Building energy supply method based on solar photovoltaic energy storage

Abstract

The invention relates to the technical field of building energy supply, and aims to solve the problem of how to efficiently utilize solar photovoltaic stored energy to supply energy to a building. According to the technical scheme, a monocrystalline silicon photovoltaic module array is laid on the roof of a building, and a lithium battery pack and an alternating-current distribution box are connected through a bidirectional energy storage converter to supply power to electric equipment of the building. The meteorological data acquisition module is used for measuring an irradiance value and a temperature value, and the load monitoring module is used for acquiring electric power. In the building energy management platform, a photovoltaic power generation prediction model generates a photovoltaic output prediction curve in combination with real-time and prediction data, a building load prediction model generates a load power demand curve according to historical and regular data, and an energy storage control decision-making model performs energy storage control according to a difference matrix between the two. A lithium battery pack charging and discharging power distribution scheme is obtained through a dynamic programming algorithm, and the bidirectional energy storage converter is controlled. The system is mainly used for realizing efficient and stable solar photovoltaic energy storage and supply of buildings.

Description

Technical Field

[0001] The present invention relates to the technical field of building energy supply. More specifically, the present invention relates to a building energy supply method based on solar photovoltaic energy storage. Background Art

[0002] With the intensification of the global energy crisis and the continuous improvement of people's awareness of environmental protection, the continuous growth of building energy consumption and the popularization of the application of renewable energy, the application of solar photovoltaic energy storage systems in the field of building energy supply has gradually become an important development direction.

[0003] Solar energy, as a clean and renewable energy source, has great development potential. Applying solar photovoltaic technology in the building field for power generation is one of the important ways to achieve building energy conservation and emission reduction. However, solar photovoltaic power generation has the characteristics of intermittency and instability, and its power generation power will be affected by factors such as weather, season and time. In the prior art, building photovoltaic energy supply systems usually combine fixed photovoltaic modules with conventional energy storage devices and achieve energy scheduling through simple charge and discharge strategies. The existing building photovoltaic energy supply systems have the following significant defects: First, the power generation power of photovoltaic power generation is significantly affected by weather fluctuations. The lack of a high-precision prediction model leads to a low matching degree between power generation and load demand, and there are often situations where excess photovoltaic electric energy cannot be effectively stored or power supply is insufficient and needs to rely on the power grid; Second, the control strategies of energy storage systems are mostly based on static thresholds, and multi-dimensional factors such as time-of-use electricity prices and battery health status are not considered, resulting in poor economy and shortened battery life; Third, traditional meteorological monitoring devices and load acquisition modules are deployed separately, and the data fusion degree is insufficient, making it difficult to support dynamic optimization scheduling. In addition, in the prior art, the coordinated control between the photovoltaic module array, the energy storage device and the power distribution system lacks a unified management platform, resulting in system response lag and low energy efficiency.

[0004] In summary, there is an urgent need for a photovoltaic energy storage energy supply method that integrates precise prediction, dynamic optimization and safety control to improve the building energy self-sufficiency rate and economy. Summary of the Invention

[0005] An object of the present invention is to solve at least the above problems and provide at least the advantages described hereinafter.

[0006] Another object of the present invention is to provide a building energy supply method based on solar photovoltaic energy storage, which can accurately predict photovoltaic output and building load, reasonably plan the charge and discharge of lithium battery packs, improve energy utilization efficiency, and ensure the stability of building energy supply.

[0007] To achieve these objects and other advantages according to the present invention, there is provided a building energy supply method based on solar photovoltaic energy storage, including: Lay a monocrystalline silicon photovoltaic module array on the surface of the building roof. The photovoltaic module array is respectively connected to a lithium battery pack and the internal AC distribution box of the building through a bidirectional energy storage inverter. The output end of the AC distribution box is connected to the building electrical equipment; Set up a meteorological data acquisition module on the building roof. The meteorological data acquisition module includes an irradiance sensor and a temperature sensor that respectively measure the irradiance value on the surface of the photovoltaic module and the temperature value on the back plate of the photovoltaic module; Install a load monitoring module on each power supply branch of the building AC distribution box. The load monitoring module collects the real-time power consumption of the air conditioning system, lighting equipment, and office equipment through a current crossbar; Set up a server of the building energy management platform in the building. The building energy management platform is configured with a photovoltaic power generation prediction model, a building load prediction model, and an energy storage control decision model. The photovoltaic power generation prediction model receives the real-time measurement data of the meteorological data acquisition module, and combines the predicted irradiance value and ambient temperature predicted value for the next 24 hours provided by the weather forecast to establish a photovoltaic output prediction curve based on time series; The building load prediction model generates a building load power demand curve that matches the time resolution of the photovoltaic power generation prediction model by analyzing the historical data of the operating power of the air conditioning system, the opening time pattern data of the lighting equipment, and the power consumption period distribution data of the office equipment provided by the load monitoring module, and combining the date type input by the built-in working day / holiday mode selector; Among them, the energy storage control decision model generates a time-domain difference matrix according to the calculated photovoltaic output prediction curve and the building load power demand curve, and according to the charge and discharge period distribution of the time-domain difference matrix, uses the dynamic programming algorithm to calculate the charge and discharge power distribution plan of the lithium battery pack within the next 24 hours, and converts it into a charge and discharge instruction set in time series to control the bidirectional energy storage inverter.

[0008] Preferably, when the deviation between the actual irradiance value and the predicted value monitored by the meteorological data acquisition module in real time exceeds ±15%, trigger the real-time correction mode of the energy storage control decision model. The correction mode dynamically adjusts the charge and discharge power of the lithium battery pack according to the difference between the current actual photovoltaic output and the real-time demand of the building load, and at the same time feeds back the deviation data to the photovoltaic power generation prediction model for rolling correction.

