SEHWS energy prediction method based on double exponential moving average and SEH wireless sensor

Abstract

This invention provides a SEHWS energy prediction method based on double exponential moving average and an SEH wireless sensor, belonging to the field of energy prediction technology for solar energy harvesting wireless sensors. The method includes: collecting and processing historical energy harvesting data, and constructing historical... D Daily Energy Matrix E ; Parameters in the calculation model μ D ( d , n ), B ( d , n ), ω and γ Based on historical data, four parameters are substituted into the DEMA energy prediction model, and the energy collected in the previous time slot is compared with the energy collected in the current time slot. n The weighted calculation is performed by taking the average of the actual energy collected over the past few days, and the result is... n Predicted value of time slot harvesting energy E(—)( d , n The SEH wireless sensor consists of a microcontroller, a data acquisition module, a solar energy collection module, a memory, a wireless communication module, and a power management module.

Description

Technical Field

[0001] This invention relates to the field of energy prediction technology for solar energy harvesting wireless sensors, and in particular to an energy prediction method based on double exponential moving average (SEHWS) and an SEH wireless sensor. Background Technology

[0002] With the rapid development of IoT technology, an increasing number of devices and systems require interconnectivity, creating a huge demand for wireless sensors. Wireless sensor networks (WSNs) are widely used in environmental monitoring, human health detection, and public safety. They consist of multiple wireless sensor nodes distributed within a monitoring area. These nodes sense physical or chemical quantities in the environment, perform preliminary data processing, and then transmit the data to a aggregation node via wireless communication protocols, using single-hop or multi-hop methods. The aggregation node then transmits the data to a central server or cloud platform for centralized processing and storage. Throughout this process, wireless sensors typically rely on battery power, but traditional batteries have limited capacity, restricting node lifespan. When the battery runs out, the wireless sensor node fails and needs battery replacement. However, because many wireless sensor nodes are deployed in harsh or sparsely populated areas, battery replacement is extremely costly, and in some cases, impossible, which has become a major limiting factor in the development of wireless sensor networks.

[0003] To address this issue, researchers have proposed introducing Solar Energy Harvesting (SEH) technology into wireless sensors. By harvesting solar energy from the environment and converting it into electrical energy to power wireless sensor nodes, the wireless sensor network becomes self-powered and its lifespan is significantly extended. In this way, sensor nodes can operate continuously without battery replacement, preventing node failure. Wireless sensor nodes powered by solar energy harvesting technology are called Solar Energy Harvesting Wireless Sensors (SEHWS). During the solar energy harvesting process, energy prediction is a necessary prerequisite to ensure the rational allocation of energy by the solar energy harvesting wireless sensor nodes.

[0004] Existing energy prediction technologies mainly focus on prediction algorithms based on machine learning and deep learning models, as well as data flow-based prediction methods. Since EHWS is a strictly resource-constrained system, its limitations in computing power and storage space place more stringent demands on prediction methods. While prediction algorithms based on machine learning and deep learning models offer advantages in predictive performance and achieve higher accuracy through processing and training on large amounts of data, they require significant computing resources, have poor interpretability, and lead to higher energy consumption. Many researchers have proposed data flow-based prediction methods and algorithms, such as EWMA, WCMA, and Pro-Energy.

[0005] EWMA performs well in relatively stable weather conditions, but it is susceptible to weather changes. WCMA introduces a GAP factor to reduce prediction errors caused by weather variations. Pro-Energy considers a set of energy harvesting models describing different weather conditions (cloudy, sunny, or rainy) and combines the latest measurements with the weather profile closest to the current energy distribution to generate predicted energy values, further improving prediction accuracy. However, these methods are inadequate in capturing energy trend changes when facing environments with drastic or trend-driven solar energy fluctuations. They are prone to lag effects and insufficient sensitivity to trends when a sustained increase or decrease occurs, resulting in lower prediction accuracy. Summary of the Invention

[0006] In view of this, the present invention provides an SEHWS energy prediction method and an SEH wireless sensor based on double exponential moving average. The double exponential moving average prediction method is used to predict the energy collected by the solar energy harvesting wireless sensor, which can respond more quickly to fluctuations in environmental energy, improve prediction accuracy, and ensure the stable operation of the wireless sensor node.

