Method and system for improving energy utilization rate by utilizing light storage resource prediction

Abstract

The invention discloses a method and system for improving the energy utilization rate by utilizing light storage resource prediction, and the method comprises the steps: collecting irradiance temperature historical output data and load consumption feature data through a distributed sensor and an edge gateway, processing time sequence features through a sequence prediction model, and integrating multi-dimensional input through the fusion of an integrated learning model; a future prediction result of photovoltaic output and load demand is obtained; further acquiring real-time parameters of the charge state and the charge and discharge efficiency of the energy storage equipment, constructing an energy storage full life cycle state evaluation model to calculate dynamic indexes, and determining an available capacity boundary; and if the boundary exceeds a preset threshold value, peak-valley period characteristics are extracted, and an optimal strategy sequence including photovoltaic peak charging and load peak discharging is generated by adopting an intelligent optimization algorithm in combination with power grid peak-valley electricity prices and carbon emission constraints. According to the invention, the photovoltaic consumption rate is obviously improved, the operation cost is reduced, the carbon emission requirement is met, and intelligent and efficient management of the optical storage system is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of light energy and energy storage technology, and particularly discloses a method and system for predicting and improving energy utilization rate using light storage resources. BACKGROUND

[0002] Under the background of energy transformation and green development, the development and utilization of new energy have become an important pillar to promote the sustainable development of society. In particular, the combination of light energy and energy storage technology is considered a key path to solve unstable energy supply and improve energy utilization efficiency. This combination of technologies not only reduces dependence on traditional fossil fuels, but also plays an important role in balancing supply and demand in the power system and reducing environmental impact. However, how to efficiently integrate these resources to maximize their value in practical applications remains a major issue that needs to be addressed.

[0003] Although existing energy management systems have achieved some degree of scheduling of light energy and energy storage resources, they generally lack the ability to adapt to dynamic changes in resources. Many methods often fail to respond promptly to the dramatic fluctuations in light energy output influenced by weather, and it is also difficult to make reasonable decisions in the face of complex energy storage device states. This limitation leads to energy waste and low system efficiency, especially when faced with complex electricity demand and external environmental changes, the system often appears to be inadequate, making it difficult to achieve precise matching and efficient use of resources.

[0004] The deeper technical difficulty lies in how to make advance judgments and dynamic adjustments of light energy output and electricity demand. This core factor directly determines whether resources can be reasonably allocated and fully utilized. Since the output of light energy is influenced by weather, time and other factors, and electricity demand changes constantly with user behavior and time period, lack of accurate grasp of future trends often leads to imbalance in resource allocation. For example, in the case of sudden weather changes, the system may fail to predict a sudden decrease in light energy output in advance, resulting in energy storage devices failing to replenish power in time, ultimately causing some electricity demand to be unsatisfied, or failing to effectively store excess light energy, resulting in wasted resources.

[0005] Therefore, how to accurately predict the future trends of light energy output and electricity demand in a complex environment of resource fluctuations and demand changes, and make dynamic resource scheduling based on this, has become a key issue to improve energy utilization efficiency. SUMMARY

[0006] The present application provides a method and system for predicting and improving energy utilization rate using light storage resources, aiming to solve at least one of the defects in the prior art.

[0007] One aspect of the present application relates to a method for predicting the improvement of energy utilization rate by using light storage resources, comprising the following steps: S100, collecting irradiance temperature history output data and load consumption characteristic data from photovoltaic equipment and meteorological sources through distributed sensors and edge gateways, processing time series characteristics using a sequence prediction model, integrating multi-dimensional input data using an ensemble learning model, and obtaining photovoltaic output and load demand future prediction results; S200, obtaining real-time parameters of state of charge and charge-discharge efficiency of energy storage equipment according to the photovoltaic output and load demand future prediction results, constructing an energy storage full life cycle state evaluation model, calculating dynamic state indicators, and determining the energy storage available capacity boundary; S300, if the energy storage available capacity boundary exceeds the preset threshold, extracting peak-valley period features from the photovoltaic output and load demand future prediction results, combining the peak-valley electricity price and carbon emission constraints, and using an intelligent optimization algorithm to generate an optimal charge-discharge strategy sequence, wherein the optimal charge-discharge strategy sequence includes photovoltaic peak charging instructions and load peak discharging instructions; S400, transmitting the optimal charge-discharge strategy sequence to the edge controller through the remote management platform, and executing real-time instructions through the photovoltaic inverter and energy storage converter to realize the collaborative response of the photovoltaic storage equipment and the controllable equipment on the load side, and obtaining the operation data after the collaborative response of the photovoltaic storage equipment and the controllable equipment on the load side; S500, extracting the absorption rate index from the operation data after the collaborative response, updating the multi-dimensional fusion prediction model parameters using a feedback mechanism, obtaining the optimized prediction accuracy, and using it for subsequent photovoltaic output and load demand prediction.

[0008] Further, step S100 comprises: S110, obtaining irradiance data, temperature data and historical output data from photovoltaic equipment and meteorological sources through distributed sensors and edge gateways, and collecting load consumption characteristic data from the load end, extracting time series characteristics from the irradiance data, temperature data, historical output data and load consumption characteristic data, filtering abnormal values using a preset threshold, and obtaining cleaned time series data sets; S120, according to the cleaned time series data set, using a sequence prediction tool to predict the future trend of photovoltaic output data and load demand data, smoothing the volatility characteristics in the prediction results, and determining the preliminary prediction data results; S130, fusing the preliminary prediction data results with multi-dimensional input data through an ensemble learning tool, obtaining a comprehensive data set containing irradiance, temperature and load consumption characteristics, if the deviation of a certain dimension data in the comprehensive data set exceeds the preset threshold, performing secondary correction on the dimension data exceeding the preset threshold, and determining the adjusted comprehensive prediction data set; S140, generating a prediction curve of future photovoltaic output and load demand by using a data visualization tool according to the adjusted comprehensive prediction dataset, labeling key nodes of the prediction curve to obtain a future prediction result of photovoltaic output and load demand.

[0009] Further, the step S200 comprises: S210, obtaining a real-time parameter dataset of the state of charge and the charging and discharging efficiency of the energy storage device from the energy storage device, combining historical operation data and environmental temperature influence records, filtering abnormal values by using a data cleaning tool to obtain the real-time parameter dataset after arrangement according to the future prediction result of photovoltaic output and load demand; S220, obtaining the charging and discharging cycle number and the capacity attenuation characteristic of the energy storage device according to the real-time parameter dataset after arrangement, and if a parameter in the charging and discharging cycle number and the capacity attenuation characteristic exceeds a preset threshold, marking the parameter exceeding the preset threshold to determine an abnormal state parameter set; S230, obtaining historical data related to the health of the energy storage device for the abnormal state parameter set, extracting features of the historical data related to the health of the energy storage device by using a time series processing tool to determine a dynamic state index range; S240, calibrating by using a data fusion tool to obtain a final available capacity boundary result by combining the capacity attenuation characteristic of the energy storage device with the dynamic state index range.

[0010] Further, the step S300 comprises: S310, if the available capacity boundary of the energy storage exceeds a preset threshold, obtaining distribution details of peak-valley period characteristics from the future prediction result of photovoltaic output and load demand, classifying the distribution details of the peak-valley period characteristics by using a data arrangement tool to obtain a divided peak-valley period interval set; S320, for the peak-valley period interval set, combining the peak-valley electricity price of the power grid and the carbon emission constraint to perform period priority ranking by using a rule matching tool to determine a period priority list suitable for charging and discharging; S330, if the charging period in the period priority list matches the photovoltaic peak, generating a photovoltaic peak charging instruction, and if the discharging period matches the load peak, generating a load peak discharging instruction to obtain a preliminary charging and discharging instruction combination; S340, adjusting the instruction by using a data calibration tool to determine an optimal charging and discharging strategy sequence by combining the state of the energy storage device and the available capacity boundary of the energy storage device with the preliminary charging and discharging instruction combination.

[0011] Further, the step S400 comprises: S410, after transmitting the optimal charging and discharging strategy sequence to the edge controller through the remote management platform, the optimal charging and discharging strategy sequence is analyzed and processed by using a data distribution tool to obtain a decomposed charging and discharging instruction set; S420, according to the decomposed charging and discharging instruction set, the photovoltaic inverter and the energy storage converter are linked, the corresponding instructions are distributed to the corresponding equipment by using a real-time instruction issuing tool, and the execution state of the light storage equipment and the load side controllable equipment is determined; S430, if the execution state of the light storage equipment and the load side controllable equipment meets the preset threshold, the running data after the cooperative response is obtained from the equipment running log by using a data acquisition tool.