[0009] Preferably, the specific implementation steps for the photovoltaic power generation prediction model to establish a photovoltaic output prediction curve based on time series are as follows: The meteorological data acquisition module uploads the irradiance measurement value and the temperature value of the back plate of the photovoltaic module in real time, and the meteorological bureau API interface provides the predicted irradiance data and ambient temperature predicted data for the next 24 hours in real time; The sliding window algorithm is used to fuse the real-time uploaded irradiance data with the irradiance prediction data provided by the meteorological bureau to generate an irradiance prediction curve: The irradiance prediction data with a 1h resolution provided by the meteorological bureau API is converted to a 15min time granularity through cubic spline interpolation. Within each 15min time period, an irradiance prediction curve G(t) is generated based on the irradiance prediction value provided by the meteorological bureau; the deviation rate δ between the average value G1 of the real-time uploaded irradiance data in the previous 10min and the irradiance prediction data G2 in the corresponding period is calculated to adjust the irradiance prediction curve G(t): When the absolute value of δ ≤ 15%, the actual irradiance prediction curve G’(t) = G(t); When the absolute value of δ > 15%, the actual irradiance prediction curve G’(t) = G(t) × (1 + 0.7 × δ / 100); Based on the deviation between the real-time uploaded irradiance data and the theoretical irradiance value, the dust deposition coefficient γ on the surface of the photovoltaic module is calculated: ; The efficiency ε of the photovoltaic module is corrected according to the real-time backplane temperature data: ; A photovoltaic output power prediction curve P(t) is generated at a 15mn time step: P(t) = ε × (1 - γ) × G’(t) × A; G L is the theoretical irradiance value, G L = G0 × cos(θ), where G0 is the theoretical solar irradiance that can be received on the surface of the photovoltaic module under ideal clear weather conditions without clouds and atmospheric pollution, W / m 2 ; θ is the solar incidence angle; ε0 is the nominal conversion efficiency of the photovoltaic module under standard test conditions; △T is the difference between the backplane temperature value of the photovoltaic module and the standard test temperature, and the standard test temperature = 25°C; A is the effective sunlight-receiving area of the photovoltaic module, m 2 .

[0010] Preferably, the specific steps for the building load prediction model to generate the building load power demand curve are as follows: Obtain historical data from the load monitoring module and preprocess it: Record the time sequence matrix of the switching time and power level of lighting equipment in each area every day; Statistical probability distribution of the daily usage duration of office equipment; Modeling based on the basic load component: For air-conditioning load: ; For lighting load: ; For office equipment load: ; Correct the building load power according to the date type: P 负荷 (t) = [P 空调 (t) + P 照明 (t) + P 办公 (t)] × σ(D); Generate the building load power demand curve: Discretize P 负荷 (t) at a 15 - minute time granularity to generate a 24 - hour building load power demand curve; Among them, P 空调 (0) is the reference power of the air - conditioning system, in kW, taking the average value under the condition of 25°C in historical data; △ 空调 T(t) is the difference between the real - time outdoor temperature and the set temperature; ρ(t) is the real - time personnel density, in persons / m 2 ; N i is the number of the i - th type of lighting equipment; P i is the rated power of a single i - th type of lighting equipment, in kW; S i (t) is the switch - state function of the i - th type of lighting equipment, 0 for off and 1 for on, which is obtained from the time - series matrix of the daily switch - on / off time and power levels of lighting equipment in each area; B is the ambient light intensity measured in real - time inside the building, in lux; M j is the number of the j - th type of office equipment; P standby is the standby power of the equipment, in kW; P full - load is the full - load power of the equipment, in kW; F(t) is the probability of equipment usage in a time period, which is obtained from the probability distribution of the daily usage duration of office equipment, and takes values from 0 to 1; σ(D) is the date correction coefficient, taking 1.0 on weekdays, 0.75 on weekends, and 0.5 on legal holidays.

[0011] Preferably, the method for obtaining the time - domain difference matrix is as follows: The energy - storage control decision model aligns the photovoltaic power output prediction curve and the load power demand curve along the time axis, calculates the power difference at each time node, and generates the time - domain difference matrix. The positive - value time periods in the time - domain difference matrix are the energy - storage charging periods, and the negative - value time periods are the energy - storage discharging periods.

[0012] Preferably, the objective function of the dynamic programming algorithm is ; The constraint condition is ; Among them, S(t) is the percentage of the remaining power of the lithium - battery pack in the total capacity; C(t) is the time-of-use electricity price of the power grid in period t, Yuan / kWh; H(t) is the electricity purchase quantity from the power grid in period t, kW; ; β is the battery state of charge balancing weight factor, taking 0.05; P ch (t) is the charging power in period t, kW; P di (t) is the discharging power in period t, kW; K(t) is the power of photovoltaic power generation exceeding the load demand in period t, kW; L(t) is the power of the load demand exceeding the photovoltaic power generation in period t, kW; P ch-max is the maximum charging power of the lithium battery pack, kW; P dis-max is the maximum discharging power of the lithium battery pack, kW; E 总 is the total capacity of the lithium battery pack, kWh; △t = 15 min.

[0013] Preferably, a battery management module is provided on the lithium battery pack to real-time monitor the voltage of each single battery, the temperature of the battery pack, and the number of charge and discharge cycles. When it is detected that the voltage of any single battery is less than 2.5 V or the voltage of any single battery is greater than 3.65 V, the battery protection circuit is triggered to cut off the charge and discharge circuit, and an alarm signal is sent to the building energy management platform to start the standby diesel generator set to supply power to the building's critical load. The diesel generator set shuts down after maintaining 5 minutes of transitional power supply when the voltage of the single battery is normal.

[0014] Preferably, the charge and discharge instruction set is transmitted to the bidirectional energy storage converter through an optical fiber ring network, and the transmission protocol adopts the IEC61850 GOOSE message format. Each instruction frame contains three groups of data: the start timestamp, the duration, and the charge and discharge power value. The instruction execution period is synchronized with the time resolution of the dynamic programming algorithm at 15 minutes.

[0015] Preferably, the irradiance sensor is calibrated once every quarter.

[0016] The present invention has at least the following beneficial effects: First, the building energy supply method based on solar photovoltaic energy storage provided by the present invention analyzes the time-domain difference matrix between the photovoltaic output prediction curve and the building load demand curve, and combines the dynamic programming algorithm to optimize the energy storage charge and discharge strategy, so that the lithium battery pack preferentially charges during the low electricity price period and discharges and supplies power during the peak period, improving the utilization rate of photovoltaic energy, reducing the building's dependence on the power grid at the same time, and realizing the dynamic matching of photovoltaic power generation and building load; Second, the building energy supply method based on solar photovoltaic energy storage provided by the present invention adopts a fusion correction mechanism of real-time irradiance data and weather forecast, combines the dust deposition coefficient and temperature compensation algorithm, controls the photovoltaic output prediction error, significantly improves the system response reliability under extreme weather conditions, and enhances the anti-interference ability of the photovoltaic prediction model; Third, the building energy supply method based on solar photovoltaic energy storage provided by the present invention improves the accuracy of building load prediction by establishing sub-item load models for air conditioners, lighting, and office equipment, and combining multi-parameter coupling calculations such as personnel density and ambient light. Especially when switching between holiday modes, it reduces the prediction deviation and strengthens the building load prediction accuracy; Fourth, the building energy supply method based on solar photovoltaic energy storage provided by the present invention uses a battery management module to monitor key parameters such as voltage and temperature in real time, combines a hierarchical protection mechanism and seamless switching of standby diesel generators, can reduce the risk of overcharging / overdischarging of lithium batteries, ensure the continuity of power supply for critical loads, and guarantee the safe operation of the energy storage system; Fifth, the building energy supply method based on solar photovoltaic energy storage provided by the present invention considers the time-of-use electricity price and the cost of battery life attenuation through a dynamic programming algorithm, combines a 15-minute granularity command fast response, reduces the comprehensive electricity cost of the building, extends the cycle life of lithium batteries at the same time, and optimizes the economy of photovoltaic energy; Sixth, the building energy supply method based on solar photovoltaic energy storage provided by the present invention calibrates the irradiance sensor quarterly and establishes a dust deposition compensation model, reduces the annual attenuation rate of photovoltaic efficiency, avoids the problem of inaccurate prediction models caused by equipment aging, and guarantees the continuous and efficient operation of the system.

