;

[0196] A deep learning module, which is used to construct a point prediction model and an interval prediction model of power based on a deep learning model;

[0197] A power prediction module, which is used to independently predict hydropower, wind power, or photovoltaic power based on the point prediction model and the interval prediction model, and directly predict the medium- and long-term total power with the total power as the label.

[0198] The present invention further includes a verification module, which is used to evaluate the point prediction by using indicators such as the Nash-Sutcliffe efficiency coefficient NSE, bias coefficient Bias, root mean square error RMSE, and mean absolute error MAE;

[0199] For interval prediction, an improved coverage width composite criterion CWC, interval coverage probability PICP, and interval normalized average width PINAW indicators are used for evaluation;

[0200] The total power of the water-wind-solar multi-energy complementary system is obtained by accumulating the separately predicted hydropower, wind power, and photovoltaic power, and compared with the combined prediction results of the medium- and long-term hydropower, wind power, and photovoltaic power obtained in step 7.

[0201] The present invention further provides a water-wind-solar power joint prediction device considering spatio-temporal complementarity, including a processor and a memory for storing a computer program that can run on the processor. When the processor is used to run the computer program, it executes the steps of the water-wind-solar power joint prediction method considering spatio-temporal complementarity described in any one of the above.

[0202] The memory in the embodiments of the present invention is used to store various types of data to support the operation of the water-wind-solar power joint prediction device considering spatio-temporal complementarity. Examples of these data include: any computer program for operating on the water-wind-solar power joint prediction device considering spatio-temporal complementarity.

[0203] The method for joint prediction of water, wind and photovoltaic power considering spatio-temporal complementarity disclosed in the embodiments of the present invention can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the method for joint prediction of water, wind and photovoltaic power considering spatio-temporal complementarity can be completed by the integrated logic circuit of the hardware in the processor or the instructions in the form of software. The above-mentioned processor may be a general-purpose processor, a digital signal processor (DSP, Digital Signal Processor), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or any conventional processor, etc. Combining the steps of the method disclosed in the embodiments of the present invention can be directly embodied as being executed and completed by the hardware decoding processor, or executed and completed by the combination of the hardware and software modules in the decoding processor. The software module may be located in the storage medium, and the storage medium is located in the memory. The processor reads the information in the memory and combines its hardware to complete the steps of the method for joint prediction of water, wind and photovoltaic power considering spatio-temporal complementarity provided in the embodiments of the present invention.

[0204] In an exemplary embodiment, the device for joint prediction of water, wind and photovoltaic power considering spatio-temporal complementarity can be implemented by one or more application-specific integrated circuits (ASICs, Application Specific Integrated Circuits), DSPs, programmable logic devices (PLDs, Programmable Logic Devices), complex programmable logic devices (CPLDs, Complex Programmable Logic Devices), FPGAs, general-purpose processors, controllers, microcontroller units (MCUs, MicroController Units), microprocessors (Microprocessors), or other electronic components, and is used to execute the foregoing method.

[0205] It can be understood that the memory can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM, Read Only Memory), a programmable read-only memory (PROM, Programmable Read-Only Memory), an erasable programmable read-only memory (EPROM, Erasable Programmable Read-Only Memory), an electrically erasable programmable read-only memory (EEPROM, Electrically Erasable Programmable Read-Only Memory), a ferromagnetic random access memory (FRAM, ferromagnetic random access memory), a flash memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM, Compact Disc Read-Only Memory); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a random access memory (RAM, RandomAccessMemory), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as a static random access memory (SRAM, Static Random Access Memory), a synchronous static random access memory (SSRAM, Synchronous Static Random Access Memory), a dynamic random access memory (DRAM, Dynamic Random Access Memory), a synchronous dynamic random access memory (SDRAM, SynchronousDynamic Random Access Memory), a double data rate synchronous dynamic random access memory (DDRSDRAM, Double Data Rate Synchronous Dynamic Random Access Memory), an enhanced synchronous dynamic random access memory (ESDRAM, Enhanced Synchronous Dynamic Random AccessMemory), a synchronous link dynamic random access memory (SLDRAM, SyncLink Dynamic Random AccessMemory), a direct rambus random access memory (DRRAM, Direct Rambus Random Access Memory). The memory described in the embodiments of the present invention is intended to include but not be limited to these and any other suitable types of memory.