[0012] Further, step S500 comprises: S510, according to the running data after the cooperative response, the corresponding consumption rate index value is calculated by using a data processing tool, if the consumption rate index value is lower than the preset threshold, an adjustment signal is generated by using a feedback tool, and the direction of feedback adjustment is determined; S520, the adjustment signal is transmitted to the prediction parameter database by using the direction of feedback adjustment, the adjustment signal is corrected by using a parameter updating tool, and the optimized prediction accuracy is obtained; S530, according to the optimized prediction accuracy, the prediction basis suitable for the subsequent period is generated by using a data integration tool combined with historical running data, and whether the prediction basis meets the load demand is judged.

[0013] Another aspect of the present application relates to a system for predicting and improving energy utilization rate by using light storage resources, for executing the method for predicting and improving energy utilization rate by using light storage resources described above, comprising: A photovoltaic output and load demand future prediction result acquisition module is configured to collect irradiance temperature historical output data and load consumption characteristic data from photovoltaic equipment and meteorological sources through distributed sensors and edge gateways, process time sequence characteristics by using a sequence prediction model, integrate multi-dimensional input data by using an ensemble learning model, and obtain photovoltaic output and load demand future prediction results. An energy storage available capacity boundary determination module is configured to obtain energy storage equipment state of charge and real-time parameters of charging and discharging efficiency according to the photovoltaic output and load demand future prediction results, construct an energy storage full life cycle state evaluation model, calculate dynamic state indexes, and determine the energy storage available capacity boundary. An optimal charging and discharging strategy sequence generation module is configured to extract peak-valley period characteristics from the photovoltaic output and load demand future prediction results if the energy storage available capacity boundary exceeds a preset threshold, combine grid peak-valley electricity price and carbon emission constraints, and generate an optimal charging and discharging strategy sequence by using an intelligent optimization algorithm, wherein the optimal charging and discharging strategy sequence comprises photovoltaic peak charging instructions and load peak discharging instructions. The cooperative response operation data acquisition module is configured to transmit the optimal charging and discharging strategy sequence to the edge controller through the remote management platform, and link the photovoltaic inverter and the energy storage converter to execute real-time instructions, so that the photovoltaic and energy storage equipment and the controllable equipment on the load side are cooperatively responsive, and operation data after the cooperative response of the photovoltaic and energy storage equipment and the controllable equipment on the load side is acquired. The prediction accuracy acquisition module is configured to extract the consumption rate index from the operation data after the cooperative response, update the multi-dimensional fusion prediction model parameters by using a feedback mechanism, obtain the optimized prediction accuracy, and predict the photovoltaic output and load demand in a subsequent period.

[0014] The present application has the following beneficial effects: The present application provides a method and system for predicting and improving energy utilization rate by using photovoltaic and energy storage resources, which solves the problems of unstable energy storage capacity boundary and low consumption rate caused by photovoltaic output volatility and load demand uncertainty. The problems further integrate economic and environmental optimization requirements under the constraints of peak-valley electricity price and carbon emission, forming a logically related comprehensive challenge, i.e. how to dynamically evaluate the state of energy storage and generate the optimal charging and discharging strategy to achieve efficient cooperative response. The present application collects irradiance temperature historical output data and load consumption characteristic data through distributed sensors and edge gateways, processes time series characteristics by using a sequence prediction model, integrates multi-dimensional input by using an ensemble learning model, and obtains future prediction results of photovoltaic output and load demand. Then, real-time parameters of energy storage device state of charge and charging and discharging efficiency are acquired, a full life cycle state evaluation model of energy storage is constructed to calculate dynamic indicators and determine the available capacity boundary. If the boundary exceeds the preset threshold, the peak-valley period characteristics are extracted, the optimal strategy sequence including photovoltaic peak charging and load peak discharging is generated by combining the constraints of peak-valley electricity price and carbon emission and using an intelligent optimization algorithm. The optimal strategy sequence is transmitted to the edge controller through the remote management platform, and the photovoltaic inverter and the energy storage converter are linked to execute real-time instructions, so that the photovoltaic and energy storage equipment and the controllable equipment on the load side are cooperatively responsive. Finally, the consumption rate index is extracted from the operation data after the response, the prediction model parameters are updated by using a feedback mechanism, the prediction accuracy after optimization is improved, and the prediction is used for subsequent periods. The technical effect of the present application is to significantly improve the photovoltaic consumption rate, reduce the operating cost, meet the carbon emission requirements, and realize intelligent and efficient management of the photovoltaic and energy storage system. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 FIG. 1 is a flowchart of an embodiment of the system for predicting and improving energy utilization rate by using photovoltaic and energy storage resources according to the present application; Figure 2 FIG. 2 is a functional block diagram of an embodiment of the system for predicting and improving energy utilization rate by using photovoltaic and energy storage resources according to the present application.

[0016] DETAILED DESCRIPTION OF DRAWINGS 10. Module for obtaining future forecast results of photovoltaic power output and load demand; 20. Module for determining the available capacity boundary of energy storage; 30. Module for generating the optimal charging and discharging strategy sequence; 40. Module for obtaining coordinated response operation data; 50. Module for obtaining prediction accuracy. Detailed Implementation

[0017] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0018] like Figure 1 As shown, the first embodiment of the present invention proposes a method for improving energy efficiency by utilizing photovoltaic and energy storage resources, comprising the following steps: Step S100: Collect historical irradiance and temperature data and load consumption characteristic data from photovoltaic equipment and meteorological sources through distributed sensors and edge gateways. Process the time series characteristics using a sequence prediction model and integrate the multi-dimensional input data using an ensemble learning model to obtain future prediction results of photovoltaic output and load demand.

[0019] By deploying distributed sensors (such as irradiance sensors, temperature sensors, and power sensors) and edge gateways at photovoltaic power plants, two types of core data are collected simultaneously: first, irradiance, operating temperature, historical output data (such as hourly power generation over the past 72 hours), and meteorological data (such as weather forecasts for sunny / rainy conditions and wind speed); second, consumption characteristic data on the load side (such as the time-of-day distribution of electricity load, power fluctuation amplitude, and historical peak electricity consumption). A sequence prediction model adapted to the characteristics of energy data (such as LSTM (Long Short-Term Memory) prediction) is employed. The system utilizes Memory (Long Short-Term Memory Network), Transformer, or Prophet models to mine time-series features (such as diurnal rhythms, seasonal variations, and trend and periodic patterns) from data, transforming them into quantifiable feature vectors. It integrates multi-dimensional input data (including time-series feature vectors, meteorological influencing factors, and load consumption characteristic parameters) through ensemble learning models (such as gradient boosting tree ensemble and random forest fusion frameworks), reducing the prediction bias of a single model through weighted fusion. Finally, it outputs future predictions of photovoltaic power output (such as hourly power generation predictions for the next 24 / 48 hours) and load demand (such as hourly electricity load predictions for the next 24 / 48 hours), requiring a photovoltaic power output prediction accuracy of ≥90% and a load demand prediction accuracy of ≥92% (root mean square error ≤8%), providing reliable prediction basis for subsequent energy storage status assessment and charging / discharging strategy formulation.

[0020] Step S200, according to the photovoltaic output and load demand future prediction results, the state of charge of the energy storage device and the real-time parameters of the charge and discharge efficiency are obtained, the energy storage full life cycle state evaluation model is constructed, the dynamic state index is calculated, and the available capacity boundary of the energy storage is determined.

[0021] With the photovoltaic output and load demand future prediction results (such as future 24-hour supply and demand trend) output by step S100 as reference, the key operation parameters of the energy storage device are obtained in real time through sensors and edge collection modules. The key operation parameters of the energy storage device include the current state of charge (such as SOC (State Of Charge, remaining capacity ratio of total capacity)), charge and discharge efficiency (such as charge efficiency ≥ 90%, discharge efficiency ≥ 88%), battery cell voltage / temperature, cycle number, etc. Based on these parameters, the energy storage full life cycle state evaluation model (combining battery attenuation model, thermal management model and reliability evaluation algorithm) is constructed, the dynamic state index (such as remaining cycle life, real-time charge and discharge power upper limit, attenuation rate, safe operation temperature interval) is calculated; combined with the dynamic state index and the load pressure of the future supply and demand prediction, the available capacity boundary (i.e. maximum safe charging capacity, maximum safe discharging capacity, which needs to avoid equipment damage caused by overcharging and overdischarging) of the energy storage device in the current and future period is determined, the state evaluation accuracy is required to be ≥ 95%, the available capacity boundary calculation error is ≤ 5%, and the "capacity matching" constraint condition is provided for the subsequent charge and discharge strategy generation.

