Self-powered wireless sensor node energy efficiency control method and system for indoor weak light scene

Abstract

The invention provides a self-powered wireless sensor node energy efficiency control method and system for an indoor weak light scene, and belongs to the technical field of self-powered wireless sensors of the Internet of Things. Comprising the following steps: constructing an energy collection and energy perception calculation model which comprises an energy collection module, an energy perception calculation module, an energy prediction module and an energy management module; the energy sensing calculation module dynamically adjusts the capacitor capacity by using a temperature correction model in the process of calculating available energy according to the super-capacitor voltage; the energy prediction module performs prediction by adopting a mixed entropy weight rolling adaptive energy prediction algorithm based on low energy density and high-frequency fluctuation characteristics of an indoor weak light scene, and dynamically adjusts a prediction weight according to an energy fluctuation condition by distinguishing situations, controlling dormancy and quantifying energy fluctuation. Balancing the indoor energy prediction precision and the energy consumption in the energy collection process; and the energy management module performs energy efficiency control based on the model, calculates an energy prediction value of the next time slot and performs working mode control of the node.

Description

Technical Field

[0001] This invention relates to the field of self-powered wireless sensor technology in the Internet of Things (IoT), and in particular to a method and system for energy efficiency control of self-powered wireless sensor nodes in indoor low-light scenarios. Background Technology

[0002] With the widespread application of IoT technology in smart buildings, smart homes, and other fields, the deployment scale of indoor self-powered wireless sensors is growing exponentially. Due to the significant "low energy" and "high dynamic" characteristics of indoor environments, indoor self-powered wireless sensors must meet stringent ultra-low power consumption requirements. Therefore, to minimize circuit complexity and reduce energy losses during transmission, most mainstream indoor micro-energy systems currently adopt a simplified "energy harvesting-energy storage-energy consumption" direct operating mode. In this mode, compared to traditional chemical batteries, supercapacitors, with their advantages of high charge-discharge efficiency, high power density, and long cycle life, have become a more suitable energy storage element for indoor self-powered systems.

[0003] However, when applying this mode for energy efficiency control of self-powered wireless sensor nodes, existing technologies suffer from insufficient energy sensing accuracy, poor predictive logic adaptability, and poor energy efficiency control capabilities. On the one hand, because the actual capacity of supercapacitors is affected by operating temperature and undergoes "dynamic drift," existing solutions lack targeted dynamic compensation, making it impossible to accurately deduce the actual energy from the voltage. On the other hand, existing energy efficiency control methods still follow the indirect control approach of "theoretical output energy + efficiency conversion + energy efficiency control" for outdoor strong light. This approach introduces too many intermediate conversion steps, which not only increases computational energy consumption but also makes it difficult to accurately quantify the complex losses in the process from theoretical output to actual storage. This seriously violates the low-power intention of the "direct working mode" of indoor micro-energy systems, leading to the failure of energy efficiency control of self-powered wireless sensor nodes, energy overflow of sensor nodes, or cold start, failing to meet the requirements of indoor micro-energy self-powered wireless sensors for high precision and low power consumption for long-term operation. Summary of the Invention

[0004] This invention provides an energy efficiency control method and system for self-powered wireless sensor nodes in indoor low-light scenarios. It overcomes the shortcomings of existing energy harvesting and prediction technologies, such as insufficient accuracy caused by energy sensing calculation models and poor adaptability of energy efficiency control methods to indoor low-light environments. By constructing an "energy sensing calculation" model, it achieves accurate measurement of the meager available energy indoors. It also combines the hybrid entropy weight rolling adaptive energy prediction algorithm HERA to adapt to energy fluctuations in low-light environments, thereby reducing losses in the energy sensing calculation and prediction process and significantly improving prediction accuracy. Finally, it adjusts the node's operating mode based on the predicted value, providing reliable support for the long-term stable operation of the self-powered wireless sensor.

[0005] The technical solution adopted by the present invention to solve its technical problem is as follows:

[0006] The first aspect of this invention provides a method for energy efficiency control of self-powered wireless sensor nodes in indoor low-light scenarios, comprising:

[0007] An energy harvesting and energy sensing calculation model is constructed, including an energy harvesting module, an energy sensing calculation module, an energy prediction module, and an energy management module. The energy harvesting module achieves efficient energy harvesting by adjusting indoor solar panels through MPPT. The energy sensing calculation module calculates available energy based on the supercapacitor voltage, and dynamically adjusts the capacitor capacity using a temperature correction model during the calculation process. The energy prediction module is based on the low energy density and high-frequency fluctuation characteristics of indoor low-light scenarios. The model uses the hybrid entropy weight rolling adaptive energy prediction algorithm HERA for prediction. By distinguishing scenarios, implementing sleep control, quantifying energy fluctuations, and dynamically adjusting prediction weights according to energy fluctuations, it balances the accuracy of indoor energy prediction with energy consumption during energy harvesting. The energy management module adjusts the node's operating mode based on the predicted value of available energy in the next time slot and the current supercapacitor energy value.

[0008] Model-based energy efficiency control of self-powered wireless sensor nodes:

[0009] Step 1: System initialization, completing the calibration of each component in the system and configuring system parameters;

[0010] Step 2: Collect data periodically, including the real-time voltage and operating temperature of the supercapacitor, and add context tags according to the current context type;

[0011] Step 3, Energy Sensing Calculation: Input the operating temperature into the temperature dynamic correction model to obtain the dynamic capacity of the supercapacitor, and combine it with the voltage to calculate the usable energy collected;

[0012] Step 4, sleep control mechanism judgment: If the sum of available energy collected in S consecutive time slots is lower than the threshold, the sleep gating strategy is triggered and a simplified prediction value is output; otherwise, continue to step 5.

[0013] Step 5, Context Unit Matching: Based on available energy and context labels, match the corresponding context units in the historical benchmark library and extract the historical benchmark energy values;

[0014] Step 6: Perform dynamic energy prediction using the HERA algorithm: quantify the degree of energy fluctuation, calculate the normalized Shannon entropy of the energy values ​​in the last S time slots, and further obtain dynamic weights through linear interpolation; based on historical baseline energy values ​​and recent... The actual energy values ​​of similar time periods are used to calculate the dynamic environment correction factor; combined with the current time slot prediction error, the energy prediction value of the next time slot is calculated.

[0015] Step 7, Operating mode adjustment: The operating mode of the sensor node is controlled by the supercapacitor voltage in the current time slot and the predicted energy value available in the next time slot, so as to achieve a dynamic balance between energy supply and demand.

[0016] Step 8, Historical reference energy value update: Calculate the new reference value based on the actual energy value of the current time slot and the historical reference energy value before the update;

[0017] Step 9: After completing the daily forecast task, update the status: Store the available energy collected in all time slots of the day into the corresponding historical benchmark database. When the number of data days under the same scenario label reaches... When the new data is updated, delete the historical data of the day with the highest similarity to the new data, and update the historical data of the corresponding context unit.

