WO2019205528A1

WO2019205528A1 - Antenna data-based algorithm for predicting solar energy collection power of wireless sensor network node

Info

Publication number: WO2019205528A1
Application number: PCT/CN2018/111550
Authority: WO
Inventors: 孙力娟; 任恒毅; 韩崇; 郭剑; 肖甫; 王娟; 周剑; 王汝传
Original assignee: 南京邮电大学
Priority date: 2018-04-26
Filing date: 2018-10-24
Publication date: 2019-10-31
Also published as: KR102439722B1; KR20210013565A; CN108573327B; CN108573327A

Abstract

Provided in the present invention is an antenna data-based algorithm for predicting the solar energy collection power of a wireless sensor network node, which may accurately predict the collection power of a wireless sensor network node that has a solar energy collection function while using a smaller space and shorter time. The present discovery fully considers the effect of changes in weather on predicting solar energy collection power; by means of analyzing the comparison between changes in weather and solar energy collection power, the acting relationship of changes in weather on solar energy collection is discovered, thereby accurately predicting the solar energy collection power of a node when the weather changes significantly. At the same time, the present method may achieve extremely high prediction precision both when the weather changes drastically and when the weather changes smoothly.

Description

Solar Energy Collecting Power Prediction Algorithm for Wireless Sensor Network Node Based on Weather Data

Technical field

The invention relates to a wireless sensor network node solar energy collection power prediction algorithm based on weather data, and belongs to the field of wireless sensor networks.

Background technique

Wireless Sensor Network (WSN) is a wireless network composed of a large number of stationary or mobile sensors in a self-organizing and multi-hop manner. Due to their low cost and high maneuverability, WSNs have become more and more popular in the past, and they can complete many application scenarios that traditional cable or wired networks cannot handle, such as smart home, radiation monitoring, microscopic observation of biomes and intelligent transportation. The basic composition of the WSN is the sensor node, and the researchers distribute it in the monitoring area by means of fixed point placement or random throwing.

The conventional sensor node mainly includes four modules of data acquisition, data processing, data transmission and power supply, wherein the power module is responsible for providing energy to the other three modules. Power management of resource-constrained embedded systems remains a serious challenge in the face of a mismatch in growth between batteries and other embedded system components. Due to the remote scene deployment of the WSN and the lack of reliable power, sensors in the network typically rely on batteries to provide energy to achieve their intended tasks. Battery-powered WSNs, regardless of their energy efficiency, will eventually cause the network to fail due to limited power supply unless the battery is replaced. If the network is deployed in a harsh environment or an unreachable scenario, battery replacement is even more expensive and can't even be realized. A promising solution to this problem is to use environmental energy harvesting technology with batteries. The common environmental energy that can be collected and utilized is solar energy, vibration energy, wind energy, noise, and the like. Solar energy is widely used because of its high energy density and easy collection.

The introduction of energy harvesting capabilities into wireless sensor networks will cause many design issues regarding the construction of such systems, such as energy prediction, energy management, and routing protocols. Among them, energy prediction is the premise of other problem design, because whether it is energy management or routing protocol, energy consumption must be considered in design, that is, its performance level is specified according to the available energy. Since the available energy of the WSN with energy harvesting technology includes the remaining energy and the environmental energy collected in a certain period of time, in order to know the available energy, it must be able to predict the environmental energy collected in a certain period of time in the future.

Taking WSN energy management with solar energy collection capability as an example, WSN with environmental energy harvesting technology can theoretically continue to operate because environmental energy harvesting technology can collect environmental energy to supplement sensor consumption. However, it has been shown that it is difficult to ensure the long-term uninterrupted operation of the WSN by the sensor power supply device with energy harvesting capability. The main reason is the dependence on uncontrollable solar energy. Solar energy is difficult to simulate and predict due to meteorological factors. Time and space changes make it exhibit high short-term fluctuations. The uncertainty of the amount of energy radiated by the solar energy to the surface causes the uncertainty of the energy collected by the node, making it difficult for energy management of the WSN to achieve energy neutral operation (ENO), that is, the energy consumption of the system is less than or equal to the environment within a certain period of time. The energy collected. WSN usually uses highly variable environmental energy by dynamically adjusting the performance level of the system at runtime. If the system is running at the lowest performance level, ENO can be realized, but energy is wasted, and the system is difficult at this time. Meet most of the mission requirements; if the system is running at a higher performance level, it is likely that the energy collected by the environment is less than the system energy consumption, and the network life is terminated prematurely. Therefore, reasonable energy management is an important condition for WSN to operate permanently. The premise of energy management is to know the available energy of the system, that is, to predict the environmental energy collected in a certain period of time in the future.