[0017] Other advantages, objectives, and features of the present invention will be partially reflected by the following description and partially understood by those skilled in the art through the research and practice of the present invention. Description of the Drawings

[0018] Figure 1 It is a schematic diagram of the building structure of the building energy supply method according to one technical solution of the present invention; Figure 2 It is a schematic flowchart of the building energy supply method according to another technical solution of the present invention. Detailed Embodiments

[0019] The following further describes the present invention in detail with reference to the drawings, so that those skilled in the art can implement it according to the description in the specification.

[0020] It should be understood that the terms such as "having", "including", and "comprising" used herein do not exclude the existence or addition of one or more other elements or their combinations.

[0021] The present invention provides a building energy supply method based on solar photovoltaic energy storage, and the building structure for realizing the method is as follows Figure 1 shown, including: Laying a monocrystalline silicon photovoltaic module array on the surface of the building roof, and the photovoltaic module array is respectively connected to a lithium battery pack and an internal AC distribution box of the building through a bidirectional energy storage inverter, and the output end of the AC distribution box is connected to building electrical equipment; Setting a meteorological data acquisition module on the building roof, and the meteorological data acquisition module includes an irradiance sensor and a temperature sensor that respectively measure the irradiance value on the surface of the photovoltaic module and the temperature value of the back plate of the photovoltaic module; Installing a load monitoring module on each power supply branch of the building AC distribution box, and the load monitoring module collects the real-time power consumption of the air conditioning system, lighting equipment and office equipment through a current crossbar; Setting a server of a building energy management platform in the building, and a photovoltaic power generation prediction model, a building load prediction model and a energy storage control decision model are configured in the building energy management platform. The photovoltaic power generation prediction model receives the real-time measurement data of the meteorological data acquisition module, and combines the predicted irradiance value and the predicted ambient temperature value for the next 24 hours provided by the weather forecast to establish a photovoltaic output prediction curve based on time series; the building load prediction model generates a building load power demand curve that matches the time resolution of the photovoltaic power generation prediction model by analyzing the historical data of the operating power of the air conditioning system, the opening time rule data of the lighting equipment and the power consumption period distribution data of the office equipment provided by the load monitoring module, and combining the date type input by the built-in working day / holiday mode selector; Wherein, the energy storage control decision model generates a time-domain difference matrix according to the calculated photovoltaic output prediction curve and the building load power demand curve, and according to the charge and discharge period distribution of the time-domain difference matrix, adopts a dynamic programming algorithm to calculate the charge and discharge power distribution scheme of the lithium battery pack within 24 hours of the next day, and converts it into a charge and discharge instruction set in time series to control the bidirectional energy storage inverter.

[0022] In the above technical solution, a monocrystalline silicon photovoltaic module array with a conversion efficiency of 18 - 22% is laid on the surface of the building roof. The array consists of 6 modules connected in series in a string, and every 4 strings are connected in parallel to a bidirectional energy storage inverter. The DC side of the bidirectional energy storage inverter is connected to a lithium battery pack, and the AC side is connected to the internal building AC distribution box. The output end of the AC distribution box is connected to the building electrical equipment. The capacity of the lithium battery pack is configured to be 100 kWh, using lithium iron phosphate battery cells with a cycle life of up to 6000 times. The photovoltaic modules are fixed on the roof surface through aluminum alloy brackets, and the bidirectional energy storage inverter can be installed in the roof equipment room or the building basement. The present invention constructs a complete solar power generation, energy storage, and power supply link. The monocrystalline silicon photovoltaic module array can efficiently convert solar energy into electrical energy, and the bidirectional energy storage inverter realizes the bidirectional flow of electrical energy between the lithium battery pack and the building electrical equipment. It can not only store the excess electrical energy in the lithium battery pack but also supply power to the building by the lithium battery pack when the photovoltaic power is insufficient, improving the energy utilization rate and power supply stability.

[0023] In the above technical solution, a meteorological data acquisition module is set on the building roof, which includes an irradiance sensor and a temperature sensor, respectively measuring the irradiance value on the surface of the photovoltaic module and the temperature value of the backplane of the photovoltaic module. Among them, the irradiance sensor can select a silicon-based sensor with a spectral response range of 400 - 1100 nm, with a measurement accuracy of ±5%, a range of 0 - 1500 W / m², and a sampling frequency of 1 Hz. The temperature sensor can select a PT100 platinum resistance type, with a temperature measurement range of -40°C - 120°C and an accuracy of ±0.5°C. The two sensors are fixed at the center position of the backplane of the photovoltaic module through M4 stainless steel bolts, and the signal transmission line uses RVVP shielded twisted pair wires, which are laid along the roof cable tray through PVC pipes to the server of the building energy management platform. The meteorological data acquisition module can be configured with an RS485 communication interface. A load monitoring module is installed on each power supply branch of the building AC distribution box to collect the real-time power consumption of the air conditioning system, lighting equipment, and office equipment through a current bar sensor. The meteorological data acquisition module provides basic data support for subsequent energy management. Accurate meteorological data helps to more accurately predict the photovoltaic power generation, and the real-time power consumption acquisition can enable the system to understand the electricity consumption situation of the building, thereby providing a basis for the rational allocation of energy and making the energy supply and demand better match.

[0024] In the above technical solution, as Figure 2, a server of a building energy management platform is set up inside the building, and a photovoltaic power generation prediction model, a building load prediction model, and a energy storage control decision-making model are configured. The photovoltaic power generation prediction model combines real-time measurement data and weather forecast data to establish a photovoltaic power output prediction curve, and the building load prediction model generates a building load power demand curve by analyzing historical electricity consumption data and date types. This multi-model construction method can scientifically predict from two dimensions of power generation and electricity consumption, master the energy supply and demand situation in advance, provide a reliable decision-making basis for the control of the energy storage system, and avoid energy waste and shortage.

