TWI781872B - Method for predicting solar power generation with shading effect - Google Patents

Abstract

The present invention relates to a method for predicting solar power generation with a shading effect, especially a weight ratio table using a corresponding weight ratio table of the illuminance data of a shading photosensitive element and the illuminance data of an unshaded photosensitive element, so as to more accurately know the shading condition, and a method for mathematical modeling prediction of a solar power generation system due to shading effect.

Description

Prediction method of solar power generation with shading effect

The present invention relates to a method for predicting solar power generation with shading effect, in particular to a method of using a weight ratio table corresponding to the illuminance data of a photosensitive element with shading and the illuminance data of an unshaded photosensitive element to more accurately know the shading status and target A method for mathematical modeling and forecasting of solar power generation forecasting systems based on shading effects.

From the literature, there are some studies on the impact of solar shading effect, which are analyzed from the properties of solar cells to obtain the maximum power point for tracking, but for solar power generation prediction models, it is difficult to integrate this method into And use it to model, after all, the mathematical model of environmental factors for solar power generation has not been fully disclosed in the literature. The maximum power point tracking of the shading effect is different from the solar power generation model of the shading effect established in this project, although as in the patent case of the Republic of China Patent No. I751507 applied by the inventor of this case, the establishment of a prediction model for fixed solar power generation, Or other similar modeling systems, if the actual power generation data is simply collected, the collected data should include the current voltage, current, illuminance, atmospheric temperature and humidity, solar panel temperature, the horizontal angle of the sun and the actual solar cell power generation. Pitch angle, air pollution index and other information, the modeling system cannot know whether there is shading in the system, so it is impossible to perform mathematical modeling on the shading mode more accurately.

The main purpose of the present invention is to provide a method for predicting solar power generation with shading effect. In order to obtain the shading condition more accurately and to mathematically model the solar power generation system for the shading effect, the present invention proposes an array constructed by using a plurality of illuminance sensors to detect the shading condition of the solar cell. And according to the shading situation, different neural network weight outputs are selected to complete more advanced shading effect power generation prediction. This project is expected to collect at least 4 consecutive months of data to become the data of the training group, and to verify through subsequent data accumulation as the control group. The overall data will be planned to span at least two seasons.

，其中，m _ij 、σ _ij,n 、

、

為可調整之控制參數，σ _ij,L 為中心點在m _ij 之歸屬函數左側寬度參數，σ _ij,R 為中心點在m _ij 之歸屬函數右側寬度參數。 較佳地，步驟(SA7)之該模糊類神經網路其數學方程式如下：