Currently, for solar energy, many researchers have proposed prediction algorithms such as EWMA, WCMA and Pro-Energy. It is mainly the analysis of historical data collected to predict future short-term collection power. In the scene with stable weather, the above algorithm can control the prediction error to a small extent. In particular, WCMA can adapt to the simple change of the weather, but when the weather changes greatly, the above algorithm is difficult to ensure the accuracy of the prediction. Sex. It is well known that the vast majority of scenes are very common in weather conditions during the day, so it is necessary to consider real-time weather changes for solar energy prediction algorithms.

Summary of the invention

The invention provides a solar energy collection node solar energy collection prediction algorithm based on weather data, which can accurately predict the collection power of a wireless sensor network node with a solar energy collection function while using a small space and time. And it realizes the accurate prediction of the node solar energy collection in the scene with significant weather changes, and at the same time can achieve very high prediction accuracy in the scenes with significant weather changes and smooth weather changes.

The solar data collection power prediction algorithm of the wireless sensor network node based on the weather data according to the present invention, the specific execution steps are as follows:

Step 1: Select the real-time weather type data and solar energy collection data of the past 5 years in the monitoring area, and divide the weather type into 6 types, which are sunny (C), sunny (PC), and cloud (SC). Cloudy (MCI), cloudy (O), rain and snow (RS); statistically process the solar energy collected in different time slots of different weather types, and map the magnification relationship to the weather according to the magnification relationship between the weather data. Type value index W, proceeds to step 2;

Step 2: Obtain the actual weather forecast data of the day W(d+1,i) (i=1,2...24), and predict the weather forecast data W(d+1,i) of the day and the weather of the previous day. The prediction data W(d, i) is compared, and the weather change coefficient WCS of the two days before and after is obtained, and the process proceeds to step 3;

Step 3: w _t is a threshold value reflecting whether the weather has changed significantly. The size is obtained by analyzing the historical real-time weather type data and the collected power data of the node, and comparing WCS with w _t ; if |WCS|<w _t , it is considered The forecasting day is basically the same as the weather of the previous day, and directly jumps to step 4; if |WCS|≥w _t , it is considered that the weather of the forecast day and the previous day changes significantly, and jumps directly to step 5;

Step 4: First consider the case where the weather of the day is basically the same as the previous day, and adopt the idea of WCMA algorithm; M _D is the average value of the energy collection value of the ith time slot of the D day in front of the wireless sensor network node, N=[n ₁ , n ₂ ,,, n _k ] is a 1×k matrix that records the weather changes of the first k time slots of the time slot to be predicted, where n _j is calculated;

V=[v ₁ ,v ₂ ,,,v _k ] is used to reflect the weight, where v _j is calculated by calculation, and finally calculates the degree of weather change reflecting the predicted day and the previous day (by historical collection of predicted locations) The power data is determined); at this time, the solar energy acquisition prediction value can be directly calculated, and then jumps to step 7;

Step 5: When predicting a significant change in the weather between the day and the previous day, use the actual weather forecast data for the day of the forecast to correct the prediction error; WF = [wf ₁ , wf ₂ , ,, wf _k ] is a 1 × k matrix, To store the time slots to be predicted and the weather changes of the previous k-1 time slots, where wf _j is calculated by the weather forecast data, and finally calculate the time slot i and the previous k-1 times when the weather changes significantly. The degree of weather change of the time slot, proceeds to step 6;

Step 6: When the weather changes significantly, it is divided into two cases, namely, the weather is significantly better and the weather is significantly worse; the effects of the two cases on the solar energy collection are corrected by the constants a and b; finally, the weather is calculated. The solar energy collection power P(d+1, i) predicted by the i-th time slot on the d+1th day when a significant change occurs, proceeds to step 7;

Step 7: Feed the prediction results back to the sensor node for energy management.