[0022] Step S300, if the available capacity boundary of the energy storage exceeds the preset threshold, the peak and valley period features are extracted from the photovoltaic output and load demand future prediction results, combined with the peak and valley electricity price and carbon emission constraints, the intelligent optimization algorithm is adopted to generate the optimal charge and discharge strategy sequence, wherein the optimal charge and discharge strategy sequence includes photovoltaic peak charging instruction and load peak discharging instruction.

[0023] First, determine whether the energy storage available capacity boundary determined in step S200 exceeds the preset threshold (e.g., available capacity ≥ 20% of total capacity, ensuring that there is a dispatch adjustment space); if the condition is met, extract the peak and valley period characteristics (e.g., photovoltaic output peak period 9:00-15:00, load power consumption peak period 7:00-9:00 and 18:00-22:00, and grid load valley period 0:00-6:00) from the photovoltaic output and load demand future prediction results in step S100; with the goal of "improving photovoltaic consumption rate, reducing electricity cost, and reducing carbon emissions", incorporate the grid peak-valley price difference (charging during low-valley price period, discharging for profit during peak price period) and carbon emission constraint (preferentially consuming photovoltaic clean energy, reducing carbon emissions from fossil energy supplement), and construct a multi-objective optimization function; use an intelligent optimization algorithm (such as particle swarm optimization algorithm, genetic algorithm, reinforcement learning scheduling model) suitable for energy scheduling scenarios to solve the optimization function, generate a time-sequential optimal charging and discharging strategy sequence, and the optimal charging and discharging strategy sequence includes photovoltaic peak period energy storage charging instructions (e.g., start charging at 9:00, stop at 15:00, charge to 80% of the upper limit of available capacity), load peak period energy storage discharging instructions (e.g., start discharging at 18:30, stop at 22:00, ensure load power supply), and low-valley period charging and discharging adjustment instructions (e.g., a small amount of energy supplement during grid low-valley price period); the strategy execution requires that the photovoltaic consumption rate be increased by ≥15%, the electricity cost be reduced by ≥10%, and the carbon emissions be reduced by ≥8%, providing precise instruction basis for photovoltaic storage equipment collaborative response.

[0024] Step S400, transmit the optimal charging and discharging strategy sequence to the edge controller through the remote management platform, link photovoltaic inverters and energy storage converters to execute real-time instructions, realize photovoltaic storage equipment and load side controllable equipment collaborative response, and obtain operation data after photovoltaic storage equipment and load side controllable equipment collaborative response.

[0025] The optimal charging and discharging strategy sequence (including charging, discharging and timing adjustment instructions) generated in step S300 is input, and the instructions are encrypted and transmitted to the edge controller of the optical storage system through the remote management platform (supporting wireless / wired communication protocols such as MQTT (Message Queuing Telemetry Transport, Message Queuing Telemetry Transport Protocol), Modbus); the edge controller serves as a local execution core, links photovoltaic inverters (adjusting the grid-connected / energy storage distribution ratio of photovoltaic output) and energy storage converters (accurately executing charging and discharging power and duration instructions), and simultaneously issues coordinated control instructions (such as peak shifting and load shifting) to controllable devices on the load side (such as flexible industrial loads, intelligent charging piles, and adjustable air conditioners), realizing the linkage response of photovoltaic devices, energy storage devices and controllable devices on the load side, and ensuring the consistency of strategy execution; in the process of coordinated response, through distributed sensors and data acquisition modules, real-time device operation data (including photovoltaic actual output, energy storage charging and discharging capacity / SOC change, load actual power consumption, device operating temperature / efficiency, etc.) are obtained, the instruction transmission delay is required to be ≤200ms, and the device coordinated response accuracy is required to be ≥99%, providing real and complete operation feedback data for subsequent prediction model optimization.

[0026] In step S500, the absorption rate index is extracted from the operation data after the coordinated response, and the multi-dimensional fusion prediction model parameters are updated using a feedback mechanism to obtain an optimized prediction accuracy for subsequent photovoltaic output and load demand prediction.

[0027] From the operation data of the photovoltaic and energy storage devices and the controllable devices on the load side after the coordinated response, key absorption rate indicators (including photovoltaic self-use absorption rate, energy storage charging and discharging matching absorption rate, and light rejection rate) are extracted, and deviation analysis (such as calculating the difference between predicted output and actual output, and the root mean square error between predicted load and actual load) is performed on these actual indicators and the prediction results in step S100; based on the deviation analysis results, a feedback mechanism is constructed, and the core parameters of the multi-dimensional fusion prediction model (including the weight coefficients of the sequence prediction model, the fusion ratio of the ensemble learning model, and the influence weight of the meteorological factor) are adjusted through a parameter updating algorithm (such as gradient descent method and online learning update strategy); the prediction deviation is reduced through parameter iteration optimization, and an optimized prediction accuracy (such as the photovoltaic output prediction accuracy being improved from 90% to more than 93%) is obtained; the optimized model is directly used for photovoltaic output and load demand prediction in the subsequent period (such as the next 24 hours), forming a virtuous cycle of "prediction-execution-feedback-optimization", and the prediction error is required to be reduced by ≥5% after each iteration, and the model adaptability continuously fits the actual operation scenario.

[0028] Further, the method for predicting and improving energy utilization rate by using optical storage resources provided by the embodiment comprises the following steps: In step S110, irradiance data, temperature data and historical output data are acquired from photovoltaic devices and meteorological sources through distributed sensors and edge gateways, and load consumption characteristic data is collected from the load end. Time series feature extraction is performed on the irradiance data, temperature data, historical output data and load consumption characteristic data. A preset threshold is used to filter outliers, and a cleaned time series dataset is obtained.

[0029] The outlier filtering mechanism based on the preset threshold is described by the following formula:

[0030] wherein, represents the cleaned time series data, represents the original collected sensor data, represents the mean value of the data in the sliding window, represents the standard deviation of the data in the sliding window, represents the preset threshold coefficient. When the original data deviates from the mean value by more than times the standard deviation, it is determined to be an outlier and is replaced by the mean value.

[0031] The distributed sensor multi-source data fusion processing process is described by the following formula: (1) In formula (1), represents the fused multi-dimensional data vector, represents the irradiance data, represents the temperature data, represents the historical output data, represents the load consumption characteristic data, represents the transpose, represents the sensor weight matrix, represents the bias correction vector of the edge gateway, which is used to coordinate the data differences of different sensor nodes.

[0032] The irradiance data, temperature data, and historical output data are collected from photovoltaic devices and meteorological sources through distributed sensors and edge gateways, while load consumption characteristic data is collected from the load end. This data collection process can be understood as real-time monitoring of the operating status of the photovoltaic power station and environmental factors. Assuming that a medium-sized photovoltaic power station has deployed multiple Internet of Things sensors installed on solar panels, the sensors collect irradiance data every hour, for example, the irradiance may reach 1000 watts per square meter under sunny conditions, while the temperature data is measured by a thermistor to measure the panel surface temperature, which may exceed 50 degrees Celsius during the summer peak. Historical output data is extracted from the power station's database, covering the power generation records of the past year, such as a monthly average output of 500 kilowatt-hours. At the same time, load consumption characteristic data is collected from smart meters, recording the power consumption patterns of factories or residential areas, such as peak load demand of 800 kilowatts. This collection method ensures data diversification, providing a comprehensive basis for subsequent analysis.

[0033] Time series feature extraction is performed on the irradiance data, temperature data, historical output data, and load consumption characteristic data, and a preset threshold is used to filter outliers to obtain a cleaned time series dataset. Time series feature extraction involves using Fourier transform or wavelet analysis to identify periodic patterns in the data, such as extracting daily cycle features from irradiance data from sunrise to sunset, and extracting seasonal fluctuations from temperature data, such as the trend of output decline due to low temperature in winter. When filtering outliers, the preset threshold can be set to three times the standard deviation of the data mean, for example, if an abnormal point of more than 2000 watts per square meter appears in the irradiance data, it is considered a sensor failure and is removed, resulting in a clean time series dataset that may contain thousands of time points to ensure the accuracy of the prediction. Based on the cleaned time series dataset, a sequence prediction tool is used to predict future trends of photovoltaic output data and load demand data, and the volatility features in the prediction results are smoothed to determine the preliminary prediction data results.