[0206] A computer storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the method for jointly forecasting the water, wind and light power considering the spatio-temporal complementarity described in any one of the above are implemented.

[0207] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and its equivalent technologies, the present invention also intends to include these changes and modifications.

Claims

1. A combined prediction method for water, wind and light power considering spatio-temporal complementarity, characterized in that, Including: Step 1: Collect and organize the data of the water-wind-solar multi-energy complementary system, including large-scale hydrological and meteorological factors, the total power time series data of the water-wind-solar multi-energy complementary system, the hydropower power time series data, the wind power time series data, and the photovoltaic power time series data; Step 2: Quantify the complementarity of hydropower power, wind power, and photovoltaic power in the water-wind-solar multi-energy complementary system with the hydropower power, wind power, photovoltaic power, or total power as the target power prediction object; Step 3: Consider the combined influence of large-scale hydrological and meteorological factors on the water-wind-solar conditions in the corresponding basin of the water-wind-solar multi-energy complementary system, and screen the large-scale hydrological and meteorological factors of the corresponding basin; Step 4: Consider the autocorrelation of hydropower power, wind power, photovoltaic power, or total power over time; Step 5: Construct the mapping relationship of the power prediction model. The mapping relationship considers the complementarity of hydropower power, wind power, and photovoltaic power in the water-wind-solar multi-energy complementary system, the combined influence of large-scale hydrological and meteorological factors on the water-wind-solar conditions in the corresponding basin of the water-wind-solar multi-energy complementary system, and the autocorrelation of hydropower power, wind power, photovoltaic power, or total power over time, and obtain the spatio-temporal complementary factors related to the power of the water-wind-solar multi-energy complementary system; In step 5, the spatio-temporal complementary factors include the power P i (t) of the i-th power station complementary to the target power prediction object in the t-th period, the j-th large-scale hydrometeorological factor LC j (t) in the t-th period, and the historical time series N(t) of the target power prediction object in the t-th period, and the mapping relationship is obtained as follows: i (t), the j-th large-scale hydrometeorological factor LC j (t), the historical time series N(t) of the target power prediction object in the t-th period, and the mapping relationship is obtained as follows: In the formula, f is the simulation function of the deep learning method; n1 is the number of power stations complementary to hydropower power, wind power, or photovoltaic power; n2 is the number of large-scale hydrological and meteorological prediction factors; m is the lag period of the spatio-temporal complementary factor, and the power of the target power prediction object in the next period is predicted using the spatio-temporal complementary factors in the previous m periods; Step 6: Construct a point prediction model and an interval prediction model for power based on the deep learning model; Step 7: Based on the point prediction model and the interval prediction model, and according to the spatio-temporal complementary factors, perform independent predictions of hydropower power, wind power, or photovoltaic power, and directly perform medium- and long-term predictions of the total power with the total power as the label.

2. The method for jointly predicting water-wind-solar power considering spatio-temporal complementarity according to claim 1, characterized in that: In step 2, the complementarity of hydropower power, wind power, and photovoltaic power is quantified based on the Copula function.