Further, in the step 2, the calculation formula of the weather change coefficient WCS of the two days is:

Where W(d,i) is the weather type index of the i-th time slot on day d.

Further, in the step 4, the WCMA algorithm is described by comparing the change in the energy value of the previous time slot of the current day and the value of the energy collection value M _D of the corresponding time slot of the previous D day to reflect the change of the weather condition. :

among them,

For the predicted value of the solar energy collected in the ith time slot on day d, E(d, i) is the actual observed value of the solar energy collected in the ith time slot on day d, α is the weighting factor, and α ∈ [0, 1]; M _D (d, i) is the average value of the solar energy collected in the previous D-day time slot i, and GAP _y (i) refers to the weather change of the corresponding time slot of the current time slot and the past D days, and the calculation formula is as follows:

The V in the above formula is the weight, and the larger j is, the larger the weight v _j will be, as follows:

V=[v ₁ ,v ₂ ,,,v _k ]

N is a 1 × k matrix that records the weather changes of the first k time slots of the time slot to be predicted, as follows:

N=[n ₁ ,n ₂ ,,,n _k ]

In addition, n _{j in} step 4 is calculated by:

Further, in the steps 5 and 6, the GAP _x (i) is defined as the weather change of the corresponding time slot of the day slot and the past D day, and τ is a significant change of the weather, and τ>0 is defined as the weather significant change. The solar energy collection power is greatly increased, and τ<0 is a significant deterioration in the weather. The solar energy collection is greatly reduced, as shown in the following equation:

By setting the parameters a, b to constrain the influence of different weather changes on the solar energy collection, the parameters a, b should be determined according to the actual weather conditions and solar energy collection; WC (i) is the weather two days before and after The change is as follows:

WF is a 1×k matrix, which is used to store the time slots to be predicted and the weather changes of the previous k-1 time slots. The specific calculation of wf ₁ , wf ₂ ···wf _k is as follows:

WF=[wf ₁ ,wf ₂ ,,,wf _k ]

In addition, V is introduced as a 1×k matrix to represent the weight; wherein, the larger j is, the larger the weight v _j is, that is, the closer to the weather change of the time slot to be predicted, the more important Shown as follows:

V=[v ₁ ,v ₂ ,,,v _k ]

The WF is multiplied by V and divided by the sum of V, and the final result is determined as the weather change condition WC(i).

Further, in the step 6, the final prediction algorithm is as follows:

Where w _t is a threshold reflecting whether the weather has changed significantly, and the size of w _t is obtained by analyzing historical real-time weather type data and node collected power data.

Beneficial effect

The invention proposes a solar sensor network node solar energy collection prediction algorithm based on weather data. The method can accurately predict the collected power of the wireless sensor network node with solar energy collection function while using less space and time.

This finding fully considers the impact of weather changes on solar energy harvesting power. By comparing and analyzing the weather changes and solar energy collection, the relationship between weather changes and solar energy harvesting is found, thus achieving accurate prediction of significant weather changes. Node solar energy is collected. At the same time, the method can achieve very high prediction accuracy in the case of significant weather changes and stable weather changes.

DRAWINGS

Figure 1 is a comparison of collected power for 24 time slots in winter.

Figure 2 shows the correspondence between weather type indices for different weather types in different time slots in winter.

Figure 3 shows the energy curve of solar energy acquisition predicted by EWMA, WCMA algorithm.

Figure 4 is a flow chart of the algorithm for predicting solar energy collection.

detailed description

The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings.

The invention aims at the energy management efficiency of the existing wireless sensor network node with solar energy collection function is too low, and the conventional solar energy acquisition prediction algorithm cannot adapt to the problem of scenes with frequent weather changes, and proposes to introduce weather data of weather forecast to Solar energy harvesting power prediction algorithm. The method firstly analyzes the relationship between the weather change and the collected power of the node, and quantifies the influence of the weather on the collected power. Secondly, it divides the weather change into three kinds: no significant change, the weather is significantly better, and the weather is significantly worse. The collected power predictions for each case are discussed separately. The method not only can effectively solve the influence of weather changes on the prediction results, improve the accuracy of the prediction results, and improve the applicability of the algorithm. It can achieve good prediction results in different climate characteristics.