[0007] The technical solution adopted by the embodiments of the present invention to solve its technical problem is as follows:

[0008] A SEHWS energy prediction method based on double exponential moving average includes:

[0009] Step S1: Collect and process historical energy collection data. Construct a historical D-day energy matrix E based on the current day d. The size of E is D×N, where D represents the number of historical days and N represents the total number of time slots in a day. In matrix E, E(i,n) represents the energy value collected in the nth time slot of the i-th day. Let E(d,n) represent the energy value collected in the nth time slot of the current day d.

[0010] Step S2, establish the DEMA energy prediction model:

[0011] ;

[0012] In the formula, (d,n) is the predicted energy value collected in the nth time slot of the current day d; ω is the weighting factor used to control the recent energy data, ω∈(0,1); (d,n-1) is the energy information predicted and collected in the previous time slot; μ(d,n) represents the average value of the real energy information E(di,n) collected in the nth time slot of the past D days of the current day d, i∈[1,D]; B(d,n) is the correction value of the energy change trend in the nth time slot of the current day d;

[0013] Step S3: Calculate the parameter μ in the DEMA energy prediction model. D (d,n), B(d,n), ω and γ, where:

[0014] μ D The formula for calculating (d,n) is:

[0015] ;

[0016] In the formula, n is the current time slot, i is the index variable, i∈[1,D], and E is the matrix that stores the actual collected energy;

[0017] The correction factor B(d,n) is represented as the weighted value of the most recent energy change trend B(d,n-1) and the average energy change trend μ(d,n)-μ(d,n-1), and its expression is:

[0018] ;

[0019] In the formula, γ is a weighting factor used to adjust the ratio between B(d,n-1) and μ(d,n)-μ(d,n-1), γ∈(0,1), and the initial value of B(d,n) for iterative calculation is defined as B(0,0);

[0020] Update weighting factors ω and γ:

[0021] Based on the range of values ​​of weighting factors ω and γ, the values ​​of the two weighting factors are traversed, and the combination of the two weighting factors is brought into the DEMA energy prediction model for prediction. The mean absolute percentage error (MAPE) between the actual value and the predicted value is calculated, and the weighting factor combination with the lowest MAPE value is determined.

[0022] Step S4: Based on historical energy harvesting data, calculate the four parameters B(d,n), μ, and μ required for the DEMA energy prediction model according to step S3. D Substituting (d,n), ω, and γ into the DEMA energy prediction model, the predicted value (d,n) of the energy collected in time slot n is obtained by weighting the energy predicted in the previous time slot with the average value of the actual collected energy information in the past few days for the current time slot n.

[0023] Ideally, B(0,0) should be 0.

[0024] Preferably, the expression for the mean absolute percentage error (MAPE) is:

[0025] ;

[0026] An SEH wireless sensor includes a microcontroller, and a data acquisition module, a solar energy collection module, a memory, a wireless communication module, and a power management module connected to the microcontroller; wherein the memory stores instructions that can be executed by the microcontroller, and the instructions are executed by the microcontroller to enable the microcontroller to perform the methods shown in steps S1-S4.

[0027] Data acquisition module: used to acquire environmental parameters in real time through sensors, including temperature, humidity, light intensity, and air pressure;

[0028] Solar energy collection module: Composed of solar panels, used to convert collected sunlight into electrical energy;

[0029] Power management module: responsible for managing the electrical energy generated by the solar energy collection module, specifically including:

[0030] Charging management unit: Stores the electrical energy converted from solar energy into batteries;

[0031] Voltage regulation unit: Used to power various modules in the SEH wireless sensor and ensure stable output voltage;

[0032] Energy monitoring unit: Used to monitor the battery's charge and health status in real time;

[0033] Wireless communication module: used to transmit data collected by the SEH wireless sensor to the upper-level device via wireless signal transmission;

[0034] The memory is used to store historical energy harvesting data and the operating procedures required for sensor operation;

[0035] The microcontroller is used to process historical energy collection data and construct a historical D-day energy matrix E based on the current day d. The size of E is D×N, where N represents the total number of time slots in a day. In matrix E, E(i,n) represents the energy value collected in the nth time slot of the i-th day, and E(d,n) represents the energy value collected in the nth time slot of the current day d.