[0025] According to the above technical solution, the working process of the building energy supply method based on solar photovoltaic energy storage provided by the present invention is as follows: First, the meteorological data acquisition module and the load monitoring module continuously collect data, and transmit the irradiance value, temperature value of the photovoltaic module, and the real-time electricity consumption power of each electrical device in the building to the server of the building energy management platform. Then, the photovoltaic power generation prediction model generates a photovoltaic power output prediction curve based on the real-time data of the meteorological data acquisition module and the weather forecast information; the building load prediction model analyzes the data provided by the load monitoring module and generates a building load power demand curve in combination with the date type. Next, the energy storage control decision-making model compares the two curves, generates a time-domain difference matrix, analyzes the charge and discharge period distribution, and uses the dynamic programming algorithm to calculate the charge and discharge power distribution plan of the lithium battery pack within 24 hours of the next day, and converts it into a time-series charge and discharge instruction set. Finally, the bidirectional energy storage converter receives the charge and discharge instruction set, controls the charge and discharge process of the lithium battery pack, realizes the stable energy supply of the solar photovoltaic energy storage to the building, charges the lithium battery pack when the photovoltaic power is sufficient and meets the building electricity consumption, and supplements the power supply by the lithium battery pack when the photovoltaic power is insufficient.

[0026] According to the above technical solution, in terms of energy conservation and emission reduction, by laying a monocrystalline silicon photovoltaic module array on the building roof, solar energy is converted into electrical energy, reducing the dependence on traditional fossil energy and reducing carbon emissions during the building operation, which conforms to the concept of green building and sustainable development. At the same time, the cooperation of the bidirectional energy storage converter and the lithium battery pack can effectively store the excess electrical energy generated by the photovoltaic system, avoid energy waste, and further improve the energy utilization efficiency. From the perspective of energy management, the coordinated operation of the meteorological data acquisition module, the load monitoring module and the various models in the energy management platform realizes the accurate prediction of photovoltaic power generation and building load. The energy storage control decision-making model optimizes the charge and discharge strategy of the lithium battery pack based on the prediction results, enabling the dynamic matching of energy supply and demand, improving the stability and reliability of the building energy supply system, and reducing the impact on building electrical equipment caused by power supply fluctuations. In addition, the intelligent energy management method provided by the present invention can also reduce the electricity cost of the building by reasonably allocating energy, bringing economic benefits to the building operation.

[0027] In one of the technical solutions, when the deviation between the actual irradiance value and the predicted value monitored by the meteorological data acquisition module exceeds ±15%, the real-time correction mode of the energy storage control decision-making model is triggered. The correction mode dynamically adjusts the charge and discharge power of the lithium battery pack according to the difference between the current actual photovoltaic output and the real-time demand of the building load, and at the same time feeds back the deviation data to the photovoltaic power generation prediction model for rolling correction.

[0028] In the above technical solution, first, the irradiance sensor in the meteorological data acquisition module is used to obtain the irradiance value on the surface of the photovoltaic module in real time, and it is compared with the predicted irradiance value in the future period output by the photovoltaic power generation prediction model in real time. When the deviation between the actual value and the predicted value exceeds ±15%, such as a sudden drop in irradiance caused by a sudden rainy or cloudy weather or a sudden increase in irradiance due to strong light in the cloud gap, the system immediately triggers the real-time correction mode of the energy storage control decision model. By setting a clear deviation threshold (15%), it can not only avoid control oscillations caused by frequent triggering of corrections due to small fluctuations, but also accurately capture meteorological mutations that significantly affect photovoltaic power output, ensuring that the system starts adaptive adjustment at critical nodes. After triggering the correction mode, the energy storage control decision model will suspend the conventional charge and discharge strategy based on the prediction curve, and instead, starting from the current moment, it will collect the actual power output of the photovoltaic module in real time. Through the real-time power generation power fed back by the bidirectional energy storage converter and the real-time power consumption uploaded by the building load monitoring module (such as the instant energy consumption of devices such as air conditioners and lighting), it calculates the real-time difference between the two (i.e., the photovoltaic surplus power or power deficit). Based on this real-time difference, the model adopts dynamic response algorithms, such as model predictive control or heuristic adjustment strategies, to dynamically adjust the charge and discharge instructions within the safe charge and discharge power range of the lithium battery pack (considering battery life and safety limitations). For example, if the photovoltaic power output suddenly drops resulting in insufficient real-time power supply, the model will immediately increase the battery discharge power to make up for the gap; if sudden strong light causes photovoltaic power output surplus, the model will increase the charging power to avoid waste of electric energy. The advantage of the real-time dynamic adjustment design is that it can quickly respond to power imbalances on the time scale of seconds or minutes, avoid power outages or overcharging / overdischarging of the battery caused by the accumulation of prediction errors, and significantly improve the anti-interference ability of the system. While adjusting the charge and discharge power, the system will feedback the current irradiance deviation data (the difference between the actual value and the predicted value, deviation duration, etc.) and the corresponding photovoltaic power output correction amount to the photovoltaic power generation prediction model. After receiving the feedback data, the prediction model starts a rolling correction mechanism: that is, on the basis of the original prediction model, taking the latest measured data as input, it iteratively updates the prediction curve for the remaining future period (from the current moment to 24 hours the next day). For example, if it is detected at 10 am that the irradiance has been continuously lower than the predicted value by more than 15%, the model will automatically adjust the predicted photovoltaic power output for the afternoon of the same day to avoid control failures caused by prediction deviations in subsequent periods. The core advantage of this closed-loop feedback design is that through real-time error calibration, it continuously optimizes the parameters of the prediction model, forming a dynamic optimization closed-loop of "prediction - execution - feedback - correction", enabling the prediction accuracy to gradually improve with the running time, especially suitable for long-term stable operation in complex meteorological environments.

[0029] According to the above technical solution, when the meteorological data acquisition module real-time monitors that the irradiance deviation exceeds ±15%, the entire correction process unfolds as follows: The irradiance sensor collects data at a preset frequency and compares it with the predicted value of the corresponding period output by the prediction model. If the deviation exceeds the threshold for two consecutive times (to avoid accidental noise interference), a correction signal is sent to the energy management platform; the energy storage control decision model switches to the real-time correction mode, and the output power of the photovoltaic inverter (actual power generation value) and the real-time power consumption of the load monitoring module are obtained in real time to calculate the current power difference. If the difference is positive (surplus), the battery charging power is increased; if it is negative (deficit), the battery discharge power is increased, and the adjustment amplitude is limited by the maximum charge and discharge rate and the remaining capacity of the battery. The current deviation data (such as actual irradiance, predicted irradiance, deviation percentage, occurrence time, etc.) is input into the photovoltaic power generation prediction model. The model uses the sliding window algorithm to replace the old data with the latest measured data and retrain the prediction curve for future periods, so that subsequent control strategies are based on more accurate predicted values. After the correction is completed, the system continues to monitor the deviation in real time. If the subsequent data returns within ±15% and remains stable, the correction mode is exited and the conventional control strategy based on the prediction curve is restored; if the deviation persists, the correction mode is maintained and continuous adjustment is carried out.