[0018] Preferably, in the energy harvesting and energy sensing calculation model:

[0019] Task sequence is defined as ,in For the first Energy prediction task in one time slot, task The demand vector is defined as , The context type for the day of the task; for The voltage of the time-slot supercapacitor; for Operating temperature of time-slotted supercapacitors; dividing a day into K large time slots. , k∈[1,K], each large time slot Includes S hourly gaps , i∈[1,S];

[0020] The HERA algorithm parameter set is defined as follows: ,in, For the first Historical baseline energy values ​​for the time slot context; for The normalized Shannon entropy of a time slot represents the quantification of energy fluctuations; for Dynamic weighting of time slots; for Environmental correction factor for time slots; for Prediction error of time slots; This is the error correction factor; The sleep threshold;

[0021] HERA's dynamic prediction logic is as follows: Dynamically adjust the weighting of real-time data and historical benchmarks. , combined and Achieve environmental adaptation and error correction, and improve prediction results. The expression is: ; ;

[0022] Indicates time slot The usable energy collected.

[0023] Preferably, step 1, system initialization, includes:

[0024] System calibration: Calibrate the MCU's built-in ADC and temperature sensor;

[0025] Parameter configuration: Capacitance value of the supercapacitor at 25℃ Temperature coefficient HERA algorithm weight upper and lower bounds , Benchmark update coefficients Sleep threshold Error feedback coefficient Time slot value , Total number of time slots K, total number of hourly time slots S, total number of scenario types W, scenario types Write to the MCU;

[0026] Baseline library initialization: based on the current Read historical data from the same context and construct an energy matrix. Context library : ; ;

[0027] in, Indicates the first case under the corresponding scenario w. The usable energy collected in the Kth large time slot;

[0028] Calculate the reference value for each time slot Represented as: ;

[0029] in, This represents the number of days corresponding to the first scenario. For the first The number of days corresponding to each scenario, x∈[1,N], k∈[1,K];

[0030] Configure the timer according to the preset acquisition frequency.

[0031] Preferably, step 3, energy sensing calculation, includes:

[0032] Using a temperature correction model to correct the dynamic capacitance of supercapacitors The expression is: ;

[0033] Based on supercapacitors Energy stored in time slots The expression is: ;

[0034] When the stored energy exceeds the maximum capacity of the supercapacitor At present The energy stored in a time-slot supercapacitor is defined as ;

[0035] Calculate by combining sampling and sleep power consumption. Average power consumption per time slot The expression is: ;

[0036] in For the sensor at the sampling rate Sampling power consumption, This refers to the power consumption when the node is in sleep mode. for Sampling rate within time slot Duty cycle;

[0037] calculate Time slots consume energy The expression is: ;

[0038] Time slot Collected usable energy The expression is: ;

[0039] In the formula, for The total energy consumed by the sensor within the time slot;

[0040] Available energy collected in time slots The expression is: ;

[0041] Preferably, step 4, the sleep control mechanism determination, includes: ;

[0042] Preferably, step 5, context unit matching, includes:

[0043] Based on the day's context label Match the corresponding context unit E in the historical benchmark library. w Extract the corresponding value.

[0044] Preferably, step 6, which uses the HERA algorithm for dynamic energy prediction, includes:

[0045] Volatility Quantization Calculation: Take the first S times within a time slot The usable energy collected in the time slots is used to calculate the normalized Shannon entropy. The expression is: ;

[0046] in for Probability distribution of time slot energy sample values for In S The number of times it appears in the available energy values ​​collected in the time slot. ;

[0047] Dynamic weight calculation: based on Dynamic weights are determined using linear interpolation. ,based on Environmental correction factors were determined by comparing actual energy values ​​from nearly H similar time periods. : ; ; ; ;

[0048] in, This is a vector of ratios between actual values ​​and benchmark values. To increase the time weight, Number of similar time periods;

[0049] Error feedback correction: combining the current time slot prediction error Substitute the values ​​into the prediction formula to calculate the predicted value for the next time slot. ,error The expression is: ;

[0050] in, For the first The predicted value of the time slot, For the first The actual value of the time slot.

[0051] In step 7, the operating mode of the wireless sensor node Based on the supercapacitor voltage and predicted values, the operating mode to be used in the next time slot is determined and executed. The specific method for determining the operating mode is as follows:

[0052] The supercapacitor voltage is below the death voltage threshold. When forced into a frozen lockout working mode Level=-1, the wireless sensor node enters a cold start state.

[0053] The supercapacitor voltage has recovered to the turn-on voltage threshold. When this happens, the lock is released and the system is allowed to enter a low-power operating mode (Level=0) for energy prediction.

[0054] In low-power operating mode, according to The operating mode of the wireless sensor node at the next moment is determined as mode L, where L∈[0,Q]:

[0055] Calculate the target energy value E of the system target : ;

[0056] In the formula, This represents the system's target energy value when energy is sufficient. This represents the system's target energy value when energy is insufficient. This represents the minimum energy consumption required for a wireless sensor node to maintain a no-sampling sleep mode.

[0057] Calculate the gap between the predicted value for the next time slot and the target energy required to maintain system operation:

[0058] ;

[0059] L is determined based on the gap: Gap>0; Gap≤0;

[0060] In the formula, This represents the energy of the supercapacitor in the current time slot, and round() is the rounding function. This represents the maximum energy storage capacity of the supercapacitor. This represents the death energy value; the coefficient mk takes the value F when there is energy collection and E when there is no energy collection, where F>E; 1...Q represents the node's duty cycle. Data collection is performed with the highest sampling frequency when L=Q.

[0061] Preferably, step 8, updating the historical baseline energy value, includes:

[0062] The historical baseline values ​​of the context unit are updated using the exponential moving average algorithm. The formula is: ;

[0063] in, The historical baseline energy value before the update. This is a new baseline value that incorporates the current real-time energy values.

[0064] Preferably, step 9, which involves updating the status after completing the daily prediction task, includes:

[0065] Once all time slots for the day are predicted, the actual energy values ​​for each time slot will be calculated. Store the data in the context unit corresponding to the context tag, recalculate the historical baseline value of each context unit, and complete the update of the historical baseline library;

[0066] Determine whether the number of storage days for a context unit with the corresponding context label for the current day has reached the corresponding fixed storage number N. If so, delete the available energy data collected on the day most similar to the current day's data, and store the available energy data from the K time slots collected on the current day into the corresponding context unit. The difference in overall energy scale between the day to be determined and historical days under the same context is quantified by the daily average energy deviation rate. ; ;

[0067] in, The average daily usable energy collected on the day to be determined. This represents the average daily usable energy collected under the same historical circumstances. , These are the available energy values ​​collected in large time slots for the date to be determined and historical dates under the same circumstances, respectively. The daily average energy deviation rate is selected from historical days with the same conditions as the day to be determined, using the minimum ( () as the most similar day.

[0068] A second aspect of this invention provides an energy efficiency control system for self-powered wireless sensor nodes in indoor low-light scenarios, the system comprising:

[0069] The building unit is used to construct the energy harvesting and energy sensing calculation model, including an energy harvesting module, an energy sensing calculation module, an energy prediction module, and an energy management module. The energy harvesting module achieves efficient energy harvesting by adjusting indoor solar panels through MPPT. The energy sensing calculation module calculates available energy based on the supercapacitor voltage, and dynamically adjusts the capacitor capacity using a temperature correction model during the calculation process. The prediction module is based on the low energy density and high-frequency fluctuation characteristics of indoor low-light scenarios. The model uses the hybrid entropy weight rolling adaptive energy prediction algorithm HERA for prediction. By distinguishing scenarios, implementing sleep control, quantifying energy fluctuations, and dynamically adjusting prediction weights according to energy fluctuations, it balances the accuracy of indoor energy prediction with energy consumption during energy harvesting. The calculation unit is used to perform energy efficiency control of self-powered wireless sensor nodes based on the model, calculate the energy prediction value for the next time slot, and control the working mode of the nodes.