[0036] The microcontroller is used to establish the DEMA energy prediction model:

[0037] ;

[0038] In the formula, (d,n) is the predicted energy value collected in the nth time slot of the current day d; ω is the weighting factor used to control the recent energy data, ω∈(0,1); (d,n-1) represents the energy information collected in the previous time slot; μ(d,n) represents the average value of the real energy information E(di,n) collected in the nth time slot of the past D days of the current day d, i∈[1,D]; B(d,n) is the correction value of the energy change trend in the nth time slot of the current day d;

[0039] The microcontroller is used to calculate the parameter μ in the DEMA energy prediction model. D (d,n), B(d,n), ω and γ, where:

[0040] μ D The formula for calculating (d,n) is:

[0041] ;

[0042] In the formula, n is the current time slot, i is the index variable, and i∈[1,D];

[0043] The correction factor B(d,n) is represented as the weighted value of the most recent energy change trend B(d,n-1) and the average energy change trend μ(d,n)-μ(d,n-1), and its expression is:

[0044]

[0045] In the formula, γ is a weighting factor used to adjust the ratio between B(d,n-1) and μ(d,n)-μ(d,n-1), γ∈(0,1), and the initial value of B(d,n) for iterative calculation is defined as B(0,0);

[0046] Update weighting factors ω and γ:

[0047] Based on the range of values ​​of weighting factors ω and γ, the values ​​of the two weighting factors are traversed, and the combination of the two weighting factors is brought into the DEMA energy prediction model for prediction. The mean absolute percentage error (MAPE) between the actual value and the predicted value is calculated, and the weighting factor combination with the lowest MAPE value is determined.

[0048] The microcontroller is used to calculate the four parameters B(d,n), μ, and μ required for the DEMA energy prediction model based on historical energy harvesting data. D Substituting (d,n), ω, and γ into the DEMA energy prediction model, the predicted value (d,n) of the energy collected in time slot n is obtained by weighting the energy predicted in the previous time slot with the average value of the actual collected energy information in the past few days for the current time slot n.

[0049] Ideally, B(0,0) should be 0.

[0050] Preferably, the expression for the mean absolute percentage error (MAPE) is:

[0051] .

[0052] As can be seen from the above technical solution, the SEHWS energy prediction method and SEH wireless sensor based on double exponential moving average provided in this embodiment of the invention first collect and process historical energy collection data, construct a historical D-day energy matrix E based on the current day d, where E is of size D×N, N represents the total number of time slots equally divided in a day, and E(i,n) in matrix E represents the energy value collected in the nth time slot of the i-th day, and E(d,n) represents the energy value collected in the nth time slot of the current day d; then, a DEMA energy prediction model is established; and the parameter μ in the DEMA energy prediction model is calculated. D (d,n), B(d,n), ω, and γ; based on historical energy harvesting data, according to the four parameters B(d,n), μ, and γ required by the DEMA energy prediction model. D Substituting (d,n), ω, and γ into the DEMA energy prediction model, the predicted energy for time slot n is calculated by weighting the energy predicted in the previous time slot with the average of the actual energy collected in the current time slot n over the past few days. (d,n). This invention uses a double exponential moving average forecasting method to predict the energy collected by the solar energy harvesting wireless sensor, which can respond more quickly to fluctuations in environmental energy, improve forecast accuracy, and ensure the stable operation of the wireless sensor node. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the DEMA energy prediction model. Detailed Implementation

[0054] The technical solution and effects of the present invention will be further described in detail below with reference to the accompanying drawings.

[0055] The purpose of this invention is to provide a solar energy harvesting wireless sensor energy prediction method that can reflect the energy change trend collected by the solar energy wireless sensor. The core idea of ​​the method is to perform exponential smoothing on the data twice, introduce a correction factor, and adjust the weight between the correction value of the previous time slot and the average energy difference to further smooth the prediction results, so as to better capture the trends and changes in the data, respond more quickly to the fluctuations in environmental energy, reduce prediction errors, and ensure that the wireless sensor node can operate stably under the condition of energy changes, making the prediction more sensitive, accurate and timely.

[0056] This invention provides a SEHWS energy prediction method based on double exponential moving average, the steps of which include:

[0057] Step S1: Collect and process historical energy collection data. Construct a historical D-day energy matrix E based on the current day d. The size of E is D×N, where D represents the number of historical energy observation days and N represents the total number of time slots in a day. In matrix E, E(i,n) represents the energy value collected in the nth time slot of the i-th day, and E(d,n) represents the energy value collected in the nth time slot of the current day d.