[0030] According to the above technical solution, through the real-time correction mode, the system can effectively cope with the photovoltaic output fluctuations caused by sudden meteorological changes (such as short-term heavy rain, cloud cover, local strong light, etc.), and avoid power outages or energy storage system overloads caused by the accumulation of prediction errors. For example, in a cloudy summer day, the traditional system may not have enough electrical energy to supplement when the load suddenly increases due to a sudden drop in irradiance, while this solution can ensure the continuous power supply of key loads such as air conditioners by immediately adjusting the battery discharge power, significantly improving the stability of building energy supply. At the same time, the real-time correction mechanism reduces the waste of electrical energy caused by prediction deviations (such as not charging in time when there is photovoltaic surplus) or the additional cost of purchasing electricity from the grid (such as relying on the mains when the power supply is insufficient). By dynamically adjusting the charge and discharge strategy, the charge and discharge depth and frequency of the lithium battery pack are optimized, extending the battery life; at the same time, the rolling correction prediction model makes the control strategy for future periods closer to the actual situation, and the dependence of the building on the grid can be reduced during long-term operation. Especially in the peak-valley electricity price scenario, the electricity cost can be further saved by charging and discharging during off-peak hours. The feedback correction mechanism enables the entire energy management system to have the ability of self-learning. With the accumulation of operation data, the photovoltaic power generation prediction model continuously approaches the actual situation, and the accuracy of the control strategy continues to improve. The closed-loop architecture of "monitoring - control - learning" lays a foundation for the building energy supply system to operate more advanced intelligent and unmanned, and is particularly suitable for popularization and application in scenarios with high requirements for energy autonomy such as distributed energy systems and zero-carbon buildings.

[0031] In one of the technical solutions, the specific implementation steps for establishing the photovoltaic power generation prediction model based on the time series of photovoltaic output prediction curves are as follows: The meteorological data acquisition module uploads the irradiance measurement value and the photovoltaic module backplane temperature value in real time, and the meteorological bureau API interface provides the future 24-hour irradiance prediction data and environmental temperature prediction data in real time; The sliding window algorithm is used to fuse the real-time uploaded irradiance data and the irradiance prediction data provided by the meteorological bureau to generate an irradiance prediction curve: Interpolate the 1-hour resolution irradiance prediction data provided by the meteorological bureau API to a 15-minute time granularity. In each 15-minute time period, generate an irradiance prediction curve G(t) based on the irradiance prediction value provided by the meteorological bureau; calculate the deviation rate δ between the average value G1 of the real-time uploaded irradiance data in the previous 10 minutes and the irradiance prediction data G2 in the corresponding period, and adjust the irradiance prediction curve G(t): When the absolute value of δ ≤ 15%, the actual irradiance prediction curve G’(t) = G(t); When the absolute value of δ > 15%, the actual irradiance prediction curve G’(t) = G(t) × (1 + 0.7 × δ / 100); Calculate the dust deposition coefficient γ on the surface of the photovoltaic module based on the deviation between the real-time uploaded irradiance data and the theoretical irradiance value: ; Correct the photovoltaic module efficiency ε according to the real-time backplane temperature data: ; Generate a photovoltaic output prediction curve P(t) at a 15-minute time step: P(t) = ε × (1 - γ) × G’(t) × A; G L is the theoretical irradiance value, G L = G0 × cos(θ), where G0 is the theoretical solar irradiance that can be received on the surface of the photovoltaic module under ideal clear weather conditions without clouds and atmospheric pollution, W / m 2 ; θ is the solar incidence angle; ε0 is the nominal conversion efficiency of the photovoltaic module under standard test conditions; △T is the difference between the photovoltaic module backplane temperature value and the standard test temperature, and the standard test temperature = 25°C; A is the effective sunlight-receiving area of the photovoltaic module, m 2 .

[0032] In the above technical solution, first, the meteorological data acquisition module is used to obtain the irradiance and backplane temperature on the surface of the photovoltaic module in real time, and at the same time, the predicted irradiance and temperature data for the next 24 hours from the API of the meteorological bureau are accessed. For the 1-hour resolution prediction data provided by the meteorological bureau, the cubic spline interpolation method is used to refine it to a 15-minute time granularity, making the prediction curve more conform to the short-term fluctuations of actual power generation. By improving the time resolution, the problem of prediction lag caused by coarse data granularity is avoided, which is especially suitable for scenarios where the photovoltaic output changes frequently due to weather effects, providing a fine-grained data basis for subsequent precise control. Within each 15-minute period, the system calculates the deviation rate between the average value of the real-time irradiance in the previous 10 minutes and the predicted value for the corresponding period. If the deviation rate exceeds ±15%, the prediction curve is adjusted: the predicted value is weighted and corrected by introducing real-time data to make the prediction curve closer to the current actual irradiance situation. For example, when sudden rain and cloudiness cause the measured irradiance to be significantly lower than the predicted value, the system will automatically lower the predicted values for this period and subsequent related periods, avoiding misjudgment of the energy storage system due to overestimated predicted values. The advantage of this dynamic adjustment mechanism is that it can capture the impact of sudden weather changes on irradiance in real time, reduce the long-term deviation accumulation caused by the prediction model relying on fixed parameters, and improve the adaptability of the prediction model to short-term meteorological fluctuations.

[0033] In the above technical solution, the influence of dust deposition and temperature on photovoltaic efficiency is further considered: the dust deposition coefficient is calculated through the difference between the real-time irradiance and the theoretical value to quantify the reduction in power generation efficiency caused by dust shading; at the same time, according to the difference between the backplane temperature and the standard test temperature, the conversion efficiency of the photovoltaic module is dynamically corrected (the efficiency decreases as the temperature increases). For example, in summer when the temperature is high, the system will automatically lower the predicted value of the module efficiency according to the real-time temperature, avoiding overestimation of the output due to temperature effects. Through the quantitative processing of actual environmental factors, the prediction model changes from simply relying on weather forecasts to combining the real-time status of equipment, significantly improving the physical authenticity and engineering practicability of the prediction results.