[0070] As can be seen from the above technical solution, the present invention provides a method and system for energy efficiency control of self-powered wireless sensor nodes for indoor low-light scenarios. An energy harvesting and energy sensing calculation model is constructed, including an energy harvesting module, an energy sensing calculation module, an energy prediction module, and an energy management module. The energy harvesting module achieves efficient energy harvesting by adjusting indoor solar panels using MPPT. The energy sensing calculation module calculates available energy based on the supercapacitor voltage, dynamically adjusting the capacitor capacity using a temperature correction model during the calculation process. The energy prediction module, based on the low energy density and high-frequency fluctuation characteristics of indoor low-light scenarios, uses the HERA (Hybrid Entropy Weighted Rolling Adaptive Energy Prediction) algorithm for prediction. By distinguishing scenarios, implementing sleep control, quantifying energy fluctuations, and dynamically adjusting prediction weights based on energy fluctuations, it balances the accuracy of indoor energy prediction with energy consumption during energy harvesting. The energy management module performs energy efficiency control based on the model, calculates the predicted energy value for the next time slot, and adjusts the node's operating mode based on the predicted available energy value for the next time slot and the current supercapacitor energy value. Designed for indoor low-light scenarios, this solution utilizes the nonlinear voltage characteristics of supercapacitors for precise energy measurement and combines this with a hybrid entropy weighted rolling adaptive energy prediction algorithm for energy forecasting. This reduces losses during energy sensing calculation and prediction, significantly improving prediction accuracy and energy efficiency control capabilities, and providing reliable support for the long-term stable operation of self-powered wireless sensors. This solution is suitable for self-powered wireless sensors deployed in large-scale distributed sensing scenarios such as smart buildings and smart homes that require long-term maintenance-free operation. Attached Figure Description

[0071] Figure 1 This is a diagram illustrating the overall architecture of the self-powered wireless sensor node energy efficiency control system for indoor low-light scenarios according to the present invention.

[0072] Figure 2 This is a schematic diagram of the energy calculation process for voltage sensing in a supercapacitor.

[0073] Figure 3 This is a logic block diagram of the Hybrid Entropy Weighted Rolling Adaptive Energy Prediction Algorithm (HERA).

[0074] Figure 4 This is a flowchart illustrating the execution steps of the energy efficiency control method for self-powered wireless sensor nodes in indoor low-light scenarios according to the present invention. Detailed Implementation

[0075] The technical solution and effects of the present invention will be further described in detail below with reference to the accompanying drawings.

[0076] The purpose of this invention is to overcome the shortcomings of existing energy efficiency control methods for wireless sensor nodes, particularly the insufficient accuracy caused by energy sensing calculation models in energy harvesting prediction technology and poor adaptability to indoor low-light environments. This invention provides an energy efficiency control method for self-powered wireless sensor nodes designed for indoor low-light scenarios. By constructing an energy harvesting and energy sensing calculation model, accurate measurement of the meager available energy indoors is achieved. Furthermore, a Hybrid Entropy Weighted Rolling Adaptive Energy Prediction Algorithm (HERA) is incorporated to adapt to energy fluctuations in low-light environments, thereby reducing losses in the energy sensing calculation and prediction process. This solves the measurement errors caused by the dynamic drift of supercapacitor capacity, avoids the cumbersome conversion steps in traditional prediction algorithms, significantly improves prediction accuracy and energy efficiency control capabilities, and provides reliable support for the long-term stable operation of self-powered wireless sensors.

[0077] When implementing energy efficiency control methods for self-powered wireless sensor nodes in indoor low-light scenarios, the core principles should be "simplified measurement adapted to low light and low energy" and "algorithm dynamically matching high-frequency fluctuations." A closed-loop process of "model building - initialization - energy calculation - sleep mechanism - predictive execution - energy efficiency control" should be employed to ensure long-term stable and autonomous operation of the system in indoor low-light environments. (Reference) Figure 1 The schematic diagram shown illustrates the energy efficiency control method for self-powered wireless sensor nodes in indoor low-light scenarios provided by this invention, which includes:

[0078] Construct an energy harvesting and energy sensing computing model, including an energy harvesting module, an energy sensing computing module, an energy prediction module, and an energy management module;

[0079] Energy harvesting module: It adopts indoor-specific flexible photovoltaic modules (such as PowerFilm SF11035), which can stably output low power in low light environment (light intensity ≤200 lux) without the need for additional current amplification circuit;

[0080] Energy sensing and computing module: Utilizes a low-power chip supporting Maximum Power Point Tracking (MPPT) to improve energy utilization efficiency in low-light conditions. The energy storage element is a supercapacitor, defined as having a rated capacity of [value missing] at 25°C. Maximum stored energy is Simultaneously, a temperature correction model is used to compensate for the dynamic drift of the capacitor capacitance with temperature; a low-power MCU is used as the core, and a built-in high-precision ADC is used to acquire the supercapacitor voltage. The internal temperature sensor obtains the operating temperature of the supercapacitor. ;

[0081] Energy prediction module: integrates HERA algorithm, runs on MCU kernel, and is adapted to the resource-constrained characteristics of embedded devices;

[0082] Energy Management Module: Adjusts the operating mode of the node by comparing the predicted energy available in the next time slot with the current supercapacitor energy value.

[0083] Task Sequences and Demand Vectors: A task sequence is defined as... ,in For the first Energy prediction task in one time slot, task The demand vector is defined as , The scenario type for the day of the task is used to match the corresponding energy baseline value library; for The voltage of the time-slot supercapacitor; For the present Operating temperature of time-slotted supercapacitors; dividing a day into K large time slots. , k∈[1,K], each large time slot Includes S hourly gaps , i∈[1,S];

[0084] This invention optimizes energy calculation methods for indoor low-light scenarios: it directly calculates the collected usable energy using the supercapacitor voltage; simultaneously, it reads data from the MCU's internal temperature sensor, introducing dynamic temperature correction and an energy consumption quantification model to ensure the accuracy of energy calculation. The specific design is as follows:

[0085] Time slot division definition: Divide each day into k large time slots T (each large time slot lasts for...) minute, Each large time slot T is then subdivided into S small time slots t (each small time slot lasts for a period of time). minute, That is, S hourly time slots are aggregated into one large time slot (where, For the k-th large time slot, (This refers to the i-th hour slot within the k-th large time slot). Energy harvesting is performed in hourly time slot t to accurately capture indoor energy fluctuations; energy prediction is conducted in large time slot T to avoid energy waste caused by high-frequency prediction.

[0086] Calculation of usable energy collected in time slots: The energy harvesting module uses indoor-specific flexible photovoltaic panels and maximum power harvesting units to efficiently collect energy from indoor LED lighting and diffused natural light. Available energy collected in time slots for: (1) (2)

[0087] in for The usable energy collected in the time slot (unit: J), S is Number of time slots for Energy of a time-slotted supercapacitor (unit: J) for Energy of a time-slotted supercapacitor (unit: J) for Total energy consumed by the sensor within the time slot (unit: J).