[0034] Step S120, based on the cleaned time series dataset, a sequence prediction tool is used to predict future trends of photovoltaic output data and load demand data, and the volatility features in the prediction results are smoothed to determine the preliminary prediction data results.

[0035] The time series prediction model formula for photovoltaic output and load demand is defined to capture the trend and periodic features of the time series data: (2) In formula (2), represents the predicted value at the future th time, represents the autoregressive coefficient, denotes the moving average coefficient, denotes the autoregressive order, denotes the prediction value at the past time point, denotes the prediction error term at the past time point, denotes the moving average order, denotes the prediction error term at the future time point.

[0036] The exponential smoothing formula is defined to smooth the volatility of the prediction results, which reduces the random fluctuations of the prediction data through weighted averaging: (3) In formula (3), denotes the smoothed value at the time point, denotes the original prediction data at the time point, denotes the smoothing coefficient with a value range of 0 to 1, denotes the trend component at the previous time point.

[0037] The mean absolute percentage error formula is defined to evaluate the accuracy of the prediction results, which quantifies the precision of the prediction model and verifies the reliability of the preliminary prediction data results: (4) In formula (4), denotes the mean absolute percentage error, denotes the total number of prediction data points, denotes the actual observation value at the time point, denotes the prediction value at the time point.

[0038] The sequence prediction tool can use the LSTM neural network model, which learns from historical data sequences to predict the photovoltaic output in the next 24 hours, for example, inputting the past week's irradiance and temperature data, the model outputs the output curve of the next day, the prediction value may show that the output peak at 9 am is 600 kilowatts.

[0039] The smoothing process uses the moving average algorithm to process short-term fluctuations in the prediction curve, such as averaging every 5-minute peak value to obtain a more stable preliminary prediction result, which helps to reduce the impact of noise on decision-making.

[0040] Step S130, the preliminary prediction data result is fused with the multi-dimensional input data through the integrated learning tool to obtain a comprehensive data set containing irradiance, temperature and load consumption characteristics, if the deviation of a certain dimension data in the comprehensive data set exceeds the preset threshold, the dimension data exceeding the preset threshold is subjected to secondary correction, and the adjusted comprehensive prediction data set is determined.

[0041] The integrated learning data fusion formula is defined as follows: (5) In formula (5), represents the fused comprehensive data set, represents the preliminary prediction data result, represents the irradiance input data, represents the temperature input data, represents the load consumption characteristic data, , , , respectively represent the weight coefficients of each dimension data.

[0042] The data deviation judgment formula is defined as follows: (6) In formula (6), represents the deviation judgment result, represents the observed dimension data value, represents the expected dimension data value, represents the standardized threshold parameter, represents the preset threshold, when the calculation result is greater than the secondary correction is triggered.

[0043] The secondary correction formula is defined as follows: (7) In formula (7), represents the corrected data value, represents the original dimension data exceeding the threshold, represents the target correction value, represents the correction intensity coefficient, represents the natural constant, represents the decay coefficient, represents the correction iteration number, and the gradual data correction is realized through the exponential decay function.

[0044] Ensemble learning tools are used to fuse the preliminary prediction data with multi-dimensional input data to obtain a comprehensive dataset containing irradiance, temperature, and load consumption characteristics. If the deviation of any dimension in the comprehensive dataset exceeds a preset threshold, secondary correction is performed on the data in that dimension to determine the adjusted comprehensive prediction dataset. For example, ensemble learning tools such as the random forest algorithm fuse preliminary predictions with real-time meteorological data to generate a comprehensive dataset. The irradiance dimension may show deviations. If the deviation between the predicted and actual irradiance exceeds 10%, secondary correction is performed using linear interpolation methods. Finally, the dataset is adjusted to improve the overall prediction accuracy.

[0045] Step S140: Based on the adjusted comprehensive forecast dataset, use data visualization tools to generate forecast curves for future photovoltaic power output and load demand, mark the key nodes of the forecast curves, and obtain the future forecast results for photovoltaic power output and load demand.

[0046] The future forecast of photovoltaic power output is obtained using the following formula: (8) In formula (8), Indicates time The predicted value of photovoltaic power output, Indicates the number of prediction models. Indicates the first The weighting coefficients of each prediction model Indicates the first A prediction model at time... The output value, Indicates time The prediction error term.

[0047] The future load demand forecast is derived using the following formula: (9) In formula (9), Indicates time The predicted load demand value, This represents the weighting factor for the base load. Indicates time The base load value, Indicates the temperature effect coefficient. Indicates time Temperature value, Indicates the humidity influence coefficient. This represents the humidity value at time t. This represents a random disturbance term.

[0048] According to the adjusted comprehensive prediction data set, a data visualization tool is used to generate a prediction curve of future photovoltaic output and load demand, key nodes of the prediction curve are labeled, and future prediction results of photovoltaic output and load demand are obtained. The data visualization tool, such as Tableau software, draws the output curve to display the future power generation trend, and labels the key nodes such as the load peak time point, for example, labels the output and demand matching point at 2 pm. This visualization result helps energy managers optimize scheduling, achieve supply and demand balance, and reduce energy waste.

[0049] Further, the method for predicting and improving energy utilization rate by using light storage resources provided by the embodiment comprises the following steps: Step S210, according to the future prediction results of photovoltaic output and load demand, real-time parameters of state of charge and charging and discharging efficiency of the energy storage device are obtained from the energy storage device, historical operation data and environmental temperature influence records are combined, a data cleaning tool is used to filter abnormal values, and a sorted real-time parameter data set is obtained.

[0050] The real-time state of charge after data cleaning is obtained by the following formula: (10) In formula (10), represents the real-time state of charge after data cleaning, represents the number of energy storage devices, represents the weight coefficient of the th device, represents the state of charge of the th energy storage device at time , and represents the temperature influence correction factor of the device .

[0051] The sorted real-time parameter data set is obtained by the following formula: (11) In formula (11), represents the sorted real-time parameter data set, represents a sample, represents an original data set, represents the mean value of historical operation data, represents the standard deviation multiple threshold value of abnormal value detection, represents the standard deviation of historical data, represents the prediction value based on photovoltaic output and load demand, represents the prediction deviation tolerance threshold.

[0052] Based on future forecasts of photovoltaic power output and load demand, real-time parameters of state of charge (SOC) and charge / discharge efficiency are first obtained from energy storage devices. For example, the SOC of a lithium-ion battery energy storage system might show that the current remaining capacity accounts for 80% of the total capacity, while the charge / discharge efficiency is obtained by measuring the input-output energy ratio, such as an efficiency value of 95%. These parameters are combined with historical operating data, such as charge / discharge records from the past year, and records of the impact of ambient temperature, such as logs of battery efficiency degradation under high-temperature conditions.

[0053] Data cleaning tools such as Python's Pandas library are used to filter out outliers. The specific process involves statistical analysis. First, the mean and standard deviation of the parameters are calculated. Then, points exceeding three times the standard deviation are removed. For example, if the state of charge suddenly jumps to 120%, it is considered a sensor error and removed. This results in a cleaned real-time parameter dataset, which may contain hundreds of timestamped data points for subsequent analysis, ensuring data reliability and supporting the optimized management of energy storage devices.

[0054] Step S220: Based on the organized real-time parameter dataset, obtain the charge-discharge cycle count and capacity decay characteristics of the energy storage device. If a parameter in the charge-discharge cycle count and capacity decay characteristics exceeds a preset threshold, mark the parameter that exceeds the preset threshold and determine the abnormal state parameter set.

[0055] The following formula is used to mark abnormal parameters that exceed a preset threshold: (12) In formula (12), Indicates the first Anomaly flag status for each parameter. Indicates the first One real-time monitoring parameter value, Indicates the first Each parameter corresponds to a preset threshold. When the parameter value exceeds the threshold, it is marked as 1 to indicate an abnormality; otherwise, it is marked as 0 to indicate normality. This is used to construct a set of abnormal state parameters.

[0056] The following formula is used to calculate the total number of charge-discharge cycles for energy storage devices: (13) In formula (13), Indicates the cumulative number of charge-discharge cycles. This indicates the total number of cycles within the monitoring period. Indicates the first The counter for each cycle is used when the energy storage device completes a full charge-discharge cycle. It equals 1, otherwise it is 0.

[0057] The capacity degradation feature is derived by the following equation: (14) In equation (14), represents the capacity retention rate after the th cycle, represents the actual capacity after the th cycle, represents the initial rated capacity, and the capacity retention rate is used to quantify the degree of capacity degradation of the energy storage device. When the capacity retention rate is lower than a pre-set threshold, it is determined as an abnormal state.