，其中，σ _r 為中心點在I _rmax之歸屬函數寬度參數。 較佳地，步驟(SB)更包括諾以下步驟：(SB1)將該電池模組仍放置於戶外，以蒐集該第二組複數種有遮陰感光元件照度資料及該第二組未遮陰感光元件照度資料；(SB2)該終端機將所有資料儲存於該資料庫中；(SB3)判斷蒐集資料是否已達到一設定之第二階段時間，如果是，進入步驟(SB4)，如果否，進入步驟(SB1)； (SB4)該終端機以該遞迴式模糊類神經網路為預測器，來針對該資料庫中之該第二組未遮陰感光元件照度資料進行再一次數學模型的建立；(SB5)產生一第二組無遮陰效應之太陽能發電模型，該終端機將之存入該資料庫；(SB6)該終端機T利用該模糊類神經網路，對該第二組複數種有遮陰感光元件照度資料進行調整，提供一第二組複數種權重值模型，並存入該資料庫；(SB7)該終端機將該第二組無遮陰效應之太陽能發電模型與該第二組複數種權重值模型分別相乘，得出的複數個乘積之後，可計算出一第二平均最大發電功率，為該電池模組在穩態運行時直接上線運行一第二階段時間的一第二相關參數模組，該第二相關參數模組將持續對該訓練預測數學模型進行更新、學習及優化，以將該訓練預測數學模型成為一持續訓練預測數學模型，並將該持續訓練預測數學模型存人該資料庫；及(SB8)利用該持續訓練預測數學模型以依據在當時的一第三相關參數模組，預先評估一持續訓練預測最大發電功率模組，並將該持續訓練預測最大發電功率模組存人該資料庫中。 較佳地，步驟(SB7)之該第二相關參數模組包括一第二系統相關參數模組與一第一環境相關參數模組；步驟(SB8)之該第三相關參數模組包括一第三系統相關參數模組與一第三環境相關參數模組。 較佳地，該第二系統相關參數模組包括以下任一項資料或其產生的組合：電壓、電流、太陽能電池溫度、及太陽能電池總幅照度/輻射量；該第二環境相關參數模組包括以下任一項資料或其產生的組合：大氣溫度、相對濕度、遮陰溫度、全天日射量、場址日照強度、風速及最大功率點追蹤(MPPT；Maximum power point tracking)後的太陽能電池輸出之電壓、電流與功率；該第三系統相關參數模組包括以下任一項資料或其產生的組合：電壓、電流、太陽能電池溫 度、及太陽能電池總幅照度/輻射量；該第三環境相關參數模組包括以下任一項資料或其產生的組合：大氣溫度、相對濕度、遮陰溫度、降雨機率、全天日射量、場址日照強度、風速及最大功率點追蹤(MPPT；Maximum power point tracking)後的太陽能電池輸出之電壓、電流與功率。 The solar power generation forecasting method with shading effect of the present invention includes the following steps: (SA) continuously collect at least one solar panel of a battery module of a newly installed solar power generation forecasting system with shading effect, and run a first solar panel in a steady state A first set of illuminance data of multiple shaded photosensitive elements and a first set of illuminance data of non-shaded photosensitive elements collected in one stage, and using a recursive fuzzy neural network as a predictor, for the second A set of illuminance data of unshaded photosensitive elements generates a first set of solar power generation models without shading effect, and a fuzzy neural network is used to adjust the first set of illuminance data of multiple shaded photosensitive elements to provide a The first group of multiple weight value models, the first group of solar power generation models without shading effect and the first group of multiple weight value models are multiplied respectively, and after the obtained multiple products, a first generated can be calculated An average maximum power generation module, and derive a first relevant parameter module to become a training prediction mathematical model, and store the training prediction mathematical model in a database; (SB) continue to continuously collect the battery module For the solar panel, a second set of illuminance data of a plurality of shading photosensitive elements and a second set of illuminance data of non-shading photosensitive elements collected during the second stage of steady-state operation, and using the recursive fuzzy The neural network is a predictor, which generates a second set of solar power generation models without shading effect for the second set of illuminance data of the unshaded photosensitive element, and utilizes the fuzzy neural network to have an effect on the second set of plural types. The illuminance data of the shading photosensitive element is adjusted to provide a second set of multiple weight value models, and the terminal T multiplies the second set of solar power generation models without shading effect by the second set of multiple weight value models , after obtaining the complex number of products, a second average maximum generating power module can be calculated, and a second related parameter module can be derived, and the second related parameter module will continue to carry out the training prediction mathematical model update, learn and optimize the training predictive mathematical model into a continuous training predictive mathematical model, and store the continuous training predictive mathematical model in the database; and (SC) use the continuous training predictive mathematical model to A third related parameter module is used to pre-evaluate a continuous training predicted maximum generating power module, and store the continuous training predicted maximum generating power module in the database. Preferably, the step (SA) further includes the following steps: (SA1) place a photosensitive element on each of the N positions of the solar panel of the battery module of the solar power generation prediction system with shading effect, N≧1 ; (SA2) place the battery module outdoors to use a terminal to collect the illuminance data of the first set of plural kinds of photosensitive elements with shading and the illuminance data of the first set of non-shaded photosensitive elements; (SA3) the terminal The computer stores all data in a database; (SA4) judges whether the data collection has reached a set first stage time, if yes, enters step (SA5), if not, enters step (SA3); (SA5) the The terminal uses a recursive fuzzy neural network as a predictor to establish a mathematical model for the first group of unshaded photosensitive element illuminance data in the database; (SA6) generate a first group of unshaded The solar power generation model of the shading effect, which is stored in the database by the terminal; (SA7) The terminal uses a fuzzy neural network to adjust the illuminance data of the first group of plural kinds of photosensitive elements with shading, Provide a first set of multiple weight value models and store them in the database; and (SA8) the terminal unit associates the first set of solar power generation models without shading effect with the first set of multiple weight value models After multiplying the multiple products obtained, a first average maximum power generation can be calculated, which is a first relevant parameter module for the battery module in the first stage of steady state operation, so as to become a predictive mathematical model , and store the predictive mathematical model in the database. Preferably, the first related parameter module in step (SA8) includes a first system related parameter module and a first environment related parameter module. Preferably, wherein, the first system-related parameter module includes any one of the following data or a combination thereof: voltage, current, solar cell temperature, solar cell total illuminance/radiation; the first environment-related parameter module The group includes any one of the following data or a combination thereof: atmospheric temperature, relative humidity, shade temperature, solar radiation throughout the day, site sunlight intensity, wind speed, and solar energy after maximum power point tracking (MPPT; Maximum power point tracking) Battery output voltage, current and power. Preferably, the mathematical equation of the recursive fuzzy neural network in step (SA5) is as follows:

, where m _ij , σ _ij,n ,

,

is an adjustable control parameter, σ _ij,L is the width parameter on the left side of the membership function with the center point at m _ij , and σ _ij,R is the width parameter at the right side of the membership function with the center point at m _ij . Preferably, the mathematical equation of the fuzzy neural network in step (SA7) is as follows:

, where, σ _r is the width parameter of the membership function with the center point at I _{r max} . Preferably, the step (SB) further includes the following steps: (SB1) placing the battery module outdoors to collect the illuminance data of the second group of plural types of photosensitive elements with shade and the second group of unshaded Photosensitive element illuminance data; (SB2) the terminal stores all data in the database; (SB3) judges whether the collected data has reached a set second stage time, if yes, enter step (SB4), if not, Enter step (SB1); (SB4) The terminal uses the recursive fuzzy neural network as a predictor to perform another mathematical model for the illuminance data of the second group of unshaded photosensitive elements in the database Establish; (SB5) generate a second group of solar power generation models without shading effect, and the terminal machine stores it in the database; (SB6) the terminal machine T utilizes the fuzzy neural network to generate the second group Adjust the illuminance data of a plurality of shading photosensitive elements, provide a second set of multiple weight value models, and store them in the database; (SB7) the terminal combines the second set of solar power generation models without shading effects with The second set of multiple weight models are multiplied separately, and after the multiple products are obtained, a second average maximum generating power can be calculated, which is the second stage time when the battery module is directly on-line during steady-state operation. A second relevant parameter module, the second relevant parameter module will continue to update, learn and optimize the training prediction mathematical model, so that the training prediction mathematical model becomes a continuous training prediction mathematical model, and the continuous The training prediction mathematical model is stored in the database; and (SB8) using the continuous training prediction mathematical model to pre-evaluate a continuous training prediction maximum power generation module according to a third relevant parameter module at that time, and the continuous training The model for training and predicting the maximum generating power is stored in the database. Preferably, the second related parameter module in step (SB7) includes a second system related parameter module and a first environment related parameter module; the third related parameter module in step (SB8) includes a first Three system-related parameter modules and a third environment-related parameter module. Preferably, the second system-related parameter module includes any one of the following data or a combination thereof: voltage, current, solar cell temperature, and total illuminance/radiation of the solar cell; the second environment-related parameter module Including any of the following data or a combination thereof: atmospheric temperature, relative humidity, shading temperature, solar radiation throughout the day, site sunlight intensity, wind speed, and solar cells after maximum power point tracking (MPPT; Maximum power point tracking) Output voltage, current, and power; the third system-related parameter module includes any of the following data or a combination thereof: voltage, current, solar cell temperature, and solar cell total irradiance/radiation; the third environment The relevant parameter module includes any of the following data or a combination thereof: atmospheric temperature, relative humidity, shading temperature, rainfall probability, solar radiation throughout the day, site sunlight intensity, wind speed, and maximum power point tracking (MPPT; Maximum power The output voltage, current and power of the solar cell after point tracking).