[0088] Energy storage calculation using supercapacitors: The energy storage element uses supercapacitors, which are suitable for energy storage needs in indoor low-light scenarios. The energy sensing and calculation module collects the real-time voltage of the supercapacitors through a built-in high-precision ADC. Calculate the supercapacitor using a temperature dynamic correction model. Energy stored in time slots The formula is: (3)

[0089] in, for The dynamic capacitance of the time-slotted supercapacitor is determined by the current operating temperature of the supercapacitor, which is collected by the internal temperature sensor of the MCU. The correction is performed, and the correction model is as follows: (4)

[0090] in, This refers to the rated capacitance of a supercapacitor at 25°C (unit: F). This is the coefficient between the increase in the capacitance of a supercapacitor and the temperature difference (within the specified operating temperature range of the supercapacitor). for Operating temperature of time-slot supercapacitors (unit: °C); when the stored energy exceeds the maximum capacity of the supercapacitor. (Depend on calculate, When the voltage is the rated voltage of the supercapacitor, to avoid energy overflow and waste, currently... The energy stored by a time-slotted supercapacitor is defined as: (5)

[0091] Slot-average power calculation: combining sampling and sleep power consumption, Average power consumption per time slot The calculation formula is: (6)

[0092] in For the sensor at the sampling rate The sampling power consumption (in W) is as follows. Power consumption (in W) when the node is in sleep mode. for Sampling rate within time slot The duty cycle (value range [0,1]).

[0093] Energy consumption calculation for time slots: The energy consumed in a time slot is calculated based on the average power of the time slot, using the following formula: (7)

[0094] in For time slots Sampling rate The average power consumed by the lower sensor (in W). for Duration (in seconds).

[0095] The Hybrid Entropy Weighted Rolling Adaptive Energy Prediction Algorithm (HERA) is designed for indoor low-light scenarios characterized by "low energy density + high-frequency fluctuations." It constructs a set of HERA algorithm parameters and core logic, overcoming the adaptability limitations of traditional algorithms. The specific definitions are as follows:

[0096] The HERA algorithm parameter set is defined as follows: ,in, For the first Historical baseline energy values ​​for the time slot context; for The normalized Shannon entropy of the time slot (quantizing energy fluctuations, taking values ​​[0,1]) represents the quantification of energy fluctuations; for Dynamic weighting of time slots (balancing real-time data with historical benchmarks); for Environmental correction factor for time slots (to compensate for the impact of indoor environment); for Prediction error of time slots (error correction); This is the error correction factor; The sleep threshold is set to trigger a low-power sleep strategy, reducing unnecessary computing power consumption.

[0097] Dynamic prediction logic: through Dynamically adjust the weighting of real-time data and historical benchmarks. , combined (Environmental Correction Factor) and (Prediction error) Achieve environmental adaptation and error correction; prediction results The expression is: (8) (9)

[0098] Indicates time slot The usable energy collected.

[0099] Hybrid Entropy Weighted Rolling Adaptive Energy Prediction Algorithm (HERA):

[0100] To address the challenges of high-frequency fluctuations in indoor energy and limited sensor resources, this paper proposes a method that balances indoor energy prediction accuracy with energy consumption during energy harvesting by differentiating scenarios, implementing sleep control, quantifying energy fluctuations, and dynamically adjusting prediction weights. The specific design is as follows:

[0101] Context-aware rolling benchmark modeling: Based on the differences in indoor energy distribution under different conditions, a two-dimensional matrix is ​​used to store the historical data for each context. Useful energy data collected in time slots are used to construct energy matrices for different scenarios. (A real matrix of N rows and K columns) stores the time-slot energy data corresponding to the scenario. For a fixed number of scenarios, To represent a fixed number of storage days for the corresponding scenario, where K is the number of large time slots divided each day (ensuring data scale is controllable), the specific form is as follows: (10) (11)

[0102] in, A scenario library of length W (storing the locations of different scenario energy libraries), This represents the energy reservoir for the w-th scenario; N represents the total number of storage days in the w-th scenario unit. Indicates the first case under the corresponding scenario w. The available energy collected in the Kth large time slot of the day; this storage structure is defined as a context-aware rolling baseline model, used to distinguish energy distribution patterns under different scenarios, avoid redundant overhead of the three-dimensional matrix, and improve data access efficiency. To clarify the rolling baseline value... The computational logic is based on constructing benchmarks for different scenarios using a two-dimensional matrix for each scenario, as shown in the following formula: (12)

[0103] in, This represents the number of days corresponding to the first scenario. For the first The number of days corresponding to each scenario; by averaging historical data of similar scenarios, the baseline energy value for each time slot is obtained, reflecting the energy distribution pattern under that scenario.

[0104] Once all time slots for the day are predicted, the actual energy values ​​for each time slot will be calculated. Store the data in the context unit corresponding to the context tag, recalculate the historical baseline value for each context unit, and complete the historical baseline database update; when the number of days stored under a context tag reaches N (corresponding to N×K data entries), delete the data for the day with the highest similarity to the new data (including k large time slot data), ensuring that each context tag always maintains... Representative data points are used to avoid redundant storage while accurately reflecting the energy distribution patterns under various scenarios. The difference in overall energy scale between the day to be judged and historical days under the same scenario is quantified by the "daily average energy deviation rate," calculated as follows: (13) (14)

[0105] in, The average daily usable energy collected on the day to be determined (in J). This represents the average daily usable energy collected under the same historical conditions (in J). , These are the available energy values ​​collected in large time slots for the date to be determined and historical dates under the same circumstances, respectively. The daily average energy deviation rate is selected from historical days with the same conditions as the day to be determined, using the minimum ( () as the most similar day.

[0106] Sleep control mechanism: To adapt to the low-power requirements of resource-constrained sensors, a low-power sleep gating mechanism is set up to skip invalid calculation processes. The formula is as follows: (15)

[0107] Among them, when S consecutive Available energy collected in time slots When the sum is lower than the sleep threshold, the predicted value for the next time slot is directly output as 0.0. Since there is no effective energy input during this period (such as at night when there is no light), there is still no possibility of energy collection in subsequent time slots. The complex calculation process can be skipped, which significantly reduces the sensor's ineffective energy consumption.

[0108] Entropy weighted dynamic weighting: Calculate the last S... Normalized Shannon entropy of available energy values ​​collected in time slots The formula for quantifying the degree of energy fluctuation in an indoor low-light environment is as follows: (16)

[0109] in for Probability distribution of time slot energy sample values for In S The number of times it appears in the available energy values ​​collected in the time slot. The larger the value, the more drastic the energy fluctuation (such as sudden changes caused by light switch-on or people blocking the light); the smaller the value, the more gradual the fluctuation (such as during stable lighting periods).

[0110] according to Dynamic weights are determined using linear interpolation. Balance the present The weighting of the slot energy value in relation to the historical baseline model is calculated using the following formula: (17)

[0111] in, As upper and lower limits of the weight, when fluctuations are severe. Approaching (Focusing on current observations and responding quickly to sudden changes), when fluctuations are gradual, it tends to be close to... (Focusing on historical benchmarks to ensure forecast stability).