[0058] Based on the sorted real-time parameter dataset, the number of charge and discharge cycles and the capacity degradation feature of the energy storage device are obtained. The cycle number can be derived by counting each complete charge and discharge process, such as a battery has completed 500 cycles, while the capacity degradation feature is calculated by comparing the ratio of the current capacity and the initial capacity, for example, the decay rate is 10%. If a certain parameter exceeds the pre-set threshold, such as the cycle number exceeds 1000 or the decay rate is higher than 15%, it is marked to determine the abnormal state parameter set. This marking helps to identify potential faults in a timely manner.

[0059] Step S230, for the abnormal state parameter set, obtain the historical data related to the health of the energy storage device, use time series processing tools to extract features from the historical data related to the health of the energy storage device, and determine the dynamic state indicator range.

[0060] The following formula is used to determine the change range of the dynamic state indicator by calculating the standard deviation: (15) In equation (15), represents the range value of the th dynamic state indicator, represents the number of historical data samples for calculation, represents the value of the th health indicator in the th sample, represents the mean value of the th health indicator.

[0061] For the abnormal state parameter set, obtain the historical data related to the health of the energy storage device, such as the battery voltage curve and the internal resistance change record. Use time series processing tools such as ARIMA model to extract features from these historical data, the process includes decomposing the data into trend, seasonality and residual part, so as to determine the dynamic state indicator range, such as the voltage fluctuation range is between 3.2 volts to 4.2 volts, which helps to understand the evolution of the health of the device.

[0062] Step S240, through the dynamic state index range, combining the capacity attenuation characteristics of the energy storage device, using data fusion tools for calibration, the final available capacity boundary result is obtained.

[0063] The final available capacity boundary result is obtained by the following formula: (16) In formula (16), represents the calibrated capacity boundary value, represents the theoretical maximum capacity boundary, represents the cycle attenuation influence coefficient, represents the actual cycle number, represents the rated cycle number, represents the attenuation index, represents the temperature correction coefficient.

[0064] Through the dynamic state index range combining the capacity attenuation characteristics of the energy storage device, using data fusion tools such as Kalman filter for calibration, the fusion process involves weighted average historical data and real-time characteristics, gradually iteratively adjusting parameters, and finally obtaining the available capacity boundary result, for example, the boundary value is 70% to 90% of the initial capacity, such result can guide energy scheduling, avoid over-discharge risk and improve system efficiency.

[0065] Further, the method for predicting and improving energy utilization rate by using light storage resources provided by the embodiment includes the following steps: Step S310, if the energy storage available capacity boundary exceeds the preset threshold, the distribution details of the peak valley period characteristics are obtained from the future prediction results of the photovoltaic output and load demand, and the distribution details of the peak valley period characteristics are classified and processed by using data arrangement tools, to obtain a divided peak valley period interval set.

[0066] The process of classifying and processing the peak valley period characteristics to obtain the interval set is described by the following formula: (17) In formula (17), represents the divided peak valley period interval set, represents the peak period interval, represents the valley period interval, represents the distribution details of the peak valley period characteristics, represents the classification processing function of the data arrangement tool.

[0067] If the available capacity boundary of the energy storage exceeds the preset threshold, for example, the available capacity boundary of a lithium battery energy storage system is evaluated as 65% of the initial capacity, and the preset threshold is 70%, which indicates that the device may face over-discharge risk, and the distribution details of the peak and valley period characteristics need to be obtained from the future prediction results of photovoltaic output and load demand. Specifically, photovoltaic output prediction may show that from 9 am to 3 pm in the morning, the output is at a peak, and load demand prediction indicates that from 7 pm to 10 pm, demand is at a peak. The distribution details of these characteristics include the duration of the period, the peak intensity, and the fluctuation rate, which are extracted by analyzing historical weather data and power consumption patterns, for example, using a time series decomposition method to split the predicted data into trend and periodic parts, thereby capturing the distribution rules of peaks and valleys, such as the output peak of 500 kW during the peak period, and the valley below 100 kW. This acquisition process helps identify the imbalance point of energy supply and demand, providing basic data support for subsequent classification.

[0068] The distribution details of the peak and valley period characteristics are classified using a data processing tool to obtain a set of divided peak and valley period intervals. In this process, the data processing tool can be a clustering algorithm based on statistical software, which classifies the distribution details according to peak value thresholds and duration standards, such as classifying periods with output exceeding 300 kW as high peak intervals and periods with output below 150 kW as valley intervals. The boundaries are refined by iteratively calculating the mean and variance of each period, such as dividing a 24-hour period into 4 high peak intervals and 3 valley intervals to form a set list, which ensures logical grouping of periods and facilitates further priority evaluation.

[0069] Step S320, for the set of peak and valley period intervals, combined with the grid peak and valley electricity price and carbon emission constraints, a rule matching tool is used to sort the period priorities to determine the priority list of periods suitable for charging and discharging.

[0070] The following formula is used to determine the priority of charging periods, and the lower the electricity price and the less the carbon emission, the higher the priority of the period: (18) In formula (18), represents the charging priority value of the th period, represents the electricity price level of the th valley period, represents the carbon emission factor of the th valley period, represents the grid stability weight of the th period.

[0071] The following formula is used to evaluate the priority of discharging periods, and the higher the peak electricity price and the better the carbon emission reduction effect, the higher the priority of discharging periods: (19) In formula (19), represents the discharge priority score of the th time interval, represents the electricity price benefit of the th peak time interval, represents the carbon emission contribution factor of the th time interval, represents the discharge efficiency coefficient of the th time interval.

[0072] For the set of peak-valley time interval, combined with the grid peak-valley electricity price and carbon emission constraints, a rule matching tool is used to sort the time interval priority, and determine the time interval priority list suitable for charging and discharging. Specifically, the rule matching tool involves setting a multi-criteria decision framework, for example, taking the peak-valley electricity price as a weight factor, discharging preferentially in high peak electricity price time interval to gain profit, while the carbon emission constraint limits the charging activity in high carbon time interval, by matching each interval one by one, for example, if the valley segment interval price is 0.3 yuan / kWh and the carbon emission factor is low, it is given a charging priority of 1 and a discharging priority of 3. Finally, a sorted list is generated, arranging the charging and discharging suitable time intervals from high to low, which helps to optimize the sustainability of energy utilization.

[0073] Step S330, if the charging time interval in the time interval priority list matches the photovoltaic peak, generate photovoltaic peak charging instruction, and if the discharging time interval matches the load peak, generate load peak discharging instruction, to obtain the preliminary charge-discharge instruction combination.

[0074] The following formula is used to define the photovoltaic peak charging instruction generation condition: (20) In formula (20), represents the photovoltaic peak charging instruction generation flag, represents the set of charging time intervals in the time interval priority list, represents the set of photovoltaic peak time intervals, represents an empty set, and the charging instruction flag is 1 when the charging time interval intersects with the photovoltaic peak time interval, otherwise 0.

[0075] The following formula is used to define the load peak discharging instruction generation condition: (21) In formula (21), represents the load peak discharging instruction generation flag, represents the set of discharging time intervals in the time interval priority list, This represents the set of peak load periods. When the discharge period overlaps with the peak load period, the discharge command flag is set to 1; otherwise, it is set to 0.

[0076] The initial combination of charge and discharge commands is derived using the following formula: (twenty two) In formula (22), Indicates time The initial combination of charge and discharge commands, Indicates time The charging power, Indicates time The discharge power, Indicates time The peak matching coefficient of photovoltaic power. Indicates time The load peak matching coefficient, and These are the weighting coefficients for charging and discharging, respectively.

[0077] If the charging period in the time priority list matches the photovoltaic peak, a photovoltaic peak charging instruction is generated. Simultaneously, if the discharging period matches the load peak, a load peak discharging instruction is generated, resulting in a preliminary combination of charging and discharging instructions. In practice, this means checking whether the high-priority charging periods in the list fall within the photovoltaic output peak. For example, if the charging period at 10 AM coincides with the photovoltaic peak, the instruction specifies charging the battery at maximum power during this period. Conversely, the discharging instruction targets the load peak at 8 PM, discharging to meet demand. These instructions are combined into a sequence, such as "Charging: 9-12 AM, Discharging: 6-9 PM," thus initially planning energy storage operations.

[0078] Step S340: By combining the initial charging and discharging command combinations with the energy storage device status and energy storage capacity boundary, the commands are adjusted using a data calibration tool to determine the optimal charging and discharging strategy sequence.