In order to make the above-mentioned features and advantages of the present invention more comprehensible, preferred embodiments will be described in detail below together with the accompanying drawings.

S1-S16: Step number

1: Solar power generation prediction system with shading effect

11:Battery module

A: Solar panels

L1~L4: photosensitive element

T: terminal

2: Database

22: The first group of illuminance data of multiple types of shaded photosensitive elements

23: The first group of illuminance data of unshaded photosensitive elements

3: Recursive fuzzy neural network

24: The first group of solar power generation models without shading effect

4: Fuzzy neural network

W _TJ : The first group of complex weight value models

111: The first related parameter module

11A: Training parameter module

112: theoretical maximum generating power module

11B: Training prediction mathematical model materials

22': The second group of multiple kinds of shading photosensitive element illumination data

23': The second group of illuminance data of unshaded photosensitive elements

24': The second group of solar power generation models without shading effect

W _TJ ': The second group of complex weight value models

113: The second related parameter module

11C: Continuous training of predictive mathematical models

114: The third related parameter module

19: Continuous training to predict the maximum power generation module

1111: The first system related parameter module

1112: First environment related parameter module

1131: Second system related parameter module

1132: Second environment related parameter module

1141: The third system related parameter module

1142: The third environment related parameter module

3: Asymmetric membership function fuzzy neural network

11B: Training predictive mathematical models

11C: Continuous training of predictive mathematical models

113: The second related parameter module

1131: Second system related parameter module

1132: Second environment related parameter module

114: The third related parameter module

19: Continuous training to predict the maximum power generation module

1141: The third system related parameter module

1142: The third environment related parameter module

MP1: The first average maximum generating power module

MP2: The second average maximum generating power module

Fig. 1 and Fig. 2 are a step diagram of a solar power generation forecasting method with a shading effect of the present invention; Fig. 3 is a schematic diagram of a monitoring system for a solar power generation prediction system with a shading effect of the present invention; Fig. 4 is a shading effect of the present invention Figure 5 is a diagram of the power generation prediction module of the effect solar power generation prediction method; and Fig. 5 is a corresponding weight ratio table of the illuminance data of the shading photosensitive element and the illuminance data of the non-shading photosensitive element of a kind of solar power generation prediction method with shading effect of the present invention.

The present invention can be better understood with reference to the detailed description set forth herein and the accompanying drawings. Various embodiments are discussed below with reference to the figures. Those skilled in the art will readily appreciate, however, that the detailed description given herein with respect to the figures is for explanatory purposes only, as the methods and systems may extend beyond the described embodiments. For example, the teachings given and the requirements of a particular application may dictate many alternative and suitable ways of implementing the functionality of any detail described herein. Accordingly, any method may extend beyond the specific implementation options described and illustrated in the following examples.