[0112] Environmental correction factor: Introduced to correct for the impact of changes in environmental and equipment conditions on energy harvesting. The factors are dynamically adapted, and the influence of recent data is amplified through time-weighted averaging, as shown in the following formula: (18) (19) (20)

[0113] in, This is a vector of ratios between actual values ​​and benchmark values ​​(reflecting the deviation between current and historical values). To increase the time weight (the newer the data, the greater the weight, to amplify the impact of recent changes), For the number of similar time periods, The first time was based on the type of situation. The results are obtained from the corresponding benchmark calculations, followed by historical benchmark values ​​updated from the benchmark values. Perform the substitution calculation.

[0114] Error Feedback Mechanism: To reduce the cumulative error of the system, an error feedback mechanism is designed to correct the error in real time based on the deviation between the predicted value and the actual value in the current time slot. The calculation formula is as follows: (twenty one)

[0115] in, For the first The predicted value of the time slot, For the first The actual value of the time slot is used to dynamically adjust the prediction result of the next time slot through error correction. The final calculation formula is as follows: (twenty two) (twenty three)

[0116] in, The sleep threshold, This is the error feedback coefficient, which is used to adjust the correction strength of the prediction error for the next time slot prediction value, ensuring that the error gradually converges.

[0117] Operating mode adjustment: The operating mode of the sensor node is controlled by the supercapacitor voltage in the current time slot and the predicted available energy value for the next time slot, thereby achieving a dynamic balance between energy supply and demand. The specific steps are as follows:

[0118] (1) Parameter description: Working mode: Current time slot supercapacitor voltage Death voltage threshold , turn on voltage threshold Target energy deviation Gap, system target voltage threshold The system has sufficient energy to reach the target voltage value. The system energy is insufficient to reach the target voltage value. System target energy value: (twenty four)

[0119] in, This represents the system's target energy value when energy is sufficient. This represents the system's target energy value when energy is insufficient. It is the set system target energy threshold; where , , By ensuring the system has sufficient energy to reach the target voltage value Insufficient system energy, target voltage value System target voltage threshold Substitute into formula (3) for calculation. Working mode: -1, dead working mode (cannot perform any work), 0 represents standby mode (only supercapacitor voltage acquisition and energy prediction are performed, no other data acquisition is performed, such as scene-related data acquisition, no data transmission is performed), 1....Q represents data acquisition using different duty cycles.

[0120] (2) Hysteresis control: When the system voltage drops below When forced into a frozen, locked operating state (Level -1), all sensing, communication, and computing modules except the voltage monitoring unit are disconnected to prevent further energy depletion, and the sensor nodes enter a cold start state; only when the energy storage element voltage rises back to the second voltage threshold... At this point, the lock is released, allowing the system to enter a low-power operating mode. The strong hysteresis mechanism effectively eliminates the frequent start-stop 'oscillation' phenomenon caused by small energy fluctuations at the weak light critical point.

[0121] (3) Dynamic level adjustment: Based on the deviation between the predicted value of the next time slot in step S6 and the target energy required to maintain system operation: (25)

[0122] in, The current time slot supercapacitor voltage is The calculated energy value of the supercapacitor in the current time slot.

[0123] If Gap > 0 (surplus compensation): (26)

[0124] like (Deficit Protection): (27)

[0125] The task levels are mapped to 0 to Q. This represents the maximum energy storage capacity of the supercapacitor. yes The death energy value is calculated using formula (3). This represents the energy of the supercapacitor in the current time slot. The coefficient mk: F is used when energy harvesting is available (aggressive control), and E is used when there is no energy harvesting (conservative control) (F>E). Level Q corresponds to the highest sampling frequency; level 0 corresponds to no sampling and sleep mode. Each level corresponds to a preset MCU duty cycle parameter. .

[0126] Baseline Updates: To prevent the baseline model from becoming outdated, the rolling baseline is dynamically updated based on the scenario type. An exponential moving average (EMA) is used to fuse the latest data, balancing historical patterns with real-time changes, and updating the historical baseline values ​​for scenario units. The formula is as follows: (28)

[0127] in, As the baseline update coefficient, The historical baseline energy value before the update. To integrate the current real-time energy values ​​into a new baseline value for prediction in the next time slot, the baseline model must be able to absorb new energy characteristics in real time.

[0128] Model-based energy efficiency control of self-powered wireless sensor nodes:

[0129] Step 1: System initialization, completing the calibration of each component and configuring system parameters; specifically including:

[0130] System calibration: To meet the requirements of indoor low-light scenarios, calibrate the ADC sampling accuracy (ensuring voltage measurement error ≤0.1mV) and the error of the MCU internal temperature sensor;

[0131] Parameter configuration: Capacitance value of the supercapacitor at 25℃ Temperature coefficient HERA algorithm weight upper and lower bounds , Benchmark update coefficients Sleep threshold Error feedback coefficient Time slot value , Total number of time slots K, total number of hourly time slots S, total number of scenario types W, scenario types Duty cycle of working mode Death voltage threshold ( ), enable voltage threshold ( The system has sufficient energy and a voltage threshold. Insufficient system energy voltage threshold System target voltage threshold The maximum voltage of a supercapacitor is V. max Write to the MCU;

[0132] Baseline library initialization: based on the current Read historical data from the same context and construct an energy matrix. Context library ;

[0133] Configure the timer according to the preset acquisition frequency;

[0134] Step 2: Collect data periodically, including the real-time voltage and operating temperature of the supercapacitor, and add context tags according to the current context type;

[0135] Step 3, Energy Sensing Calculation: Input the operating temperature into the temperature dynamic correction model to obtain the dynamic capacity of the supercapacitor, and combine it with the voltage to calculate the usable energy collected;

[0136] Step 4, sleep control mechanism judgment: If the sum of available energy collected in S consecutive time slots is lower than the threshold, the sleep gating strategy is triggered and a simplified prediction value is output; otherwise, continue to step 5.

[0137] Step 5, Context Unit Matching: Based on available energy and context labels, match the corresponding context units in the historical benchmark library and extract the historical benchmark energy values;

[0138] Step 6: Perform dynamic energy prediction using the HERA algorithm: quantify the degree of energy fluctuation, calculate the normalized Shannon entropy of the energy values ​​in the last S time slots, and further obtain dynamic weights through linear interpolation; based on historical baseline energy values ​​and recent... The actual energy values ​​of similar time periods are used to calculate the dynamic environment correction factor; combined with the current time slot prediction error, the energy prediction value of the next time slot is calculated.

[0139] Step 7, Operating mode adjustment: The operating mode of the sensor node is controlled by the supercapacitor voltage in the current time slot and the predicted energy value available in the next time slot, so as to achieve a dynamic balance between energy supply and demand.

[0140] Step 8, Historical reference energy value update: Calculate the new reference value based on the actual energy value of the current time slot and the historical reference energy value before the update;

[0141] Step 9: After completing the daily forecast task, update the status: Store the available energy collected in all time slots of the day into the corresponding historical benchmark database. When the number of data days under the same scenario label reaches... When the new data is updated, delete the historical data of the day with the highest similarity to the new data, and update the historical data of the corresponding context unit.