[0079] The optimal charge / discharge strategy sequence is determined by minimizing the total cost using the following formula: (twenty three) In formula (23), Indicates time The optimal charge and discharge power, Represents the decision variable for charging and discharging power. Indicates the total number of time periods. Indicates time The charging cost coefficient, Indicates time The charging power, Indicates time The discharge benefit coefficient, Indicates time The discharge power.

[0080] By combining initial charge and discharge command combinations with the energy storage device's status and capacity boundaries, a data calibration tool is used to adjust the commands and determine the optimal charge and discharge strategy sequence. The data calibration tool can use optimization algorithms such as linear programming to iteratively adjust the commands. For example, given an input device current state of charge of 50% and a capacity boundary of 70%-90%, the calibration process involves simulating the energy flow of each command and checking for exceeding the boundaries. If a charging command causes the charge to exceed 90%, the adjustment is to reduce the charging rate. The final output strategy sequence is such as "Optimized charging: 9-11 hours (80% power), discharging: 18-20 hours (100% power)". This adjustment ensures the safety and efficiency of the strategy.

[0081] Furthermore, the method for improving energy utilization efficiency by utilizing photovoltaic and energy storage resources provided in this embodiment includes step S400: Step S410: After transmitting the optimal charging and discharging strategy sequence to the edge controller through the remote management platform, the optimal charging and discharging strategy sequence is parsed and processed by the data distribution tool to obtain the decomposed charging and discharging instruction set.

[0082] The calculation method for a single charging instruction in the charging instruction set is defined by the following formula: (twenty four) In formula (24), Indicates time The charging command, This indicates the total number of charging periods. Indicates the first Charging power coefficient for each time period This represents a function indicating the charging status. This indicates the maximum charging power.

[0083] The calculation method for a single discharge command in the discharge command set is defined by the following formula: (25) In formula (25), Indicates time The discharge command is output. Indicates the total number of discharge periods. Indicates the first Discharge power coefficient for each time period This represents the discharge state indication function. This indicates the available discharge energy.

[0084] After the optimal charging and discharging strategy sequence is transmitted to the edge controller through the remote management platform, the data distribution tool will first analyze and process the optimal charging and discharging strategy sequence. The optimal charging and discharging strategy sequence is a list of instructions containing timestamps and power levels, such as charging 80% power from 9 am to 11 am and discharging 100% power from 6 pm to 8 pm. The data distribution tool reads the structured format of the sequence, such as JSON data packets, and decomposes it into independent charging and discharging instruction sets. For example, the charging set includes the start and end time of the specific period and the target voltage, while the discharging set lists the discharging rate and duration. This parsing process involves scanning the sequence content line by line to ensure the integrity of each instruction and generates a set list for subsequent device linkage, which helps to achieve accurate execution of the strategy.

[0085] Step S420, according to the decomposed charging and discharging instruction set, link the photovoltaic inverter and energy storage converter, use real-time instruction distribution tool to distribute corresponding instructions to corresponding devices, and determine the execution state of the photovoltaic and energy storage devices and the controllable devices on the load side.

[0086] The execution state of the photovoltaic and energy storage devices and the controllable devices on the load side is obtained by the following formula: (26) In formula (26), represents the execution state of the device at time , represents the power output of the photovoltaic inverter at time , represents the power state of the energy storage converter at time , represents the power demand of the controllable device on the load side at time , represents the weight coefficient of the photovoltaic device, represents the weight coefficient of the energy storage device, represents the weight coefficient of the load device.

[0087] ​​​​In an actual photovoltaic energy storage system, the parsed charging instruction set can specify that the photovoltaic inverter directs excess power to the energy storage converter during peak hours, and the discharging instruction is aimed at the load peak period. According to these sets, the system will link the photovoltaic inverter and the energy storage converter, and use a real-time instruction issuing tool to distribute instructions. The real-time instruction issuing tool is a communication module based on the MQTT protocol, which sends charging instructions to the photovoltaic inverter, such as "start DC-AC conversion and charge at 50 kilowatts of power", and at the same time pushes discharging instructions to the energy storage converter, such as "discharge to the load side at 60 kilowatts of power". In this process, the system monitors the response time and state feedback of the equipment, determines the charging and discharging state of the photovoltaic energy storage equipment such as the battery pack, and the execution of the controllable equipment on the load side such as the intelligent air conditioner. If the inverter successfully adjusts the output and the converter confirms the discharging start, it is considered that the execution state is normal. This linkage ensures the real-time coordination of energy flow.

[0088] Step S430, if the execution state of the photovoltaic energy storage equipment and the controllable equipment on the load side meets the preset threshold, the running data after the cooperative response is obtained from the equipment running log through the data acquisition tool.

[0089] The effect of cooperative response is quantified by the following formula: (27) In formula (27), represents the comprehensive running effect evaluation index after cooperative response, represents the number of monitored running parameter types, represents the weight coefficient of the th parameter, represents the running data of the th parameter after cooperative response, represents the baseline data of the th parameter before cooperative response, represents the standard baseline value of the th parameter.

[0090] If the execution status of the light storage device and the controllable device on the load side meets the preset threshold value, such as the battery temperature being below 45 degrees Celsius and the load response rate exceeding 95%, the operation data after the coordinated response is obtained from the device operation log through a data acquisition tool. The data acquisition tool can be an embedded log reader that extracts data such as actual charging capacity and voltage curve from the log file of the photovoltaic inverter, obtains discharge efficiency and remaining capacity information from the log of the energy storage converter, and collects records such as power consumption and response delay from the log of the device on the load side. These data are summarized into a time series data set for subsequent analysis of the overall performance of the system, for example, if the log shows that the energy loss during charging is less than 5%, it indicates that the coordinated response is effective, which provides a data basis for optimizing future strategies. Assuming that in a photovoltaic energy storage application of a commercial building, after the execution status check confirms that all devices are running smoothly, the data acquisition tool automatically pulls the log and generates a report showing that the total energy is saved by 20% after the coordinated response, which reflects the actual benefits of strategy execution.

[0091] Further, the method for predicting the improvement of energy utilization rate by utilizing light storage resources provided by the embodiment includes the following steps: Step S510, according to the operation data after the coordinated response, the corresponding consumption rate index value is calculated by using the data processing tool, if the consumption rate index value is lower than the preset threshold value, the adjustment signal is generated by using the feedback tool, and the direction of feedback adjustment is determined.

[0092] The following formula is used to calculate the consumption efficiency of renewable energy by the system after the coordinated response: (28) In formula (28), represents the consumption rate index value, represents the actual consumed energy amount, represents the total available energy amount.

[0093] The following formula is used to define the generation condition of the adjustment signal: (29) In formula (29), represents the generation state of the adjustment signal, represents the preset consumption rate threshold value, and when the consumption rate is lower than the threshold value, the generation adjustment signal is 1, otherwise it is 0.

[0094] The direction of feedback adjustment is obtained by the following formula: (30) In formula (30), represents the direction and intensity of feedback adjustment, represents the target consumption rate threshold value, represents the current accommodation rate value, represents the adjustment gain coefficient, is a sign function used to determine the adjustment direction.

[0095] According to the operating data after the coordinated response, the data processing tool is used to calculate the corresponding accommodation rate index value. Here, the accommodation rate index value refers to the proportion of renewable energy that is effectively utilized in the photovoltaic energy storage system. For example, the system efficiency can be evaluated by calculating the ratio of the actual absorbed solar power to the total power generation. The data processing tool is an algorithm-based analysis module that first extracts key parameters from the operating data, such as the output power of the photovoltaic inverter and the storage capacity of the energy storage converter. Then, it uses formulas to convert these parameters into the accommodation rate, for example, by dividing the absorbed power by the total power generation to obtain a percentage value. This process involves a data cleaning step to ensure the accuracy of the input data and avoid noise interference with the calculation results, thereby providing a reliable basis for subsequent decision-making. For example, in a photovoltaic energy storage project in an industrial park, the operating data shows that the photovoltaic system generates 1000 kilowatt-hours of electricity in a day, but only 800 kilowatt-hours are actually accommodated to the load side. The data processing tool automatically calculates the accommodation rate as 80%. If this value is lower than the preset threshold, such as 85%, the next mechanism is triggered, which helps to identify energy waste problems in a timely manner.

[0096] Step S520: Transmit the adjustment signal to the prediction parameter database in the direction of feedback adjustment, and use the parameter updating tool to correct the adjustment signal to obtain the optimized prediction accuracy.