[0034] According to the above technical solution, a specific working process is as follows: The meteorological sensor uploads irradiance and backplane temperature in real time, and the meteorological bureau API synchronously provides irradiance and temperature prediction data for the next 24 hours. The system first refines the 1-hour prediction data of the meteorological bureau into a 15-minute interval through interpolation to form a preliminary prediction time series. Within each 15-minute period, the average measured irradiance of the previous 10 minutes is extracted and compared with the predicted value of the corresponding period to calculate the deviation rate. If the deviation is within ±15%, the original prediction curve is directly adopted; if it exceeds the threshold, the predicted values of this period and subsequent related periods are dynamically adjusted according to the measured data according to the rules to form a corrected irradiance prediction curve. According to the difference between the real-time irradiance and the theoretical value (irradiance under dust-free and ideal weather conditions), the dust deposition coefficient (reflecting the cleanliness of the component surface) is calculated; at the same time, according to the difference between the backplane temperature and the standard test temperature of 25°C, the influence coefficient of temperature on the component efficiency is calculated. The two are combined to perform a secondary adjustment on the corrected irradiance prediction value, and finally a photovoltaic output power prediction curve considering the actual environmental impact is generated. The above process loops every 15 minutes, continuously incorporating the latest measured data through a sliding window and replacing the old data to ensure that the prediction curve is always based on the latest environmental conditions and equipment operation data, providing real-time updated input parameters for the energy storage control decision-making model.

[0035] According to the above technical solution, through multi-source data fusion, fine-grained time processing, and dynamic deviation correction, the prediction model can more accurately capture short-term fluctuations in irradiance (such as cloud cover, sudden sunny and rainy changes). At the same time, through dust and temperature correction, the influence of the physical state of the equipment on the power generation efficiency is eliminated, making the prediction curve closer to the actual photovoltaic output. Compared with the traditional model that only relies on weather forecasts, the prediction error under complex weather conditions is reduced, providing a more reliable basis for the charge and discharge decision-making of the energy storage system. Quantifying the influence of dust deposition and temperature makes the prediction model shift from "theoretical" to "engineering", automatically identifying the trend of component efficiency decline and adjusting the predicted value to avoid the failure of the charge and discharge strategy of the energy storage system caused by environmental factors (such as overcharging or under-discharging), and improving the robustness of the entire energy supply system under different climate conditions. At the same time, the prediction model is continuously corrected by real-time measured data to form a closed-loop mechanism of "data acquisition - model prediction - deviation calibration - strategy optimization". As the operation time accumulates, the model will continuously learn local meteorological characteristics and equipment characteristics, and the prediction accuracy will continue to improve, providing more intelligent decision-making support for the building energy management platform, and finally realizing the efficient utilization of solar energy resources and the refined control of the energy storage system.

[0036] In one of the technical solutions, the specific steps for the building load prediction model to generate the building load power demand curve are as follows: Obtain historical data from the load monitoring module and preprocess it: The timing matrix that records the switching times and power levels of lighting equipment in each area daily; Statistical probability distribution of the daily usage duration of office equipment; Modeling based on the base load component: For air-conditioning load: ; For lighting load: ; For office equipment load: ; Correct the building load power according to the date type: P 负荷 (t)=[P 空调 (t)+P 照明 (t)+P 办公 (t)]×σ(D); Generate the building load power demand curve: Discretize P 负荷 (t) at a 15-minute time granularity to generate a 24-hour building load power demand curve; Among them, P 空调 (0) is the reference power of the air-conditioning system, in kW, taking the average value under the condition of 25°C in historical data; △ 空调 T(t) is the difference between the real-time outdoor temperature and the set temperature; ρ(t) is the real-time personnel density, in people / m 2 ; N i is the number of the i-th type of lighting equipment; P i is the rated power of a single i-th type of lighting equipment, in kW; S i (t) is the switching state function of the i-th type of lighting equipment, 0 for off and 1 for on, which is obtained from the timing matrix of the switching times and power levels of lighting equipment in each area daily; B is the ambient light intensity measured in real time inside the building, in lux; M j is the number of the j-th type of office equipment; P standby is the standby power of the equipment, in kW; P full load is the full-load power of the equipment, in kW; F(t) is the probability of equipment usage in a time period, which is obtained from the statistical probability distribution of the daily usage duration of office equipment, and its value ranges from 0 to 1; σ(D) is the date correction coefficient, taking 1.0 on weekdays, 0.75 on weekends, and 0.5 on legal holidays.

[0037] In the above technical solution, first, the load monitoring module collects historical data of various electrical equipment in the building, including the switching time and power levels of lighting equipment, the usage duration distribution of office equipment, etc., and integrates these data into a time series matrix and a probability distribution model. For example, record the opening hours of lighting equipment in different areas every day (such as the meeting room lights on at 9 am and off at 5 pm), and the usage probabilities of office equipment (such as computers, printers) at different times (such as the usage probability decreases during the lunch break). Among them, P 空调 (0) takes at least the average value under the 25°C working condition in the historical data of one month to ensure the accuracy of the data. Through the structured historical data storage, it provides a rich sample basis for subsequent load forecasting, avoids the prediction fluctuations caused by relying only on real-time data, and improves the model's ability to capture regular electricity consumption behaviors.

[0038] In the above technical solution, the building load is further divided into three categories: air conditioning, lighting, and office equipment, and targeted prediction models are established respectively: For air conditioning load: Based on the historical reference power, it is dynamically adjusted in combination with the difference between the real-time outdoor temperature and the set temperature, and the personnel density. For example, when the outdoor temperature rises by 1°C in summer, the air conditioning load increases by a certain proportion; when the personnel are dense (such as during the meeting period), the load further increases. The air conditioning load model can accurately reflect the electricity consumption characteristics of the air conditioning system with changes in the environment and personnel, avoiding the errors caused by the fixed power assumption in the traditional model.

[0039] For lighting load: It is comprehensively calculated according to the switching state, quantity, and rated power of lighting equipment in different areas, in combination with the indoor environmental light intensity. When the natural light is sufficient (such as on a sunny day), the power of the lighting equipment is automatically reduced; at night or on a cloudy day, it is calculated according to the actual on state. The lighting load model takes into account both manual control and environmental adaptability, reducing the prediction deviation of energy waste scenarios.

[0040] For office equipment load: Considering the difference between the standby power and the actual usage power of the equipment, it is dynamically adjusted through the usage probability of different time periods. For example, the office computer runs at full load power during working hours and switches to the standby state during non-working hours. The office equipment load model automatically distributes the load proportion of different time periods according to the historical usage rules, avoiding extreme assumptions such as "full power operation" or "ignoring standby power consumption".