The method is mainly divided into three parts. One is to analyze the relationship between different weather types and solar energy collection; the second is to calculate and analyze the weather changes on the day and the previous day; the third is to calculate the solar energy collected according to different weather changes. Predictive value.

(1) Analysis of the relationship between different weather types and solar energy collection

According to the analysis of historical data, the degree of influence of weather patterns in different seasons on solar energy collection is also different. According to the season as the standard, the data is divided into four parts: spring, summer, autumn and winter. These influencing factors are mapped to the weather type index to reflect the influence of weather type on solar energy collection.

Select real-time weather type data and node collection power data of sensor nodes in the past five years in a scene. First, the weather types are divided into six types: sunny (C), sunny (PC), cloudy (SC), cloudy (MCI), cloudy (O), and rain (RS). The amount of collection in different time periods of the day is significantly different, and the day is divided into 24 time slots with 1 hour as 1 time slot. Secondly, the solar energy collection power of different time slots of different weather types is statistically processed (Fig. 1 shows the comparison of the collected power of 24 time slots in winter). Finally, according to the magnification relationship between the weather data, the magnification relationship is mapped to the weather type index. It can be concluded from Fig. 1 that except for sunrise and sunset, the relationship between the collection amount of the other time slots and the weather type is very clear, so the weather type index of the 11-17 time slot is the magnification relationship between the weather data, and the rest The weather type index of the gap is the mean of all time slots. Due to the large amount of data on cloudy days, the weather type index W is set to unit 1, and the rest of the weather type index is equivalently converted on a cloudy basis. Figure 2 shows the different weather types of different time slots in a certain winter scene. Weather type index correspondence.

(2) Calculate and analyze the weather changes predicted on the day and the previous day

The EWMA algorithm has higher accuracy when the weather in the predicted location is basically unchanged, and the WCMA algorithm corrects the defect that the EWMA algorithm cannot adapt to the weather change, so that the algorithm can predict the weather change. With high precision. However, when the weather in the predicted location changes drastically, such as when the cloudy day turns sunny, WCMA still has a higher error. The idea of the present invention is based on WCMA, which introduces real-time weather on the network to improve the defect that EWMA and WCMA cannot adapt to the significant change of weather in the predicted location.

Due to the changing weather conditions, the calculation of the weather adjustment factor GAP(i) should be considered in many aspects. If the forecast is basically the same, significant change, and drastic change compared with the previous day's weather conditions, the solar energy collection will change differently, and the weather conditions are not a simple linear relationship with the solar power collection power, so it is necessary to consider the situation. . On this basis, it is necessary to consider how to discern the changes in the weather conditions on the day of the forecast and the previous day. Through the analysis of historical data, it is found that the main factors that determine the quality of the day's weather conditions or the level of solar energy collection are the weather conditions at noon and afternoon. If the weather conditions at noon and afternoon are sunny, then the solar energy collected on the day will be higher; on the contrary, when the weather conditions at noon and afternoon are cloudy or rainy, the solar energy collected on that day will be It will be very low. Therefore, a new variable WCS is introduced to reflect the weather changes on the day of the forecast and the previous day, as shown in equation (1):

Since the latitude and longitude of different scenes are different, the value of i in equation (1) is also uncertain. W(d,i) is the weather type index of the i-th time slot on day d.

(3) Calculate the predicted value of solar energy collection according to different weather changes:

After distinguishing between different weather changes, you need to consider the specific calculation method for each change.