[0142] Furthermore, the present invention provides an energy efficiency control system for a self-powered wireless sensor node for indoor low-light scenarios, used to implement the aforementioned energy efficiency control method for a self-powered wireless sensor node for indoor low-light scenarios, comprising:

[0143] The building unit is used to construct the energy harvesting and energy sensing calculation model, including an energy harvesting module, an energy sensing calculation module, an energy prediction module, and an energy management module; the energy storage element is a supercapacitor; the energy sensing calculation module calculates the available energy based on the supercapacitor voltage, and dynamically adjusts the capacitor capacity using a temperature correction model during the calculation process; the prediction module is based on the low energy density and high frequency fluctuation characteristics of indoor low-light scenarios, and the model uses the hybrid entropy weight rolling adaptive energy prediction algorithm HERA for prediction. By distinguishing scenarios, implementing sleep control, quantifying energy fluctuations, and dynamically adjusting the prediction weights according to energy fluctuations, it balances the accuracy of indoor energy prediction with energy consumption during energy harvesting.

[0144] The computing unit is used for energy efficiency control of self-powered wireless sensor nodes based on models, calculating the energy prediction value for the next time slot, and controlling the working mode of the nodes.

[0145] The process of the computing unit will be explained below with reference to an example:

[0146] Step 1: System initialization operations:

[0147] Before the prediction period begins, complete system calibration, parameter configuration, and benchmark library initialization to ensure the system and algorithm are ready. Specific steps include:

[0148] System calibration: The MCU's built-in ADC is calibrated using a standard voltage source to ensure that the voltage measurement error is ≤0.1mV; the MCU's built-in temperature sensor is calibrated using a high-precision constant temperature chamber to ensure that the supercapacitor's operating temperature measurement error is ≤0.2℃.

[0149] Parameter configuration: Supercapacitor capacitance value Temperature coefficient HERA algorithm weight upper and lower bounds , Benchmark update coefficients Sleep threshold Error feedback coefficient Time slot value It lasts for 15 minutes. It is 3 minutes, i.e., k=96 (number of large time slots per day), S=5 (number of small time slots contained in each large time slot), W=2 (number of scenarios), specifically divided into "weekdays ( =0)” and “Non-working days ( =1), Workday N=5 (fixed storage days on workdays), Non-workday N=2 (fixed storage days on non-workdays), Work mode duty cycle Death voltage threshold ( =2.5V), threshold voltage ( =2.8V), Vmax =5.3V, written to the MCU's non-volatile memory cell.

[0150] Baseline library initialization: based on the current (Weekdays / Non-working days (weekends)) Read historical data under the same circumstances to construct a weekday energy matrix (5×96, where 5 is the fixed number of weekday storage days and 96 is the number of large time slots per day) and a non-working day energy matrix (2×96, where 2 is the fixed number of non-working day storage days). Calculate the initial baseline value for each time slot. The formula is as follows: (29)

[0151] Timer configuration: Set timer interrupts to trigger small time slot data acquisition every 3 minutes and large time slot prediction calculation every 15 minutes, adapting to dual-layer time slot prediction requirements.

[0152] Step 2: Energy-related data collection:

[0153] The energy acquisition and calculation of relevant data are performed at a frequency of 3 minutes per execution, and the real-time voltage of the supercapacitor is read. Synchronous reading of supercapacitor operating temperature Record the current context state label;

[0154] Step 3: Energy Sensing Calculation

[0155] The usable energy collected is directly calculated using the supercapacitor voltage, leveraging the square relationship between the supercapacitor's energy and voltage. Specific steps include:

[0156] Supercapacitor dynamic capacitance correction: Calculation using temperature correction model The dynamic capacitance of a supercapacitor;

[0157] Energy calculated directly from voltage: No current data collection is required; energy is calculated directly from the supercapacitor's energy-voltage relationship. ;

[0158] Combined with Energy consumed by time slot missions The time slot can be calculated. Collected usable energy ;

[0159] Step 4: Determine the hibernation control mechanism:

[0160] If we calculate the last 5 times The sum of available energy collected in time slots If the sleep strategy is triggered, the complex algorithm process such as fluctuation quantization and environmental correction is skipped, and the simplified prediction value of 0 is directly output; otherwise, continue to step 5.

[0161] Step 5: Contextual Unit Matching

[0162] Based on the current situation status label Match the corresponding context unit in the historical benchmark database and extract the corresponding... value;

[0163] Step 6: Dynamic Energy Prediction

[0164] Based on 15 minutes / large time slot Call the HERA algorithm to complete the next time slot. Energy prediction, specific steps:

[0165] Volatility Quantization Calculation: Take the first S=5 times within a time slot The normalized Shannon entropy is calculated based on the available energy collected in the time slots. Quantify the current degree of energy fluctuation;

[0166] Dynamic weight calculation: based on Dynamic weights are determined using linear interpolation. ,based on Environmental correction factors were determined by comparing actual energy values ​​with those from five similar time periods. ;

[0167] Error feedback correction: combining the current time slot prediction error Substitute the values ​​into the prediction formula to calculate the predicted value for the next time slot. ;

[0168] Step 7: Operating mode adjustment: The operating mode of the sensor node is controlled by the supercapacitor voltage in the current time slot and the predicted energy value available in the next time slot, so as to achieve a dynamic balance between energy supply and demand.

[0169] Step 8: Baseline Update

[0170] The historical baseline values ​​of the context unit are updated using the exponential moving average algorithm. ;

[0171] Step 9: Status Update

[0172] Once all time slots for the day are predicted, the actual energy values ​​for each time slot will be calculated. The data is categorized into scenario units and stored in the historical benchmark database. The historical average of each unit is recalculated to complete the benchmark database iteration. If the number of storage days for weekday / non-weekday scenarios reaches a fixed number (N=5 for weekdays and N=2 for non-weekdays), the most similar day (the day in the same scenario as the day to be determined) is deleted. The available energy data collected on the minimum day is used to store the available energy in the K time slots collected on that day into the corresponding scenario. Simultaneously, the algorithm's temporary variable cache is cleared, releasing resources to prepare for the next day's prediction.

[0173] Cyclic execution: At the start of the next prediction cycle, Step 2-Step 9 are repeated to provide effective energy prediction values ​​for the energy management of the self-powered wireless sensor, enabling the self-powered sensor to operate continuously and ensuring the long-term stable autonomy of the system.

[0174] In the above implementation process, energy consumption during energy storage and prediction is reduced through "supercapacitor voltage direct energy calculation" and "Hybrid Entropy Weight Rolling Adaptive Energy Prediction Algorithm (HERA)". The overall solution is suitable for indoor distributed sensing scenarios such as smart buildings and smart homes, providing technical support for the long-term stable operation of self-powered wireless sensors.