[0041] In the above technical solution, finally, through the built-in working day / holiday mode selector, the system corrects the total load according to the date type (working day, weekend, legal holiday). For example, the usage of office equipment decreases on weekends, and the load correction factor is set to 0.75; on legal holidays, it further drops to 0.5. By precisely matching the activity patterns of personnel on different date types, it avoids the prediction deviation caused by the "one-size-fits-all" approach of traditional models, and is especially suitable for scenarios with obvious periodic load changes such as commercial buildings and office spaces.

[0042] According to the above technical solution, a specific implementation process is as follows: The load monitoring module continuously collects the historical operation data of each electrical device, such as the switch time series of lighting devices and the daily usage duration records of office equipment, to form a structured time series matrix and probability distribution table; starting from the historical reference power, the air conditioning load is dynamically adjusted according to the difference between the real-time outdoor temperature and the set temperature and the current personnel density according to preset rules (the higher the temperature and the more people, the higher the load). The lighting load is calculated based on the real-time on / off status of lighting devices in each area (from the time series matrix), the number of devices, the rated power, and the indoor light intensity (reducing the lighting power when the light is strong) to calculate the total lighting load for each period. The load of office equipment is calculated based on the device type (such as computers, copiers), quantity, standby / full-load power, and the usage probability in different periods (such as the usage probability decreases during the lunch break) to calculate the actual power consumption of each device at different times. Finally, the three types of loads are added together to obtain the total load, and then the corresponding correction factor is applied according to the current date type (working day / weekend / holiday) (such as reducing the load by 25% on weekends), and finally discretized at a 15-minute time granularity to generate the building load power demand curve for the next 24 hours, which is completely matched with the time resolution of the photovoltaic output prediction curve.

[0043] According to the above technical solution, through classified modeling and historical data-driven, the model can accurately reflect the dynamic electricity consumption characteristics of air conditioners, lighting, and office equipment (such as air conditioners changing with temperature, lighting adjusting with light, and office equipment fluctuating with working hours). Compared with the traditional "total load averaging" prediction method, it reduces errors and accurately predicts the load peaks caused by the operation of air conditioners and the concentrated use of office equipment, avoiding power shortages in the energy storage system due to underestimated loads. The date type correction mechanism enables the system to automatically switch to different load modes: high-frequency office electricity consumption is matched on weekdays, low-load scenarios are adapted on weekends, and the predicted values of non-essential loads are further reduced on holidays. The differential design ensures the rationality of the energy storage control strategy in different scenarios, improving the stability and energy utilization efficiency of the system during long-term operation. Based on the load prediction curve with a 15-minute time granularity, it forms an accurate time match with the photovoltaic output prediction curve, providing fine-grained power difference data for the energy storage control decision-making model. This enables the charging and discharging strategy of the lithium battery pack to be accurate to each 15-minute period, avoiding energy waste or energy storage overload caused by traditional coarse-grained control, and ultimately achieving the supply-demand balance and efficient operation of the building energy supply system.

[0044] In one of the technical solutions, the method for obtaining the time domain difference matrix is as follows: The energy storage control decision-making model aligns the photovoltaic output prediction curve and the load power demand curve along the time axis, calculates the power difference at each time node, and generates the time domain difference matrix. The positive value period in the time domain difference matrix is the energy storage charging period, and the negative value period is the energy storage discharging period.

[0045] In the above technical solution, the irradiance prediction data with a 1h resolution provided by the meteorological bureau API is refined to a 15min interval through cubic spline interpolation. Each time point corresponds to the boundary of a 15min time period. The load components such as air conditioners, lighting, and office equipment are discretized at a 15min granularity to ensure complete alignment with the time points of the photovoltaic curve. By forcing a unified time granularity, the "time misalignment" caused by different time steps is avoided, enabling each node of the two curves to correspond one by one on the time axis, laying a foundation for subsequent difference calculations. Precise timestamps are added to each data point of each curve to ensure that the starting time and interval are exactly the same. When generating the curve, if there is a slight deviation between the acquisition time of real-time monitoring data (such as irradiance and load power) and the preset time node (for example, the sensor collects data once per second), the data is calibrated to the nearest 15min node through nearest neighbor interpolation or the mean method. The two aligned curves are converted into time series matrices of the same dimension. The rows of each matrix represent time nodes, and the columns represent the corresponding power values. Among them, the photovoltaic output matrix stores the predicted power generation of each 15min node, and the load demand matrix stores the predicted power consumption of each 15min node, ensuring that the number of rows (number of time nodes) and time order of the two matrices are exactly the same. When calculating the energy difference in a certain time period, the photovoltaic power and load power at the same time node can be directly obtained through matrix indexing, without additional time matching calculations, greatly improving the data processing efficiency. According to the above technical solution, the time-domain difference matrix, as a translator between photovoltaic, load, and energy storage, ensures the real-time visualization and controllability of energy flow. The time-domain difference matrix aligns the photovoltaic output and load demand curves along the time axis, converting the abstract power change into a computable numerical matrix, clearly defining the energy surplus or deficit at each time node: a positive value represents that the photovoltaic output is greater than the load demand (charging period), and a negative value represents that the load demand is greater than the photovoltaic output (discharging period). It avoids the blindness of charging and discharging based on experience in traditional extensive control, ensures timely energy replenishment of the energy storage, and thus realizes the optimal allocation of energy in the time dimension. The dynamic programming algorithm relies on the structured data of the time-domain difference matrix and can quickly identify the distribution rules of charging and discharging periods throughout the day (such as continuous charging periods, intermittent discharging periods, etc.), and then calculate the optimal power distribution plan for the lithium battery pack. Compared with unstructured original data, the data input in matrix form significantly reduces the computational complexity of the algorithm, improves the decision-making efficiency, and avoids strategy lag or resource waste caused by data fragmentation.

[0046] In one of the technical solutions, the objective function of the dynamic programming algorithm is: ; The constraint conditions are: ; Among them, S(t) is the percentage of the remaining battery capacity of the lithium battery pack in the total capacity; C(t) is the time-of-use electricity price of the power grid in period t, yuan / kWh; H(t) is the electricity purchase quantity from the power grid in period t, kW; ; β is the battery state of charge balance weight factor, taking 0.05; P ch (t) is the charging power in period t, kW; P di (t) is the discharging power in period t, kW; K(t) is the power of photovoltaic power generation exceeding the load demand in period t, kW; L(t) is the power of the load demand exceeding the photovoltaic power generation in period t, kW; P ch-max is the maximum charging power of the lithium battery pack, kW; P dis-max is the maximum discharging power of the lithium battery pack, kW; E 总 is the total capacity of the lithium battery pack, kWh; △t = 15 min.