Figure 3 shows the EWMA, WCMA algorithm predicted energy curve and daily mean error (1 hour per time slot). The analysis shows that the WCMA algorithm reduces the prediction error compared with the EWMA algorithm, but there is still a large error when the weather changes significantly. Through a large number of experiments on the WCMA algorithm (the weight α of the experiment is fixed at 0.5), it is found that the prediction error of the algorithm for different weather changes is different. By comparison, it is found that the weather from sunny to cloudy weather has a very large prediction error; the weather from cloudy to sunny is more predictive; and the current two days are basically the same, the prediction error is very small. Therefore, considering the classification of different weather changes, there are two main categories, that is, the weather changes significantly and the weather does not change significantly, which is represented by GAP _x and GAP _y respectively.

① first discuss the more complex cases, i.e. the case when a significant change in the weather occurs GAP _x how to calculate. According to the above analysis, the changes of the two days before and after are subdivided into two types, namely, the weather changes drastically, the solar energy collection increases greatly, and the weather changes drastically. For the convenience of description, the definition of τ is a significant change in the weather, and the definition of τ>0 is that the weather is significantly better, the solar energy collection is greatly increased, and τ<0 is a significant deterioration of the solar energy collected by the weather. As shown in formula (2).

By setting the parameters a, b to constrain the influence of different weather changes on the solar energy collection, the parameters a, b should be determined according to the actual weather conditions and solar energy collection. WC(i) is the weather change for the first two days, as shown in formula (3).

WF is a 1×k matrix, which is used to store the time interval of the time slot to be predicted and the previous k-1 time slots. The calculation of specific wf ₁ , wf ₂ , , and wf _k is as shown in formula (4). ~ (5).

WF=[wf ₁ ,wf ₂ ,,,wf _k ] (4)

When the weather condition of the time slot to be predicted is reflected by the time slot of the time slot to be predicted and the weather condition of the previous k-1 time slots, there is another problem to be considered. wf ₁ , wf ₂ , , , wf _k pairs will be predicted The weather conditions of the time slots are not the same. The closer the weather changes to the time slots to be predicted, the more important it is, so the concept of weights needs to be introduced. V is also a 1 × k matrix, which is used to reflect the weight. Where j is larger, then the weight v _j will be larger, that is, the closer to the weather change of the time slot to be predicted, the more important, as shown in the formulas (6) to (7).

V=[v ₁ ,v ₂ ,,,v _k ] (6)

2 Next, consider the case where the weather does not change significantly, that is, the calculation of GAP _y . When the weather conditions in the last two days are sunny or cloudy, the energy value collected by the nodes will not change greatly, but this does not mean that it will not change, because it is also sunny, but also because of temperature, humidity and other factors. The small difference is generated. At this time, the idea of the WCMA algorithm is selected, and the change in the weather condition is reflected by comparing the energy value of the previous time slot of the current day and the value of the average value M _D of the energy collection value of the corresponding time slot of the previous D day. The WCMA algorithm is described as follows:

among them,

For the predicted value of the solar energy collected in the ith time slot on day d, E(d, i) is the actual observed value of the solar energy collected in the ith time slot on day d, α is the weighting factor, and α ∈ [0, 1]. M _D (d, i) is the average value of the solar energy collected in the previous D-day time slot i, and GAP _y (i) refers to the weather change of the corresponding time slot of the current time slot and the past D days. Specifically, it is as shown in formulas (9) to (10).

The weight in the formula (13) is the weight, and the larger the j, the larger the weight v _j , as in the formulas (6) to (7). N is a 1×k matrix that records the weather changes of the first k time slots of the time slot to be predicted, as in equations (11)–(12).

N=[n ₁ ,n ₂ ,,,n _k ] (11)

Finally, our prediction algorithm is as in formula (13):

This method is mainly to predict the specific process of solar energy collection.

Predicting solar energy collection: The solar energy collection power flow chart is shown in Figure 4. The specific steps are as follows:

Step 1: Select the real-time weather type data and solar energy collection data of the past 5 years in the monitoring area, and divide the weather type into 6 types, which are sunny (C), sunny (PC), and cloud (SC). Cloudy (MCI), cloudy (O), rain and snow (RS). The solar energy collection power of different time slots of different weather types is statistically processed, and the magnification relationship is mapped to the weather type numerical index W according to the magnification relationship between the weather data, and the process proceeds to step 2.