[0058] Before forecasting, each day is divided into N time slots, such as choosing to divide it into half-hour slots (N=48). This transforms the time series data into a two-dimensional matrix for subsequent processing and analysis. Figure 1 As shown;

[0059] Step S2, establish the DEMA energy prediction model. The prediction formula for DEMA is shown in equation (1):

[0060] (1)

[0061] In the formula, (d,n) is the energy prediction value collected in the nth time slot of the current day d; ω is the weighting factor used to control the recent energy data, ω∈(0,1). The data is weighted by the two weighting factors ω and (1-ω). This weighted average is updated with each new time slot. Its value is set between 0 and 1. The possible values ​​are traversed and appropriate values ​​are selected to improve the prediction accuracy. The most recent data point (current time slot) will have a larger weight, while the weight of earlier data points (previous time slot) will gradually decrease. (d,n-1) represents the energy information collected in the previous time slot; μ(d,n) represents the average value of the real energy information E(di,n) collected in the nth time slot of the past D days on the current day d, i∈[1,D]; B(d,n) is the correction value of the energy change trend in the nth time slot of the current day d, which is the weighted value of the most recent energy change trend and the average energy change trend;

[0062] Step S3: Calculate the parameter μ in the DEMA energy prediction model. D (d,n), B(d,n), ω and γ, where:

[0063] 3.1) Calculate the average value μ of the actual collected energy information of the current time slot over the past D days. D (d,n), the calculation formula is:

[0064] (2)

[0065] In the formula, d represents the current date, n represents the current time slot, D represents the number of days for energy observation, i is the index variable for traversing the past D days (i∈[1,D]), and E is the matrix storing the actual collected energy. The energy data from the past D days in time slot n are obtained, stored in the energy matrix E, and the average value of these data is calculated. This yields the average energy collection situation for the current time slot over the past D days. As time progresses, the calculation range of the average value will continuously expand.

[0066] 3.2) The correction factor B(d,n) is represented as the weighted value of the most recent energy change trend B(d,n-1) and the average energy change trend μ(d,n)-μ(d,n-1), and its expression is:

[0067] (3)

[0068] In the formula, γ is a weighting factor used to adjust the ratio between B(d,n-1) and μ(d,n)-μ(d,n-1), γ∈(0,1), and the weighting will be updated over time; the initial value of B(d,n) for iterative calculation is defined as B(0,0); B(0,0) takes the value of 0; the difference between the average energy difference in the current period and the previous period is calculated, and then the weighted average of it and the correction value of the previous period is taken to obtain the correction factor for the current period. The correction factor reflects the trend of the average energy change.

[0069] 3.3) Update weighting factors ω and γ:

[0070] Based on the range of values ​​of weighting factors ω and γ, the combination of the two weighting factors is traversed and then fed into the DEMA energy prediction model for prediction. The mean absolute percentage error (MAPE) between the actual value and the predicted value is calculated, and the weighting factor combination with the lowest MAPE value is determined. By continuously adjusting ω and γ, the optimal weighting factor that achieves the best prediction result is selected.

[0071] The expression for Mean Absolute Percentage Error (MAPE) is:

[0072] (4)

[0073] Step S4: Based on historical energy harvesting data (specifically, actual energy harvesting information over the past few days), calculate the four parameters B(d,n), μ, and μ required for the DEMA energy prediction model according to step S3. D Substituting (d,n), ω, and γ into the DEMA energy prediction model, the predicted energy (d,n) for time slot n is calculated by weighting the energy predicted in the previous time slot with the average of the actual energy collected in the current time slot n over the past few days. This weighted average is continuously updated with each new time slot.

[0074] The DEMA energy prediction model is established through the four steps outlined above. This invention employs two exponential smoothing calculations, one for smoothing the horizontal values ​​and the other for smoothing the trend values ​​of the data. Both smoothing operations utilize a weighted moving average method, combining historical and current data for prediction. Each new predicted or revised value is obtained by a weighted average of the values ​​from the previous time slot and the data from the current time slot. The formula is as follows: The formula we use reflects energy change trends. It employs a weighted average of the overall forecast and correction factors using two weights, ω and γ. By performing two weighted moving averages, it maintains smoothness while reducing random fluctuations in the data to show the average energy change trend. This means that DEMA responds to energy changes faster and reflects energy change trends more promptly, especially when the data shows a clear upward or downward trend, making the prediction results more stable and reliable. This is particularly crucial for solar energy data that may be affected by various environmental factors such as weather and location.