[0097] The following formula describes the process of generating an adjustment signal by feedback adjustment direction and transmitting it to the prediction parameter database: (31) In formula (31), represents the adjustment signal vector, represents the learning rate coefficient, represents the gradient of the loss function with respect to the prediction parameters, represents the feedback weight coefficient, represents the feedback signal vector.

[0098] The following formula quantifies the improvement in system prediction accuracy after parameter optimization: (32) In formula (32), represents the optimized prediction accuracy, represents the total number of prediction samples, represents the weight of the th sample, represents the true value of the th sample, Indicates the first The predicted value for each sample.

[0099] If the absorption rate falls below a preset threshold, an adjustment signal is generated via a feedback tool to determine the direction of adjustment. This feedback tool is a response mechanism that analyzes the reasons for the low absorption rate, such as load mismatch or equipment failure, and generates a signal indicating whether to increase charging periods or optimize discharge strategies. The feedback tool compares the current indicator with the threshold and generates a vector signal; for example, a positive adjustment indicates increased absorption, while a negative adjustment indicates reduced losses. This ensures the system evolves towards a more efficient state. For instance, assuming the absorption rate is only 70%, the feedback tool might generate a signal pointing to "extending peak-hour discharge." This direction is based on historical data analysis and aims to match peak loads, thereby improving overall energy utilization.

[0100] The adjustment signal is transmitted to the prediction parameter database via feedback adjustment direction. A parameter update tool then corrects the adjustment signal to obtain optimized prediction accuracy. The prediction parameter database stores historical data such as weather forecasts and load patterns. The parameter update tool is an iterative algorithm module that integrates the adjustment signal into the database, for example, by correcting the prediction model parameters using a weighted averaging method to ensure that the next prediction is more accurate. In residential energy storage systems, after the adjustment signal is transmitted, the parameter update tool corrects the prediction model's error rate, reducing it from 10% to 5%. The optimized accuracy directly improves the reliability of the strategy.

[0101] Step S530: Based on the optimized prediction accuracy, use a data integration tool to combine historical operating data to generate a prediction basis applicable to subsequent cycles, and determine whether the prediction basis meets the load demand.

[0102] The following formula is used to define the criteria for determining whether the forecast data meets the load demand: (33) In formula (33), This indicates the result of the judgment on whether the forecast is based on whether it meets the load demand. Indicates time The actual load demand value, Indicates time The predicted value, This indicates the allowable relative error threshold. When the relative error between the predicted value and the actual requirement is within the allowable range, the result is 1, indicating compliance; otherwise, it is 0, indicating non-compliance.

[0103] Based on the optimized prediction accuracy, a data integration tool is used to combine historical operational data to generate a prediction basis for the subsequent period. The data integration tool fuses multiple sources of data, such as the consumption rate and weather variables from the past week, to generate a comprehensive basis for estimating future energy demand. The integrated basis shows that the next period's load will increase by 20%, providing data support for planning.

[0104] Determining whether the prediction basis meets the load demand involves comparing the basis with the actual demand curve. If the match is high, it is confirmed to be effective, otherwise further iteration is needed to ensure that the system responds to load fluctuations. For example, in a commercial building application, if the basis estimates the load to be 500 kWh and the actual demand is 480 kWh, it is determined to be in line, reflecting the practical value of the prediction. This closed-loop mechanism ultimately improves the sustainability and economic benefits of the energy storage system.

[0105] See Figure 2The application provides a system for predicting and improving energy utilization rate by using light storage resources, which is used for executing the method for predicting and improving energy utilization rate by using light storage resources, and comprises a photovoltaic output and load demand future prediction result acquisition module 10, a storage energy available capacity boundary determination module 20, an optimal charging and discharging strategy sequence generation module 30, a cooperative response operation data acquisition module 40 and a prediction accuracy acquisition module 50. The photovoltaic output and load demand future prediction result acquisition module 10 is used for collecting irradiance temperature history output data and load consumption characteristic data from photovoltaic equipment and meteorological sources through distributed sensors and edge gateways, processing time sequence characteristics by using a sequence prediction model, integrating multi-dimensional input data by using an ensemble learning model, and obtaining photovoltaic output and load demand future prediction results. The storage energy available capacity boundary determination module 20 is used for obtaining storage energy equipment state of charge and charging and discharging efficiency real-time parameters according to the photovoltaic output and load demand future prediction results, constructing a storage energy full life cycle state evaluation model, calculating dynamic state indexes, and determining a storage energy available capacity boundary. The optimal charging and discharging strategy sequence generation module 30 is used for extracting peak-valley period characteristics from the photovoltaic output and load demand future prediction results if the storage energy available capacity boundary exceeds a preset threshold, combining power grid peak-valley electricity prices and carbon emission constraints, and generating an optimal charging and discharging strategy sequence by using an intelligent optimization algorithm, wherein the optimal charging and discharging strategy sequence comprises photovoltaic peak charging instructions and load peak discharging instructions. The cooperative response operation data acquisition module 40 is used for transmitting the optimal charging and discharging strategy sequence to an edge controller through a remote management platform, linking photovoltaic inverters and storage energy converters to execute real-time instructions, realizing cooperative response of photovoltaic storage equipment and load side controllable equipment, and acquiring operation data of the photovoltaic storage equipment and the load side controllable equipment after cooperative response. The prediction accuracy acquisition module 50 is used for extracting an inclusion rate index from the operation data after cooperative response, updating multi-dimensional fusion prediction model parameters by using a feedback mechanism, obtaining an optimized prediction accuracy, and being used for photovoltaic output and load demand prediction in a subsequent period.

[0106] Compared with the prior art, the method and system for predicting and improving energy utilization rate by using light storage resources provided by the embodiment collect irradiance temperature historical output data and load consumption characteristic data through distributed sensors and edge gateways, process time sequence characteristics by using a sequence prediction model, integrate multi-dimensional inputs by using an ensemble learning model, obtain photovoltaic output and load demand future prediction results, further obtain real-time parameters of energy storage equipment state of charge and charge-discharge efficiency, construct an energy storage full life cycle state evaluation model to calculate dynamic indexes and determine available capacity boundaries, extract peak-valley period characteristics if the boundaries exceed a preset threshold, combine power grid peak-valley electricity prices and carbon emission constraints, and generate an optimal strategy sequence including photovoltaic peak charging and load peak discharging by using an intelligent optimization algorithm. The optimal strategy sequence is transmitted to an edge controller through a remote management platform, and photovoltaic inverters and energy storage converters are linked to execute real-time instructions, so that photovoltaic storage equipment and load side controllable equipment are cooperatively responded. Finally, a consumption rate index is extracted from the operating data after the response, a prediction model parameter is updated by using a feedback mechanism, and the prediction accuracy after optimization is improved, which is used for subsequent period prediction. The embodiment significantly improves photovoltaic consumption rate, reduces operation cost, meets carbon emission requirements, and realizes intelligent and efficient management of the photovoltaic storage system.

[0107] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments by those skilled in the art once they learn of the basic inventive concepts. Therefore, the appended claims are intended to encompass within their scope all possible variations and modifications of the preferred embodiments. It is apparent that those skilled in the art can, without departing from the spirit or scope of the application, make various changes and modifications of the application. Thus, the application is intended to encompass all such changes and modifications as fall within the scope of the claims, together with all equivalents thereof.

Claims

1. A method for predicting the improvement of energy utilization rate by using light storage resources, characterized in that, The method comprises the following steps: S100, collecting irradiance temperature history output data and load consumption characteristic data from photovoltaic equipment and meteorological sources through distributed sensors and edge gateways, processing time series characteristics using a sequence prediction model, integrating multi-dimensional input data using an ensemble learning model, and obtaining photovoltaic output and load demand future prediction results; S200, obtaining real-time parameters of state of charge and charge-discharge efficiency of energy storage equipment according to the photovoltaic output and load demand future prediction results, constructing an energy storage full life cycle state evaluation model, calculating dynamic state indicators, and determining the energy storage available capacity boundary; S300, if the energy storage available capacity boundary exceeds a preset threshold, extracting peak-valley period features from the photovoltaic output and load demand future prediction results, combining grid peak-valley electricity prices and carbon emission constraints, and using an intelligent optimization algorithm to generate an optimal charge-discharge strategy sequence, wherein the optimal charge-discharge strategy sequence includes photovoltaic peak charging instructions and load peak discharging instructions; S400, transmitting the optimal charge-discharge strategy sequence to an edge controller through a remote management platform, and linking photovoltaic inverters and energy storage converters to execute real-time instructions, realize coordinated response of photovoltaic storage equipment and controllable equipment on the load side, and obtain operation data after the coordinated response of the photovoltaic storage equipment and the controllable equipment on the load side; S500, extracting the absorption rate index from the operation data after the coordinated response, updating the multi-dimensional fusion prediction model parameters using a feedback mechanism, obtaining the optimized prediction accuracy, and using it for photovoltaic output and load demand prediction in subsequent periods.