Please refer to Fig. 1 and Fig. 2, Fig. 3 and Fig. 4 at the same time, which are a step diagram of a solar power generation prediction method with a shading effect of the present invention, a schematic diagram of a monitoring system for a solar power generation prediction system with a shading effect of the present invention, and the present invention A power generation prediction module diagram of a solar power generation prediction method with shading effect. The method for predicting the maximum power generation of the shadow mode solar energy system of the present invention comprises the following steps: (S1) each of the four positions of the solar panel A of the battery module 11 of a solar power generation prediction system 1 with a shading effect newly installed Place a photosensitive element L1, L2, L3, L4; (S2) place the battery module 11 outdoors to use a terminal T to collect a first set of illuminance data 22 of a plurality of shaded photosensitive elements and a first Set the illuminance data 23 of the unshaded photosensitive element; (S3) the terminal T stores all the data in a database 2; (S4) judge whether the collected data has reached a set first stage time, if yes, enter the step (S5), if not, enter step (S2); (S5) the terminal T uses a recursive fuzzy neural network 3 as a predictor to target the first group of unshaded objects in the database 2 Photosensitive element illuminance data 23 carries out the establishment of a mathematical model; (S6) produces a first group of solar power generation models 24 without shading effect, and the terminal machine T stores it into the database 2; (S7) the terminal machine T Utilize a fuzzy neural network 4 to adjust the first group of multiple types of shaded photosensitive element illuminance data 22, provide a first group of multiple types of weight value models W _TJ , and store them in the database 2; (S8 ) The terminal T multiplies the first group of solar power generation models 24 without shading effect and the first group of multiple weight value models W _TJ respectively, and after obtaining multiple products, a first average value can be calculated The maximum power generation MP1 is a first relevant parameter module 111 of the battery module 11 operating in the first stage of steady state to become a training prediction mathematical model 11B, and the training parameter module 11A is stored in the Database 2; (S9) place the battery module 11 outdoors to collect a second set of illuminance data 22' of a plurality of shaded photosensitive elements and a second set of illuminance data 23' of non-shaded photosensitive elements; (S10) The terminal machine T stores all data in the database 2; (S11) judges that the data collection has reached a set second stage time, if yes, enters step (S12), if not, enters step (S9 ); (S12) The terminal T uses the recursive fuzzy neural network 3 as a predictor to conduct another mathematical model for the second group of unshaded photosensitive element illuminance data 23' in the database 2 (S13) generate a second group of solar power generation models 24' without shading effect, and the terminal T stores it into the database 2; (S14) the terminal T utilizes the fuzzy neural network 4 , to adjust the second group of plural kinds of shaded photosensitive element illumination data 22', provide a second group of plural kinds of weight value models W _TJ ', and store them in the database 2; (S15) the terminal T will The second group of solar power generation models 24' without shading effect is multiplied by the second group of plural weight value models W _TJ ' respectively, and after obtaining the plural products, a second average maximum power generation MP2 can be calculated , for the battery module 11 is stable A second relevant parameter module 113 of a second stage of running directly on-line during the state operation, the second relevant parameter module 113 will continue to update, learn and optimize the training prediction mathematical model 11B, so that the training prediction The mathematical model 11B becomes a continuous training predictive mathematical model 11C, and stores the continuous training predictive mathematical model 11C in the database 2; and (S16) utilizes the continuous training predictive mathematical model 11C according to a third relevant parameter at that time The module 114 pre-evaluates a continuous training predicted maximum generating power module 19, and stores the continuous training predicted maximum generating power module 19 in the database 2. . Through the above steps, the continuous training prediction maximum generating power module 19 is the final result of the present invention, that is, the final predicted generating power.

，其中，m _ij 、σ _ij,n 、

、

為可調整之控制參數，σ _ij,L 為中心點在m _ij 之歸屬函數左側寬度參數，σ _ij,R 為中心點在m _ij 之歸屬函數右側寬度參數。 該模糊類神經網路(4)其數學方程式如下：：

，其中，σ _r 為中心點在I _rmax之歸屬函數寬度參數。 該第二階段時間為至少三至六個月。該第二相關參數模組113包括一第二系統相關參數模組1131與一第二環境相關參數模組1132。該第二系統相關參數模組1131包括電壓、電流、太陽能電池溫度、太陽能電池總幅照度/輻射量等資料。該第二環境相關參數模組1132包括大氣溫度、相對濕度、遮陰溫度、全天日射量、場址日照強度、風速、最大功率點追蹤(MPPT；Maximum power point tracking)後的太陽能電池輸出之電壓、電流、功率等資料。 該第三相關參數模組114包括一第三系統相關參數模組1141與一第三環境相關參數模組1142。該第三系統相關參數模組1141包括電壓、電流、太陽能電池溫度、太陽能電池總幅照度/輻射量等資料。該第三環境相關參數模組1142包括大氣溫度、相對濕度、遮陰溫度、降雨機率、全天日射量、場址日照強度、風速、最大功率點追蹤(MPPT；Maximum power point tracking)後的太陽能電池輸出之電壓、電流、功率等資料。 However, the duration of this first phase is at least three months. The first related parameter module 111 includes a first system related parameter module 1111 and a first environment related parameter module 1112 . The first system-related parameter module 1111 includes data such as voltage, current, solar cell temperature, and total irradiance/radiation of the solar cell. The first environment-related parameter module 1112 includes atmospheric temperature, relative humidity, shading temperature, solar radiation throughout the day, site sunlight intensity, wind speed, and output of solar cells after maximum power point tracking (MPPT; Maximum power point tracking). Voltage, current, power and other data; the mathematical equation of the recursive fuzzy neural network 3 is as follows:

, where m _ij , σ _ij,n ,

,

is an adjustable control parameter, σ _ij,L is the width parameter on the left side of the membership function with the center point at m _ij , and σ _ij,R is the width parameter at the right side of the membership function with the center point at m _ij . Its mathematical equation of this fuzzy class neural network (4) is as follows::

, where, σ _r is the width parameter of the membership function with the center point at I _{r max} . This second phase lasts at least three to six months. The second related parameter module 113 includes a second system related parameter module 1131 and a second environment related parameter module 1132 . The second system-related parameter module 1131 includes data such as voltage, current, solar cell temperature, and total irradiance/radiation of the solar cell. The second environment-related parameter module 1132 includes atmospheric temperature, relative humidity, shading temperature, solar radiation throughout the day, site sunlight intensity, wind speed, and output of solar cells after maximum power point tracking (MPPT; Maximum power point tracking). Voltage, current, power and other data. The third related parameter module 114 includes a third system related parameter module 1141 and a third environment related parameter module 1142 . The third system-related parameter module 1141 includes data such as voltage, current, solar cell temperature, and total irradiance/radiation of the solar cell. The third environment-related parameter module 1142 includes atmospheric temperature, relative humidity, shading temperature, rainfall probability, solar radiation throughout the day, site sunshine intensity, wind speed, and solar energy after maximum power point tracking (MPPT; Maximum power point tracking). Battery output voltage, current, power and other data.

Please refer to FIG. 5 , which is a corresponding weight ratio table of the illuminance data of the shading photosensitive element and the illuminance data of the non-shading photosensitive element in a solar power generation prediction method with shading effect of the present invention. The details are as follows: the terminal T calculates the corresponding weight value model (proportion) of the illuminance for the plurality of illuminance data of the illuminance of the shading photosensitive element in the database 2 and the illuminance data of the non-shading photosensitive element, that is, design A fuzzy neural network is used to make adjustments, so that the system provides multiple weight value models under multiple shading states to allow the system to provide a mathematical model under shading states, and proposes a method for shading judgments and weights. The fuzzy neural network structure of the seed-value model is used to achieve a plurality of weight-value models W _T0 ~ W _T15 , and store them in the database 2 . In terms of W _T0 , the value of each photosensitive element L1, L2, L3, and L4 is 0, which means that no shadow can be detected in each photosensitive element; in terms of W _T1 , the value of photosensitive element L4 is 1, means that the photosensitive element L4 has detected shadows. Therefore, the corresponding weight ratio of the illuminance data of the shading photosensitive element to the illuminance data of the non-shading photosensitive element is W _T1 . Certainly, when the value of each of the four photosensitive elements L1, L2, L3, and L4 of W _T15 in the bottom row is 1, it means that the situation of "no shadow can be detected by each photosensitive element" is wrong. What will happen, that is, "every photosensitive element detects the generation of shadow", that is, the state of "complete shading".

The description of the present invention is as above, but it is not intended to limit the scope of patent rights claimed by the present invention. The scope of its patent protection shall depend on the scope of the appended patent application and its equivalent fields. All changes or modifications made by those with common knowledge in the field without departing from the spirit or scope of this patent belong to the equivalent change or design completed under the spirit disclosed by the present invention, and should be included in the scope of the following patent application Inside.