[0075] Furthermore, the present invention provides an SEH wireless sensor, including a microcontroller, and a data acquisition module, a solar energy collection module, a memory, a wireless communication module, and a power management module connected to the microcontroller; wherein, the memory stores instructions executable by the microcontroller, and the instructions are executed by the microcontroller to enable the microcontroller to perform steps S1-S4 and Figure 1 The method shown;

[0076] Data acquisition module: used to acquire environmental parameters in real time through sensors. These environmental parameters include temperature, humidity, light intensity, air pressure, etc.

[0077] Solar energy collection module: mainly composed of solar panels, used to convert collected sunlight into electrical energy; photovoltaic materials, such as silicon crystals, are typically used to improve energy conversion efficiency;

[0078] Power management module: responsible for managing the electrical energy generated by the solar energy collection module, specifically including:

[0079] Charging management unit: efficiently stores the electrical energy converted from solar energy into the battery;

[0080] Voltage regulation unit: Used to power various modules in the SEH wireless sensor and ensure stable output voltage;

[0081] Energy monitoring unit: Used to monitor the battery's charge and health status in real time to prevent overcharging or over-discharging.

[0082] Wireless communication module: Used to transmit data collected by the SEH wireless sensors to upper-level devices, such as a central server or other devices, via wireless signals. Common communication technologies include Wi-Fi, LoRa, Zigbee, Bluetooth, etc., with the specific choice depending on the application scenario.

[0083] The memory is used to store historical energy harvesting data and the operating procedures required for sensor operation;

[0084] A microcontroller is used to process historical energy harvesting data. Based on the current day d, a historical D-day energy matrix E is constructed. The size of E is D×N, where N represents the total number of time slots in a day. In matrix E, E(i,n) represents the energy value collected in the nth time slot of the i-th day, and E(d,n) represents the energy value collected in the nth time slot of the current day d.

[0085] Microcontrollers are used to build DEMA energy prediction models.

[0086] ;

[0087] A microcontroller is used to calculate the parameter μ in the DEMA energy prediction model. D (d,n), B(d,n), ω and γ are calculated in the same way as step S2 above;

[0088] A microcontroller is used to calculate the four parameters B(d,n), μ, and μ required for the DEMA energy prediction model based on historical energy harvesting data. DSubstituting (d,n), ω, and γ into the DEMA energy prediction model, the predicted value (d,n) of the energy collected in time slot n is obtained by weighting the energy predicted in the previous time slot with the average value of the actual collected energy information in the past few days for the current time slot n.

[0089] This invention adds trend smoothing calculations to the single exponential moving average (DEMA), enabling DEMA to better capture trend changes in collected solar energy data. It smooths the data while retaining more trend information, which is particularly important for solar energy harvesting data, which exhibits certain trends. By more accurately predicting future energy changes, it more precisely reflects the actual situation and improves forecast accuracy.

[0090] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the invention. Those skilled in the art can make various improvements and modifications without departing from the principles of the invention, and all such improvements and modifications should be included within the scope of the claims. Therefore, the scope of protection of the present invention should include any modifications or equivalent alternatives to the technical solution, as long as they still fall within the spirit and scope of the technical solution.

Claims

1. A SEHWS energy prediction method based on double exponential moving average, characterized in that, include: Step S1: Collect and process historical energy collection data. Construct a historical D-day energy matrix E based on the current day d. The size of E is D×N, where N represents the total number of time slots in a day. Let E(d,n) represent the energy value collected in the nth time slot of the current day d. Step S2, establish the DEMA energy prediction model: ； In the formula, (d,n) is the predicted energy value collected in the nth time slot of the current day d; ω is a weighting factor used to control the weighting of recent energy data, ω∈(0,1); (d,n-1) represents the energy information predicted and collected in the previous time slot; μ(d,n) represents the average value of the actual energy information E(di,n) collected in the nth time slot of the past D days of the current day d, i∈[1,D]. B(d,n) is the correction value for the energy change trend in the nth time slot of the current day d; Step S3: Calculate the parameter μ in the DEMA energy prediction model. D (d,n), B(d,n), ω and γ, where: μ D The formula for calculating (d,n) is: ； In the formula, n is the current time slot, i is the index variable, and i∈[1,D]; The correction factor B(d,n) is represented as the weighted value of the most recent energy change trend B(d,n-1) and the average energy change trend μ(d,n)-μ(d,n-1), and its expression is: ； In the formula, γ is a weighting factor used to adjust the ratio between B(d,n-1) and μ(d,n)-μ(d,n-1), γ∈(0,1), and the initial value of B(d,n) for iterative calculation is defined as B(0,0); Update weighting factors ω and γ: Based on the range of values ​​of weighting factors ω and γ, the values ​​of the two weighting factors are traversed, and the combination of the two weighting factors is brought into the DEMA energy prediction model for prediction. The mean absolute percentage error (MAPE) between the actual value and the predicted value is calculated, and the weighting factor combination with the lowest MAPE value is determined. Step S4: Based on historical energy harvesting data, calculate the four parameters B(d,n), μ, and μ required for the DEMA energy prediction model according to step S3. D Substituting (d,n), ω, and γ into the DEMA energy prediction model, the predicted energy for time slot n is calculated by weighting the energy predicted in the previous time slot with the average of the actual energy collected in the current time slot n over the past few days. (d,n).