3. The method for jointly predicting water-wind-solar power considering spatio-temporal complementarity according to claim 2, characterized in that: The specific steps of step 2 include: Step 201: Determine the marginal distributions of hydropower power, wind power, and photovoltaic power based on kernel density estimation, and the kernel density estimation formula is as follows: where x i is a sample point of the same distribution as X; K(t) is the kernel function; h is the window width; N is the number of sample points; the kernel function K(t) needs to satisfy in the real number domain: Step 202: Solve the joint distribution of pairwise variables based on the Copula function, and the calculation formula is as follows: wherein, is an n-dimensional joint distribution function; are the marginal distribution functions of the hydropower, wind power, and photovoltaic power; C is the Copula function of the connection variable; Select the Euclidean distance d 2 Evaluate the fitting degree: where C n (u, v) is the empirical Copula function, I is the indicator function, when F n (x ic ) ≤ u, I takes 1, otherwise it takes 0; Based on the established joint distribution function, the conditional distribution when X≤x given Y = y can be established. Variables X and Y are any two of hydropower power, wind power, and photovoltaic power, and the binary conditional probability distribution function is shown as follows: In the formula, f(x,y) is the joint probability density function of variables X and Y; f(x) and f(y) are the probability density functions of variables X and Y respectively; u and v are the cumulative distribution functions of variables X and Y; C(u,v) is the probability density function of the Copula function; Step 203: Based on the binary conditional probability distribution function, establish the conditional expectation distribution function as shown in the following formula: Based on Equation (7), the average value of variable X can be obtained when the given variable Y = y, and the conditional expectation value of another related variable can be obtained when the distribution values of hydropower, wind power, or photovoltaic power are known. Thus, the complementarity between any two of them in the hydropower-wind power-photovoltaic multi-energy complementary system can be quantified, and based on this, the power of the power station complementary to hydropower, wind power, or photovoltaic power can be determined.

4. The method for joint forecasting of hydropower, wind power and photovoltaic power considering spatio-temporal complementarity according to claim 1, wherein: In step 3, large-scale hydrometeorological factors corresponding to the basin are screened based on the maximum mutual information coefficient Step 3 specifically includes: Step 301. There exists an ordered pair dataset D = {(x mi , y mi ), i = 1, 2, …, n}. By dividing the X-axis into x m parts and the Y-axis into y m parts, a grid partition G of x m × y m is obtained, and the mutual information value of each grid is calculated: Wherein, p(x m ) and p(y m ) are the marginal probability density functions of variables X m , Y m ; p(x m , y m ) is the joint probability density function of the two variables, that is, the ratio of the number of points in the current grid to the total number of data points; Step 302: Normalize the maximum value row of mutual information in the grid: Step 303: Perform grid division for different schemes, and repeat Step 301 and Step 302. The maximum normalized mutual information value among all schemes is the maximum mutual information coefficient: where x m y m <B is the total number of grid division constraint conditions, and B is set to the 0.6th power of the total number of data.

5. The method for joint forecasting of hydropower, wind power and photovoltaic power considering spatio-temporal complementarity according to claim 1, wherein: In Step 6, an interval forecasting model is constructed by using the upper and lower limit estimation method, and the deep learning is trained in the direction of better forecasting interval performance by constructing a loss function. Considering that the forecasting interval needs to meet the criterion of having a smaller interval width while covering as many true values as possible, the loss function used is shown in the following formula: Loss = f1 + f2 (12) f2 = k2|u q (t) - l q (t)| (14) In the formula, f1 represents the distance between the interval median and the actual value. The closer the interval median is to the actual value, the more accurate the result of the prediction interval. At the same time, a penalty factor λ is set within f1 q to penalize the case where the actual value is not within the prediction interval, so as to improve the interval coverage rate during the network training process; f2 is used to calculate the interval width. Under the condition of the same interval coverage rate, the narrower the interval width, the higher the quality of the corresponding prediction interval; by adjusting the proportionality coefficients k1 and k2, the network can be trained to be targeted towards a certain index. When the ratio of k1 and k2 is appropriate, by adjusting the penalty factor λ q the actual value coverage rate and width of the interval can be made to meet the requirements; N(t) represents the actual power value at the t-th time period, with the unit of MW; u q (t), l q (t) are the prediction results of the model at the t-th time period, representing the upper bound and lower bound of the prediction respectively; λ q is the penalty coefficient; k1 and k2 are both proportionality coefficients.