Step 2: Obtain the actual weather forecast data W(d+1,i) (i=1,2...24) for the day, and use the formula (1) to predict the weather forecast data W(d+1,i) for the day. Compared with the weather forecast data W(d, i) of the previous day, the weather change coefficient WCS of the two days before and after is obtained, and the process proceeds to step 3.

Step 3: w _t is a threshold that reflects whether the weather has changed significantly. The size is obtained by analyzing historical real-time weather type data and node collected power data, and comparing WCS with w _t . If |WCS|<w _t , it is considered that the weather on the day of the forecast is basically the same as that of the previous day, and jumps directly to step 4; if |WCS|≥w _t , it is considered that the weather on the day of the forecast and the previous day changes significantly, and jumps directly. Step 5.

Step 4: Consider the case where the weather is basically the same as the previous day, and adopt the idea of WCMA algorithm. As shown in equation (9), M _D is the average value of the energy acquisition values of the ith time slot of the D-day in front of the wireless sensor network node, and N=[n ₁ , n ₂ ,,, n _k ] is a 1×k matrix. , the weather change of the first k time slots of the time slot to be predicted is recorded, where n _j is calculated by the formula (12). V=[v ₁ ,v ₂ ,,,v _k ] is used to reflect the weight, where v _j is calculated by the formula (7), and finally the weather change reflecting the predicted day and the previous day is calculated by the formula (10). Degree (determined by historically collecting power data from predicted locations). At this time, the solar energy acquisition prediction value can be directly calculated using the formula (5), and then jumps to step 7.

Step 5: When predicting a significant change in the weather between the day and the previous day, use the actual weather forecast data for the day of the forecast to correct the forecast error. WF=[wf ₁ ,wf ₂ ,,,wf _k ] is a 1×k matrix for storing the time slots to be predicted and the weather changes of the previous k-1 time slots, where wf _j is through the weather forecast data. To calculate, specifically as the formula (5), finally calculate the weather change degree of the time slot i and the previous k-1 time slots when the weather changes significantly by the formula (3), and proceeds to step 6.

Step 6: When the weather changes significantly, it is divided into two situations, namely, the weather is significantly better and the weather is significantly worse. The effect of these two cases on solar energy collection is different, using the constants a, b to correct, as shown in equation (2). Finally, the solar energy collection power P(d+1,i) predicted by the i-th time slot on the d+1th day when the weather changes significantly is calculated using the formula (13), and the process proceeds to step 7.

The above is only the preferred embodiment of the present invention, and the scope of the present invention is not limited to the above embodiments, but equivalent modifications or variations made by those skilled in the art according to the disclosure of the present invention should be incorporated. Within the scope of protection stated in the claims.

Claims

A wireless sensor network node solar energy collection prediction algorithm based on weather data, characterized in that the specific execution steps are as follows:

Step 1: Select the real-time weather type data and solar energy collection data of the past 5 years in the monitoring area, and divide the weather type into 6 types, which are sunny (C), sunny (PC), and cloud (SC). Cloudy (MCI), cloudy (O), rain and snow (RS); statistically process the solar energy collected in different time slots of different weather types, and map the magnification relationship to the weather according to the magnification relationship between the weather data. Type value index W, proceeds to step 2;

Step 2: Obtain the actual weather forecast data W(d+1,i) (i=1,2...24) of the day, and predict the weather forecast data W(d+1,i) of the day and the weather forecast data of the previous day. W(d, i) is compared, and the weather change coefficient WCS is obtained two days before and after, and the process proceeds to step 3;

Step 3: w t is a threshold value reflecting whether the weather has changed significantly. The size is obtained by analyzing the historical real-time weather type data and the collected power data of the node, and comparing WCS with w t ; if |WCS|<w t , it is considered The forecasting day is basically the same as the weather of the previous day, and directly jumps to step 4; if |WCS|≥w t , it is considered that the weather of the forecast day and the previous day changes significantly, and jumps directly to step 5;

Step 4: First consider the case where the weather of the day is basically the same as the previous day, and adopt the idea of WCMA algorithm; M D is the average value of the energy collection value of the ith time slot of the D day in front of the wireless sensor network node, N=[n 1 , n 2 ,,, n k ] is a 1×k matrix that records the weather changes of the first k time slots of the time slot to be predicted, where n j is calculated;