2. The SEHWS energy prediction method based on double exponential moving average as described in claim 1, characterized in that, B(0,0) takes the value 0.

3. The SEHWS energy prediction method based on double exponential moving average as described in claim 1, characterized in that, The expression for Mean Absolute Percentage Error (MAPE) is: 。 4. A SEH wireless sensor, characterized in that, The invention includes a microcontroller, and a data acquisition module, a solar energy collection module, a memory, a wireless communication module, and a power management module connected to the microcontroller; wherein the memory stores instructions that can be executed by the microcontroller, and the instructions are executed by the microcontroller to enable the microcontroller to perform the method described in any one of claims 1-3. The data acquisition module is used to acquire environmental parameters in real time through sensors, including temperature, humidity, light intensity, and air pressure. The solar energy collection module consists of solar panels and is used to convert collected sunlight into electrical energy. The power management module is responsible for managing the electrical energy generated by the solar energy collection module, specifically including: Charging management unit: Stores the electrical energy converted from solar energy into batteries; Voltage regulation unit: Used to power various modules in the SEH wireless sensor and ensure stable output voltage; Energy monitoring unit: Used to monitor the battery's charge and health status in real time; The wireless communication module is used to transmit the data collected by the SEH wireless sensor to the upper-level device via wireless signal transmission. The memory is used to store historical energy harvesting data and the operating procedures required for sensor operation; The microcontroller is used to process historical energy collection data and construct a historical D-day energy matrix E based on the current day d. The size of E is D×N, where N represents the total number of time slots in a day. In matrix E, E(i,n) represents the energy value collected in the nth time slot of the i-th day, and E(d,n) represents the energy value collected in the nth time slot of the current day d. The microcontroller is used to establish the DEMA energy prediction model: ； In the formula, (d,n) is the predicted energy value collected in the nth time slot of the current day d; ω is the weighting factor used to control the recent energy data, ω∈(0,1); (d,n-1) represents the energy information collected in the previous time slot; μ(d,n) represents the average value of the real energy information E(di,n) collected in the nth time slot of the past D days of the current day d, i∈[1,D]; B(d,n) is the correction value of the energy change trend in the nth time slot of the current day d; The microcontroller is used to calculate the parameter μ in the DEMA energy prediction model. D (d,n), B(d,n), ω and γ, where: μ D The formula for calculating (d,n) is: ； In the formula, n is the current time slot, i is the index variable, and i∈[1,D]; The correction factor B(d,n) is represented as the weighted value of the most recent energy change trend B(d,n-1) and the average energy change trend μ(d,n)-μ(d,n-1), and its expression is: ； In the formula, γ is a weighting factor used to adjust the ratio between B(d,n-1) and μ(d,n)-μ(d,n-1), γ∈(0,1), and the initial value of B(d,n) for iterative calculation is defined as B(0,0); Update weighting factors ω and γ: Based on the range of values ​​of weighting factors ω and γ, the values ​​of the two weighting factors are traversed, and the combination of the two weighting factors is brought into the DEMA energy prediction model for prediction. The mean absolute percentage error (MAPE) between the actual value and the predicted value is calculated, and the weighting factor combination with the lowest MAPE value is determined. The microcontroller is used to calculate the four parameters B(d,n), μ, and μ required for the DEMA energy prediction model based on historical energy harvesting data. D Substituting (d,n), ω, and γ into the DEMA energy prediction model, the predicted energy for time slot n is calculated by weighting the energy predicted in the previous time slot with the average of the actual energy collected in the current time slot n over the past few days. (d,n).

5. The SEH wireless sensor as described in claim 4, characterized in that, B(0,0) takes the value 0.

6. The SEH wireless sensor as described in claim 4, characterized in that, The expression for Mean Absolute Percentage Error (MAPE) is: 。

Images