6. The method for joint forecasting of hydropower, wind power and photovoltaic power considering spatio-temporal complementarity according to claim 1, wherein: It further includes Step 8. For point forecasting, various indexes such as Nash efficiency coefficient NSE, bias coefficient Bias, root mean square error RMSE and mean absolute error MAE are used for evaluation; For interval forecasting, improved coverage width comprehensive criterion CWC, interval coverage probability PICP and interval normalized average width PINAW indexes are used for evaluation, and the specific calculation formulas are as follows; In the formula, N(t) represents the actual power value at the t-th time period, MW; T is the total number of sample time periods; represents the mean power, MW; β and η are the penalty coefficients of CWC; VR is the difference between the maximum and minimum values of the forecast sample; μ is the confidence level; Compare the long-term result of the total power obtained in Step 7 with the result of the sum of hydropower, wind power and photovoltaic power.

7. A combined forecasting device for water, wind and light power considering spatio-temporal complementarity, characterized in that Including: A data collection module for collecting and collating data of the hydropower, wind power and photovoltaic power multi-energy complementary system, including large-scale hydrometeorological factors, time series data of the total power of the hydropower, wind power and photovoltaic power multi-energy complementary system, time series data of hydropower, time series data of wind power and time series data of photovoltaic power; A power complementarity quantification module for quantifying the complementarity of hydropower, wind power and photovoltaic power in the hydropower, wind power and photovoltaic power multi-energy complementary system with hydropower, wind power, photovoltaic power or total power as the target power forecasting object; A large-scale hydrometeorological factor determination module for considering the combined influence of large-scale hydrometeorological factors on the hydropower, wind power and photovoltaic power conditions in the corresponding basin of the hydropower, wind power and photovoltaic power multi-energy complementary system, and screening the large-scale hydrometeorological factors of the corresponding basin; An autocorrelation determination module for considering the autocorrelation of hydropower, wind power, photovoltaic power or total power in time; A spatio-temporal complementary factor determination module is used to construct the mapping relationship of the power prediction model. The mapping relationship takes into account the complementarity of hydropower, wind power, and photovoltaic power in the water-wind-solar multi-energy complementary system, the combined influence of large-scale hydrometeorological factors on the water-wind-solar conditions in the corresponding basins of the water-wind-solar multi-energy complementary system, and the autocorrelation of hydropower, wind power, photovoltaic power, or total power over time, and obtains spatio-temporal complementary factors related to the power of the water-wind-solar multi-energy complementary system; the spatio-temporal complementary factors include the power P of the i-th power station complementary to the target power prediction object at the t-th time period i (t), the j-th large-scale hydrometeorological factor LC j (t) at the t-th time period, and the historical time series N(t) of the target power prediction object at the t-th time period, and the mapping relationship is obtained as follows: In the formula, f is the simulation function of the deep learning method; n1 is the number of power stations complementary to hydropower, wind power or photovoltaic power; n2 is the number of large-scale hydrometeorological forecasting factors; m is the lag period of the spatio-temporal complementary factor, and the power of the target power forecasting object in the next time period is forecasted by using the spatio-temporal complementary factors in the previous m time periods; A deep learning module for constructing a point forecasting model and an interval forecasting model of power based on a deep learning model; A power forecasting module for independently forecasting hydropower, wind power or photovoltaic power based on the point forecasting model and the interval forecasting model according to spatio-temporal complementary factors, and directly forecasting the long-term and medium-term total power with the total power as the label.

8. A combined water, wind and solar power forecasting device considering spatio-temporal complementarity, characterized in that, It includes a processor and a memory for storing a computer program that can run on the processor. When the processor is used to run the computer program, it executes the steps of the method for jointly predicting the water, wind, and light power considering spatio-temporal complementarity according to any one of claims 1-7 above.

9. A computer storage medium, characterized in that: A computer program is stored in the computer storage medium, characterized in that when the computer program is executed by a processor, it realizes the steps of the method for jointly predicting the water, wind, and light power considering spatio-temporal complementarity according to any one of claims 1-7 above.

Images