[0175] The operating mode of the sensor node is controlled by the supercapacitor voltage in the current time slot and the predicted energy value for the next time slot, thereby achieving a dynamic balance between energy supply and demand. The specific steps are as follows:

[0176] (1) Parameter description: Working mode: (duty cycle is) Death voltage threshold (2.5V), threshold voltage for activation (2.8V), target energy deviation Gap, system target voltage threshold (2.5V), the system has sufficient energy for the target voltage value. (4.0V), System energy insufficient target voltage value (3.6V), System target energy value , This represents the system's target energy value when energy is sufficient. This represents the system's target energy value when energy is insufficient. It is the set system target energy threshold;

[0177] (2) Hysteresis control: When the system voltage drops below When forced into a frozen, locked operating state (Level -1), all sensing, communication, and computing modules except the voltage monitoring unit are disconnected to prevent further energy depletion, and the sensor nodes enter a cold start state; only when the energy storage element voltage rises back to the second voltage threshold... At this point, the lock is released, allowing the system to enter a low-power operating mode. The strong hysteresis mechanism effectively eliminates the frequent start-stop 'oscillation' phenomenon caused by small energy fluctuations at the weak light critical point.

[0178] (3) Dynamic level adjustment: Based on the deviation between the predicted value of the next time slot in step S6 and the target energy required to maintain system operation. If Gap > 0 (surplus compensation): ,like (Deficit Protection): The task level is mapped to levels 0 to Q. This represents the maximum energy storage capacity of the supercapacitor. yes The death energy value is calculated using formula (3). Coefficient mk: F (aggressive control) is used when energy is being collected, and E (conservative control) is used when there is no energy collection (F>E). Level Q corresponds to the highest sampling frequency; level 0 corresponds to no-sampling sleep. Each level corresponds to a preset MCU duty cycle parameter. .

[0179] Compared with the prior art, the present invention has the following advantages:

[0180] (1) Strong scene adaptability: It is designed specifically for indoor low-light scenes. The algorithm mechanism is deeply adapted to the core characteristics of μW-level low energy density and high-frequency fluctuations. It achieves dynamic context matching (calling the historical benchmark library of the corresponding context). Data (avoiding cross-situation prediction bias) and energy fluctuation quantification (based on normalized Shannon entropy) (Determine the intensity of energy fluctuations) and accurately capture the energy change patterns caused by indoor LED light switching, temporary obstruction by people, etc., effectively solving the problem of poor adaptability of traditional algorithms in indoor low-light environments.

[0181] (2) Reduce computational energy consumption and improve prediction accuracy: The HERA algorithm proposed in this invention significantly reduces ineffective computational energy consumption through a sleep gating strategy; simultaneously, it eliminates the additional power consumption caused by current sampling by directly calculating the model using supercapacitor voltage. This is combined with context awareness (calling historical baseline energy values ​​corresponding to the context). ) and dynamic weighting (based on Linear interpolation adjusts dynamic weights Mechanisms such as [list of mechanisms] adapt to indoor energy fluctuation characteristics. The two work together to significantly improve prediction accuracy while ensuring low power consumption, providing technical support for the long-term stable operation of self-powered wireless sensors.

[0182] (3) Improve the energy efficiency control capability of nodes. Through the control architecture of "prediction + working mode adjustment", this invention not only solves the problem of oscillation of self-powered nodes at the critical point of weak light, but also realizes the function of actively adjusting the task load according to the ambient light situation. On the basis of ensuring the long-term stable autonomous operation of the system, it maximizes the utilization efficiency of micro energy.

[0183] 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 will understand that implementing all or part of the above embodiments and making equivalent changes in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A method for energy efficiency control of self-powered wireless sensor nodes in indoor low-light scenarios, characterized in that... Construct an energy harvesting and energy sensing computing model, including an energy harvesting module, an energy sensing computing module, an energy prediction module, and an energy management module; The energy harvesting module achieves efficient energy harvesting by adjusting the indoor solar panels using MPPT. The energy sensing and calculation module calculates available energy based on the supercapacitor voltage, and dynamically adjusts the capacitor capacity using a temperature correction model during the calculation process. The energy prediction module is based on the low energy density and high frequency fluctuation characteristics of indoor low-light scenarios. The model uses the hybrid entropy weight rolling adaptive energy prediction algorithm HERA for prediction. By distinguishing scenarios, implementing sleep control, quantifying energy fluctuations, and dynamically adjusting prediction weights according to energy fluctuations, it balances the accuracy of indoor energy prediction with energy consumption during energy harvesting. The energy management module adjusts the node's operating mode based on the predicted energy available in the next time slot and the current supercapacitor energy value; Model-based energy efficiency control of self-powered wireless sensor nodes: Step 1: System initialization, completing the calibration of each component in the system and configuring system parameters; Step 2: Collect data periodically, including the real-time voltage and operating temperature of the supercapacitor, and add context tags according to the current context type; Step 3, Energy Sensing Calculation: Input the operating temperature into the temperature dynamic correction model to obtain the dynamic capacity of the supercapacitor, and combine it with the voltage to calculate the usable energy collected; Step 4, sleep control mechanism judgment: If the sum of available energy collected in S consecutive time slots is lower than the threshold, the sleep gating strategy is triggered and a simplified prediction value is output; Otherwise, proceed to step 5; Step 5, Context Unit Matching: Based on available energy and context labels, match the corresponding context units in the historical benchmark library and extract the historical benchmark energy values; Step 6: Use the HERA algorithm to perform dynamic energy prediction: quantify the degree of energy fluctuation, calculate the normalized Shannon entropy of the energy values ​​of the last S time slots, and further obtain the dynamic weights through linear interpolation. Based on historical baseline energy values ​​and recent The actual energy values ​​of similar time periods are used to calculate the dynamic environment correction factor; combined with the current time slot prediction error, the energy prediction value of the next time slot is calculated. Step 7, Operating mode adjustment: The operating mode of the sensor node is controlled by the supercapacitor voltage in the current time slot and the predicted energy value available in the next time slot, so as to achieve a dynamic balance between energy supply and demand. Step 8, Historical reference energy value update: Calculate the new reference value based on the actual energy value of the current time slot and the historical reference energy value before the update; Step 9: After completing the daily forecast task, update the status: Store the available energy collected in all time slots of the day into the corresponding historical benchmark database. When the number of data days under the same scenario label reaches... When the new data is updated, delete the historical data of the day with the highest similarity to the new data, and update the historical data of the corresponding context unit.

2. The energy efficiency control method for self-powered wireless sensor nodes in indoor low-light scenarios as described in claim 1, characterized in that, In the energy harvesting and energy sensing calculation model: Task sequence is defined as ,in For the first Energy prediction task in one time slot, task The demand vector is defined as , The context type for the day of the task; for The voltage of the time-slot supercapacitor; for Operating temperature of time-slotted supercapacitors; dividing a day into K large time slots. , k∈[1,K], each large time slot Includes S hourly gaps , i∈[1,S]; The HERA algorithm parameter set is defined as follows: ,in, For the first Historical baseline energy values ​​for the time slot context; for The normalized Shannon entropy of a time slot represents the quantification of energy fluctuations; for Dynamic weighting of time slots; for Environmental correction factor for time slots; for Prediction error of time slots; This is the error correction factor; The sleep threshold; HERA's dynamic prediction logic is as follows: Dynamically adjust the weighting of real-time data and historical benchmarks. , combined and Achieve environmental adaptation and error correction, and improve prediction results. The expression is: ; ; Indicates time slot The usable energy collected.