V=[v 1 ,v 2 ,,,v k ] is used to reflect the weight, where v j is calculated by calculation, and finally calculates the degree of weather change reflecting the predicted day and the previous day (by historical collection of predicted locations) The power data is determined); at this time, the solar energy acquisition prediction value can be directly calculated, and then jumps to step 7;

Step 5: When predicting a significant change in the weather between the day and the previous day, use the actual weather forecast data for the day of the forecast to correct the prediction error; WF = [wf 1 , wf 2 , ,, wf k ] is a 1 × k matrix, To store the time slots to be predicted and the weather changes of the previous k-1 time slots, where wf j is calculated by the weather forecast data, and finally calculate the time slot i and the previous k-1 times when the weather changes significantly. The degree of weather change of the time slot, proceeds to step 6;

Step 6: When the weather changes significantly, it is divided into two cases, namely, the weather is significantly better and the weather is significantly worse; the effects of the two cases on the solar energy collection are corrected by the constants a and b; finally, the weather is calculated. The solar energy collection power P(d+1, i) predicted by the i-th time slot on the d+1th day when a significant change occurs, proceeds to step 7;

Step 7: Feed the prediction results back to the sensor node for energy management.
The weather data-based wireless sensor network node solar energy collection power prediction algorithm according to claim 1, wherein in the step 2, the calculation formula of the weather change coefficient WCS of the two days is:

Where W(d,i) is the weather type index of the i-th time slot on day d.
The weather data-based wireless sensor network node solar energy collection power prediction algorithm according to claim 1, wherein in the step 4, by comparing and predicting the energy collection value of the previous time slot of the day corresponding to the previous D day, The change in the mean value of the energy collection value of the gap, M D , reflects the change in weather conditions. The WCMA algorithm is described as follows:

among them,
For the predicted value of the solar energy collected in the ith time slot on day d, E(d, i) is the actual observed value of the solar energy collected in the ith time slot on day d, α is the weighting factor, and α ∈ [0, 1]; M D (d, i) is the average value of the solar energy collected in the previous D-day time slot i, and GAP y (i) refers to the weather change of the corresponding time slot of the current time slot and the past D days, and the calculation formula is as follows:

The V in the above formula is the weight, and the larger j is, the larger the weight v j will be, as follows:

V=[v 1 ,v 2 ,,,v k ]

N is a 1 × k matrix that records the weather changes of the first k time slots of the time slot to be predicted, as follows:

N=[n 1 ,n 2 ,,,n k ]

In addition, n j in step 4 is calculated by:
The weather data-based wireless sensor network node solar energy collection power prediction algorithm according to claim 1, wherein in the steps 5 and 6, the GAP x (i) is defined as a day time slot and a past D day. The weather change of the time slot, τ is a significant change in the weather, and the definition of τ>0 is that the weather is significantly better, the solar energy collection is greatly increased, and τ<0 is the significant deterioration of the solar energy collected by the weather, as shown in the following formula:

By setting the parameters a, b to constrain the influence of different weather changes on the solar energy collection, the parameters a, b should be determined according to the actual weather conditions and solar energy collection; WC (i) is the weather two days before and after The change is as follows:

WF is a 1×k matrix, which is used to store the time slots to be predicted and the weather changes of the previous k-1 time slots. The specific calculation of wf 1 , wf 2 ···wf k is as follows:

WF=[wf 1 ,wf 2 ,,,wf k ]

In addition, V is introduced as a 1×k matrix to represent the weight; wherein, the larger j is, the larger the weight v j is, that is, the closer to the weather change of the time slot to be predicted, the more important Shown as follows:

V=[v 1 ,v 2 ,,,v k ]

The WF is multiplied by V and divided by the sum of V, and the final result is determined as the weather change condition WC(i).
The weather data-based wireless sensor network node solar energy collection power prediction algorithm according to claim 1, wherein in the step 6, the final prediction algorithm is as follows:

Where w t is a threshold reflecting whether the weather has changed significantly, and the size of w t is obtained by analyzing historical real-time weather type data and node collected power data.