2. The method of claim 1, wherein, Step S100 comprises: S110, obtaining irradiance data, temperature data, and historical output data from photovoltaic equipment and meteorological sources through distributed sensors and edge gateways, and collecting load consumption characteristic data from the load end, extracting time series characteristics from the irradiance data, the temperature data, the historical output data, and the load consumption characteristic data, filtering outliers using a preset threshold, and obtaining a cleaned time series data set; S120, according to the cleaned time series data set, using a sequence prediction tool to predict future trends of photovoltaic output data and load demand data, smoothing the volatility characteristics in the prediction results, and determining preliminary prediction data results; S130, fusing the preliminary prediction data results with multi-dimensional input data through an ensemble learning tool, obtaining a comprehensive data set containing irradiance, temperature, and load consumption characteristics, if the deviation of a certain dimension data in the comprehensive data set exceeds a preset threshold, performing secondary correction on the dimension data exceeding the preset threshold, and determining an adjusted comprehensive prediction data set; S140, according to the adjusted comprehensive prediction data set, using a data visualization tool to generate a prediction curve of future photovoltaic output and load demand, labeling key nodes of the prediction curve, and obtaining photovoltaic output and load demand future prediction results.

3. The method of claim 1, wherein, Step S200 comprises: S210, according to the photovoltaic output and load demand future prediction results, by obtaining the energy storage device state of charge and charging and discharging efficiency real-time parameters from the energy storage device, combining historical operation data and environmental temperature influence records, using a data cleaning tool to filter outliers, obtaining the sorted real-time parameter data set; S220, according to the sorted real-time parameter data set, obtaining the charging and discharging cycle number and capacity attenuation characteristics of the energy storage device, if a parameter in the charging and discharging cycle number and capacity attenuation characteristics exceeds a preset threshold, marking the parameter exceeding the preset threshold to determine the abnormal state parameter set; S230, for the abnormal state parameter set, obtaining the energy storage device health-related historical data, using a time series processing tool to extract features from the energy storage device health-related historical data, and determining the dynamic state index range; S240, by the dynamic state index range, combining the capacity attenuation characteristics of the energy storage device, using a data fusion tool to calibrate, obtaining the final available capacity boundary result.

4. The method for predicting the promotion of energy utilization by utilizing light storage resources according to claim 1, characterized in that, Step S300 includes: S310, if the energy storage available capacity boundary exceeds a preset threshold, obtain the distribution details of the peak and valley period characteristics from the photovoltaic output and load demand future prediction results, use a data arrangement tool to classify and process the distribution details of the peak and valley period characteristics, and obtain the divided peak and valley period interval set; S320, for the peak and valley period interval set, combining the grid peak and valley electricity price and carbon emission constraint, using a rule matching tool to perform period priority sorting, and determining a period priority list suitable for charging and discharging; S330, if the charging period in the period priority list matches the photovoltaic peak, generate a photovoltaic peak charging instruction, and if the discharging period matches the load peak, generate a load peak discharging instruction, and obtain a preliminary charging and discharging instruction combination; S340, through the preliminary charging and discharging instruction combination, combining the energy storage device state and the energy storage capacity boundary, using a data calibration tool to adjust the instructions, and determining the optimal charging and discharging strategy sequence.

5. The method for predicting the promotion of energy utilization by utilizing light storage resources according to claim 1, characterized in that, Step S400 includes: S410, after transmitting the optimal charging and discharging strategy sequence to the edge controller through the remote management platform, using a data distribution tool to analyze and process the optimal charging and discharging strategy sequence, and obtaining a decomposed charging and discharging instruction set; S420, according to the decomposed charging and discharging instruction set, linking the photovoltaic inverter and the energy storage converter, using a real-time instruction issuing tool to distribute corresponding instructions to corresponding devices, and determining the execution state of the light storage device and the load side controllable device; S430, if the execution state of the light storage device and the load side controllable device meets the preset threshold, obtaining the operation data after the coordinated response from the device operation log through the data acquisition tool.

6. The method for predicting the promotion of energy utilization by utilizing light storage resources according to claim 5, characterized in that, In step S410, the calculation method of a single charging instruction in the charging instruction set is defined by the following formula: wherein, denotes a charging instruction at a time instant , denotes a total number of charging periods, denotes a charging power coefficient of the th period, denotes a charging state indication function, denotes a maximum charging power; The calculation method of a single discharging instruction in the discharging instruction set is defined by the following formula: wherein, denotes the time instant the discharge command output, denotes the total number of discharge periods, denotes the discharge power coefficient of the period, denotes the discharge status indication function, denotes the available discharge energy.

7. The method of claim 6, wherein, In step S420, the execution state of the light storage device and the load side controllable device is obtained by the following formula: wherein, denotes the state of the device at time , denotes the power output of the photovoltaic inverter at time , denotes the power state of the energy storage converter at time , denotes the power demand of the controllable load-side device at time , denotes the weight coefficient of the photovoltaic device, denotes the weight coefficient of the energy storage device, denotes the weight coefficient of the load device. 8.The method of claim 7, wherein, In step S430, the following formula is used to quantify the effect of coordinated response: wherein, represents the comprehensive operation effect evaluation index after the cooperative response, represents the number of monitored operation parameter types, represents the weight coefficient of the th parameter, represents the operation data of the th parameter after the cooperative response, represents the baseline data of the th parameter before the cooperative response, represents the standard baseline value of the th parameter. 9.The method of predicting and improving energy utilization rate by utilizing light storage resource according to claim 1, wherein, Step S500 comprises: S510, according to the operation data after the cooperative response, using a data processing tool to calculate the corresponding accommodation rate index value, if the accommodation rate index value is lower than the preset threshold, generating an adjustment signal through a feedback tool to determine the direction of feedback adjustment; S520, transmitting the adjustment signal to the prediction parameter database through the direction of feedback adjustment, using a parameter updating tool to correct the adjustment signal to obtain the optimized prediction accuracy; S530, according to the optimized prediction accuracy, using a data integration tool to combine historical operation data to generate prediction basis suitable for subsequent periods, and judging whether the prediction basis meets the load demand.

10. A system for predicting the improvement of energy utilization using light storage resources, for performing the method for predicting the improvement of energy utilization using light storage resources according to any one of claims 1 to 9, characterized in that, Comprise: The photovoltaic output and load demand future prediction result acquisition module (10) is used for collecting irradiance temperature history output data and load consumption characteristic data from photovoltaic equipment and meteorological sources through distributed sensors and edge gateways, processing time sequence characteristics by using a sequence prediction model, integrating multi-dimensional input data by using an ensemble learning model, and obtaining photovoltaic output and load demand future prediction results; The energy storage available capacity boundary determination module (20) is used for obtaining real-time parameters of energy storage equipment state of charge and charge-discharge efficiency according to the photovoltaic output and load demand future prediction results, constructing an energy storage full life cycle state evaluation model, calculating dynamic state indexes, and determining the energy storage available capacity boundary; The optimal charge-discharge strategy sequence generation module (30) is used for extracting peak-valley period characteristics from the photovoltaic output and load demand future prediction results if the energy storage available capacity boundary exceeds a preset threshold, combining power grid peak-valley electricity price and carbon emission constraints, and generating an optimal charge-discharge strategy sequence by using an intelligent optimization algorithm, wherein the optimal charge-discharge strategy sequence includes photovoltaic peak charging instructions and load peak discharging instructions; The cooperative response operation data acquisition module (40) is used for transmitting the optimal charge-discharge strategy sequence to an edge controller through a remote management platform, linking photovoltaic inverters and energy storage converters to execute real-time instructions, realizing cooperative response of photovoltaic storage equipment and load side controllable equipment, and acquiring operation data of the photovoltaic storage equipment and load side controllable equipment after the cooperative response; The prediction accuracy acquisition module (50) is used for extracting an accommodation rate index from the cooperative response operation data, updating multi-dimensional fusion prediction model parameters by using a feedback mechanism, obtaining optimized prediction accuracy, and predicting photovoltaic output and load demand in subsequent periods.

Images