S1-S17: Step number

Claims (9)

A method for predicting solar power generation with shading effect, comprising the following steps: (SA) continuously collecting at least one solar panel (A) of a battery module (11) of a newly installed solar power generation prediction system (1) with shading effect , a first group of multiple shaded photosensitive element illuminance data (22) and a first group of non-shaded photosensitive element illuminance data (23) collected during the first stage of steady-state operation, and using a recursive The fuzzy neural network (3) is a predictor, which produces a first group of solar power generation models (24) without shading effects for the first group of unshaded photosensitive element illuminance data (23), and uses a fuzzy class The neural network (4) adjusts the illuminance data (22) of the first group of plural kinds of photosensitive elements with shading, and provides a first group of plural kinds of weight value models (W _TJ ), and the first group of non-shading effect The solar power generation model (24) is multiplied respectively with the first group of multiple weight value models (W _TJ ), and after the multiple products obtained, a first average maximum power generation module (MP1) produced can be calculated, And derive a first relevant parameter module (111), to become a training prediction mathematical model (11B), and this training prediction mathematical model (11B) is stored in a database (2); (SB) continue to collect continuously The solar panel (A) of the battery module (11), a second group of plural kinds of illuminance data (22') collected by a second group of shaded photosensitive elements during a second stage of steady-state operation and a second group of unidentified Shading photosensitive element illuminance data (23'), and using the recursive fuzzy neural network (3) as a predictor to generate a second group of unshaded photosensitive element illuminance data (23') Solar power generation model (24') without shading effect, and using the fuzzy neural network (4) to adjust the illuminance data (22') of the second set of multiple kinds of photosensitive elements with shading to provide a second set Multiple weight value models (W _TJ '), the terminal T multiplies the second group of solar power generation models without shading effect (24') with the second group of multiple weight value models (W _TJ ') respectively , after the multiple products obtained, a second average maximum generating power module (MP2) generated can be calculated, and a second related parameter module (113) is derived, the second related parameter module (113) The training predictive mathematical model (11B) will be continuously updated, learned and optimized so that the training predictive mathematical model (11B) becomes a continuous training predictive mathematical model (11C), and the continuous training predictive mathematical model (11C) Store in the database (2); and (SC) use the continuous training prediction mathematical model (11C) to pre-evaluate a continuous training prediction maximum power generation module based on a third relevant parameter module (114) at that time (19), and deposit the continuous training prediction maximum generating power module (19) in the database (2). The solar power generation prediction method with shading effect as described in claim 1, wherein step (SA) further includes the following steps: (SA1) the battery module (11) of the solar power generation prediction system (1) with shading effect ) to place a photosensitive element (L1, L2, L3, L4) in each of the N positions of the solar panel (A), N≧1; (SA2) place the battery module (11) outdoors to utilize a terminal The machine (T) collects the illuminance data (22) of the first group of plural types of shading photosensitive elements and the illuminance data (23) of the first group of unshaded photosensitive elements; (SA3) the terminal (T) stores all the data In a database (2); (SA4) judge whether the data collection has reached a set first stage time, if yes, enter step (SA5), if not, enter step (SA3); (SA5) the terminal (T) Using a recursive fuzzy neural network (3) as a predictor to establish a mathematical model for the first group of unshaded photosensitive element illuminance data (23) in the database (2) ; (SA6) generate a first group of solar power generation models (24) without shading effect, and the terminal (T) stores it in the database (2); (SA7) the terminal (T) uses a fuzzy Neural network-like (4), adjust the first group of plural types of illuminance data of shaded photosensitive elements (22), provide a first group of plural types of weight value models (W _TJ ), and store them in the database ( 2); and (SA8) the terminal (T) multiplies the first group of solar power generation models (24) without shading effect and the first group of multiple weight value models (W _TJ ) respectively to obtain After multiple products, a first average maximum generating power (MP1) can be calculated, which is a first relevant parameter module (111) of the battery module (11) running in the first stage of steady state time, so as to become A training prediction mathematical model (11B), and storing the training prediction mathematical model (11B) into the database (2). The method for predicting solar power generation with shading effect as described in claim 2, wherein the first related parameter module (111) in step (SA8) includes a first system related parameter module (1111) and a first environment Related parameter module (1112). The solar power generation prediction method with shading effect as described in claim 3, wherein the first system-related parameter module (1111) includes any of the following data or a combination thereof: voltage, current, solar cell temperature, solar energy The total illuminance/radiation of the battery; the first environment-related parameter module (1112) includes any one of the following data or a combination thereof: atmospheric temperature, relative humidity, shading temperature, solar radiation throughout the day, and sunlight intensity at the site , wind speed and maximum power point tracking (MPPT; Maximum power point tracking) output voltage, current and power of the solar cell. The solar power generation prediction method with shading effect as described in claim 2, wherein the mathematical equation of the recursive fuzzy neural network (3) in step (SA5) is as follows:

, where m _ij , σ _ij,n ,

,

is an adjustable control parameter, σ _ij,L is the width parameter on the left side of the membership function with the center point at m _ij , and σ _ij,R is the width parameter at the right side of the membership function with the center point at m _ij . The method for predicting solar power generation with shading effect as described in Claim 2, wherein the mathematical equation of the fuzzy neural network (4) in step (SA7) is as follows:

, where, σ _r is the width parameter of the membership function with the center point at I _{r max} . The method for predicting solar power generation with shading effect as described in Claim 1, wherein step (SB) further includes the following steps: (SB1) placing the battery module (11) outdoors to collect the second set of A plurality of illuminance data of photosensitive elements with shading (22') and the second group of illuminance data of photosensitive elements without shading (23'); (SB2) the terminal (T) stores all the data in the database 2; (SB3) Judging whether the data collection has reached a set second stage time, if yes, enter step (SB4), if not, enter step (SB1); (SB4) the terminal (T) fuzzy with the recursive The neural network (3) is a predictor to establish a mathematical model again for the second group of unshaded photosensitive element illuminance data (23') in the database (2); (SB5) generate a first Two groups of solar power generation models (24') without shading effect, which are stored in the database (2) by the terminal (T); (SB6) the terminal T utilizes the fuzzy neural network 4 to The second group of plural kinds of shaded photosensitive element illumination data 22' is adjusted to provide a second group of plural kinds of weight model W _TJ ', and stored in the database 2; (SB7) the terminal (T) will The solar power generation model (24') of the second group without shading effect is multiplied respectively by the second group of multiple weight value models (W _TJ '), and after the multiple products obtained, a second average maximum The power generation (MP2) is a second relevant parameter module (113) for the battery module (11) to run directly online for a second stage of time during steady state operation, and the second relevant parameter module (113) will Continuously update, learn and optimize the training prediction mathematical model (11B), so that the training prediction mathematical model (11B) becomes a continuous training prediction mathematical model (11C), and store the continuous training prediction mathematical model (11C) and (SB8) using the continuous training prediction mathematical model (11C) to pre-evaluate a continuous training prediction maximum generating power module ( 19), and store the continuous training predicted maximum generating power module (19) in the database (2). The method for predicting solar power generation with shading effect as described in Claim 8, wherein the second related parameter module (113) of step (SB7) includes a second system related parameter module (1131) and a first environment Related parameter module (1132); the third related parameter module (114) of step (SB8) includes a third system related parameter module (1141) and a third environment related parameter module (1142). The method for predicting solar power generation with shading effect as described in Claim 8, wherein the second system-related parameter module (1131) includes any of the following data or a combination thereof: voltage, current, solar cell temperature, and The total illuminance/radiation of the solar cell; this second environment-related parameter The module (1132) includes any of the following data or a combination thereof: atmospheric temperature, relative humidity, shading temperature, solar radiation throughout the day, site sunlight intensity, wind speed and maximum power point tracking (MPPT; Maximum power point tracking ) output voltage, current and power of the solar cell; the third system-related parameter module (1141) includes any one of the following data or a combination thereof: voltage, current, solar cell temperature, and total irradiance of the solar cell / radiation amount; the third environment-related parameter module (1142) includes any one of the following data or a combination thereof: atmospheric temperature, relative humidity, shading temperature, rainfall probability, solar radiation throughout the day, site sunlight intensity, Wind speed and maximum power point tracking (MPPT; Maximum power point tracking) output voltage, current and power of solar cells.

Images