3. The energy efficiency control method for self-powered wireless sensor nodes in indoor low-light scenarios as described in claim 2, characterized in that, Step 1, system initialization, includes: System calibration: Calibrate the MCU's built-in ADC and temperature sensor; Parameter configuration: Capacitance value of the supercapacitor at 25℃ Temperature coefficient HERA algorithm weight upper and lower bounds , Benchmark update coefficients Sleep threshold Error feedback coefficient Time slot value , Total number of time slots K, total number of hourly time slots S, total number of scenario types W, scenario types Write to the MCU; Baseline library initialization: based on the current Read historical data from the same context and construct an energy matrix. Context library : ; ; in, Indicates the first case under the corresponding scenario w. The usable energy collected in the Kth large time slot; Calculate the initial reference value for each time slot Represented as: ; in, This represents the number of days corresponding to the first scenario. For the first The number of days corresponding to each scenario, x∈[1,N], k∈[1,K]; Configure the timer according to the preset acquisition frequency.

4. The energy efficiency control method for self-powered wireless sensor nodes in indoor low-light scenarios as described in claim 3, characterized in that, Step 3, energy sensing calculation, includes: Using a temperature correction model to correct the dynamic capacitance of supercapacitors The expression is: ; Based on supercapacitors Energy stored in time slots The expression is: ; When the stored energy exceeds the maximum capacity of the supercapacitor At present The energy stored in a time-slot supercapacitor is defined as ; Calculate by combining sampling and sleep power consumption. Average power consumption per time slot The expression is: ; in For the sensor at the sampling rate Sampling power consumption, This refers to the power consumption when the node is in sleep mode. for Sampling rate within time slot Duty cycle; calculate Time slots consume energy The expression is: ; Time slot Collected usable energy The expression is: ; In the formula, for The total energy consumed by the sensor within the time slot; Available energy collected in time slots The expression is: 。 5. The energy efficiency control method for self-powered wireless sensor nodes in indoor low-light scenarios as described in claim 4, characterized in that, The step 4 sleep control mechanism judgment includes: 。 6. The energy efficiency control method for self-powered wireless sensor nodes in indoor low-light scenarios as described in claim 5, characterized in that, Step 5, context unit matching, includes: Based on the day's context label Match the corresponding context unit E in the historical benchmark library. w Extract the corresponding value.

7. The energy efficiency control method for self-powered wireless sensor nodes in indoor low-light scenarios as described in claim 6, characterized in that, Step 6, which uses the HERA algorithm for dynamic energy prediction, includes: Volatility Quantization Calculation: Take the first S times within a time slot The usable energy collected in the time slots is used to calculate the normalized Shannon entropy. The expression is: ; in for Probability distribution of time slot energy sample values for In S The number of times it appears in the available energy values ​​collected in the time slot. ; Dynamic weight calculation: based on Dynamic weights are determined using linear interpolation. ,based on Environmental correction factors were determined by comparing actual energy values ​​from nearly H similar time periods. : ; ; ; ; in, This is a vector of ratios between actual values ​​and benchmark values. To increase the time weight, Number of similar time periods; Error feedback correction: combining the current time slot prediction error Substitute the values ​​into the prediction formula to calculate the predicted value for the next time slot. ,error The expression is: ; in, For the first The predicted value of the time slot, For the first The actual value of the time slot.

8. The energy efficiency control method for self-powered wireless sensor nodes in indoor low-light scenarios as described in claim 7, characterized in that, In step 7, the operating mode of the wireless sensor node Based on the supercapacitor voltage and predicted values, the operating mode to be used in the next time slot is determined and executed. The specific method for determining the operating mode is as follows: The supercapacitor voltage is below the death voltage threshold. When forced into a frozen lockout working mode Level=-1, the wireless sensor node enters a cold start state. The supercapacitor voltage has recovered to the turn-on voltage threshold. When this happens, the lock is released and the system is allowed to enter a low-power operating mode (Level=0) for energy prediction. In low-power operating mode, according to The operating mode of the wireless sensor node at the next moment is determined as mode L, where L∈[0,Q]: Calculate the target energy value E of the system target : ; In the formula, This represents the system's target energy value when energy is sufficient. This represents the system's target energy value when energy is insufficient. It is the set system target energy threshold; Calculate the gap between the predicted value for the next time slot and the target energy required to maintain system operation: ; L is determined based on the gap: ,Gap>0; ,Gap≤0; In the formula, This represents the energy of the supercapacitor in the current time slot, and round() is the rounding function. This represents the maximum energy storage capacity of the supercapacitor. It is the death energy value; The coefficient mk is taken as F when there is energy harvesting and as E when there is no energy harvesting, where F>E; 1...Q represents the nodes according to different duty cycles. Data collection is performed with the highest sampling frequency when L=Q.

9. The energy efficiency control method for self-powered wireless sensor nodes in indoor low-light scenarios as described in claim 8, characterized in that, Step 8, updating the historical baseline energy value, includes: The historical baseline values ​​of the context unit are updated using the exponential moving average algorithm. The formula is: ; in, The historical baseline energy value before the update. This is a new baseline value that incorporates the current real-time energy values; Step 9, after completing the daily forecast task, includes updating the status, which includes: Once all time slots for the day are predicted, the actual energy values ​​for each time slot will be calculated. Store the data in the context unit corresponding to the context tag, recalculate the historical baseline value of each context unit, and complete the update of the historical baseline library; Determine whether the number of storage days for a context unit with the corresponding context label for the current day has reached the corresponding fixed storage number N. If so, delete the available energy data collected on the day most similar to the current day's data, and store the available energy data from the K time slots collected on the current day into the corresponding context unit. The difference in overall energy scale between the day to be determined and historical days under the same context is quantified by the daily average energy deviation rate. ; ; in, The average daily usable energy collected on the day to be determined. This represents the average daily usable energy collected under the same historical circumstances. , These are the available energy values ​​collected in large time slots for the date to be determined and historical dates under the same circumstances, respectively. The daily average energy deviation rate is selected from historical days with the same conditions as the day to be determined, using the minimum ( () as the most similar day.

10. A self-powered wireless sensor node energy efficiency control system for indoor low-light scenarios, characterized in that, The system for implementing the energy efficiency control method for self-powered wireless sensor nodes in indoor low-light scenarios as described in any one of claims 1-9 includes: The system comprises an energy harvesting module, an energy sensing and calculation module, an energy prediction module, and an energy management module. The energy harvesting module achieves efficient energy collection by adjusting indoor solar panels using MPPT (Multi-Phase Power Theory). The energy sensing and calculation module calculates available energy based on the supercapacitor voltage, dynamically adjusting the capacitor capacity using a temperature correction model during the calculation process. The energy prediction module, based on the low energy density and high-frequency fluctuation characteristics of indoor low-light scenarios, employs the HERA (Hybrid Entropy Weighted Rolling Adaptive Energy Prediction) algorithm for prediction. It balances indoor energy prediction accuracy with energy consumption during energy harvesting by differentiating scenarios, implementing sleep control, quantifying energy fluctuations, and dynamically adjusting prediction weights based on energy fluctuations. The energy management module adjusts the node's operating mode based on the predicted available energy for the next time slot and the current supercapacitor energy value. The computing unit is used for energy efficiency control of self-powered wireless sensor nodes based on models, calculating the energy prediction value for the next time slot, and controlling the working mode of the nodes.

Images