[0047] In the above technical solution, the present invention further constructs an objective function including the power grid electricity purchase cost and the battery state stability, realizes the balance between the economy and safety of the energy storage system, and simultaneously realizes the minimization of the electricity purchase cost and the stabilization of the battery state. In terms of minimizing the electricity purchase cost, with the electricity prices of each period as the weights, the system is guided to preferentially use photovoltaic electric energy and energy storage electricity, reduce the electricity purchase expenditure during peak periods. For example, during peak electricity price periods, the battery discharging power is increased as much as possible and the power grid electricity purchase is reduced; in terms of stabilizing the battery state, by setting a deviation penalty term for the remaining battery power and the ideal state (55%), overcharging and over-discharging of the battery (the battery power is lower than 20% or higher than 90%) are avoided, and the service life of the lithium battery is extended. Both the short-term electricity cost and the long-term operation reliability of the battery are considered.

[0048] In the above technical solution, the safety and controllability of the energy storage system are ensured by three sets of constraint conditions: the power range constraint, the power boundary constraint, and the state transition constraint. The power range constraint limits the remaining battery power between 20% and 90% to prevent overcharging (damaging the battery life) and over-discharging (causing power supply interruption). Especially when the photovoltaic output is insufficient in extreme weather, sufficient power is reserved to cope with sudden loads. The power boundary constraint sets the maximum charge and discharge power according to the physical characteristics of the battery to avoid potential safety hazards (overheating, short circuit) caused by current exceeding the limit. At the same time, combined with the real-time photovoltaic surplus or load deficit, the feasible power interval is dynamically adjusted. The state transition constraint ensures that the charge and discharge behavior in each time period accurately reflects the change of the battery state (charging increases the power, discharging decreases the power) through the power balance equation, and considers the loss of charge and discharge efficiency (charging efficiency 95%, discharging efficiency 97%) to make the model closer to the actual operation characteristics.

[0049] In the above technical solution, the dynamic programming algorithm uses a 15-minute time step, which is exactly the same as the time resolution of the photovoltaic output prediction curve and the load demand curve (96 time period data). The energy flow is processed period by period to accurately capture the power fluctuations in each short cycle (such as the photovoltaic mutation caused by the start and stop of air conditioners and cloud cover), and avoid strategy lag caused by coarse time granularity. When a sudden increase in photovoltaic output is detected, the charging power in that time period can be immediately adjusted to make full use of the sudden electric energy.

[0050] According to the above technical solution, a specific implementation process is as follows: Input the predicted curve of photovoltaic power output for the next 24 hours, the building load demand curve (both with a granularity of 15 minutes), the grid peak-valley electricity price data (C(t) takes a high price during the day and a low price at night on weekdays), and the basic parameters of the lithium battery pack (total capacity, maximum charge and discharge power). Set the initial remaining battery power S(0) (obtained according to the state at the end of the previous day), and determine the ideal power target (55%) as the benchmark for deviation penalty. Starting from the first time period (00:00 - 00:15), judge whether the current time period is a charging period (photovoltaic surplus, the difference is positive) or a discharging period (load deficit, the difference is negative) according to the time-domain difference matrix. For the charging period: On the premise of meeting the maximum charging power and the upper limit of the battery capacity, calculate the optimal charging power to minimize the comprehensive cost of the electricity purchase cost (if electricity needs to be purchased from the grid) and the deviation of the battery state; for the discharging period: Under the constraints of the maximum discharging power and the lower limit of the battery capacity, determine the optimal discharging power, give priority to meeting the load demand, and purchase electricity from the grid for the insufficient part. Update the remaining battery power S(t + 1) at the next moment through the state transition equation, and recursively calculate the optimal strategies for all subsequent time periods to ensure that the impact of the current decision on future time periods is fully considered (overcharging during the day may result in no electricity available at night, automatically balancing the electricity distribution throughout the day). After traversing all 24 hours of time periods, output the charge and discharge power values for each 15-minute time period as the charge and discharge power distribution plan for the next 24 hours, convert it into the control command of the bidirectional energy storage converter, and record the electricity purchase amount from the grid for each time period for subsequent cost accounting.

[0051] According to the above technical solution, through the peak-valley electricity price guidance and the optimal dispatching of the energy storage system, the economic operation of "peak shaving and valley filling" is realized: charging is prioritized during the period of photovoltaic surplus and low electricity price, and discharging is prioritized during the period of insufficient photovoltaic power and high electricity price, reducing the proportion of high-price electricity purchase, and is especially suitable for high-load scenarios such as commercial buildings. The battery state stability constraint avoids excessive power consumption or saturation (controlled within the range of 20% - 90%), and combined with the charge and discharge power limit, reduces the performance degradation of the battery caused by deep charge and discharge or current over-limit, and reduces the operation and maintenance cost and replacement frequency of the energy storage system. At the same time, the global optimization characteristics of the dynamic programming algorithm enable the system to cope with complex energy flow scenarios (such as intermittent photovoltaic output and random load fluctuations), and ensure the continuity and stability of power supply by adjusting the charge and discharge strategy in real time. At the same time, the modular design of the objective function and constraint conditions reserves an expansion space for the subsequent introduction of more complex optimization objectives (such as carbon emissions and equipment maintenance costs), and helps the building energy supply system to upgrade to multi-objective intelligent optimization. At the same time, the accurate data of photovoltaic output prediction and load demand prediction are used to clarify the energy supply and demand relationship through the time-domain difference matrix, forming a complete chain of "prediction - optimization - control". The multi-module cooperation mechanism enables each charge and discharge decision of the energy storage system to be based on the global optimal solution rather than local adjustment, and finally realizes the maximum utilization of solar energy resources and the refined management of building energy consumption.

[0052] In one of the technical solutions, a battery management module is provided on the lithium battery pack to monitor the voltage of each single battery, the temperature of the battery pack and the number of charge and discharge cycles in real time. When it is detected that the voltage of any single battery is less than 2.5V or the voltage of any single battery is greater than 3.65V, the battery protection circuit is triggered to cut off the charge and discharge circuit, and an alarm signal is sent to the building energy management platform to start the standby diesel generator set to supply power to the building's critical loads. The diesel generator set shuts down after maintaining 5 minutes of transitional power supply when the voltage of the single battery is normal.

[0053] In the above technical solution, the battery management module monitors the single-cell battery voltage (accuracy up to ±0.01 V), the battery pack temperature (resolution ±1 °C), and the charge and discharge cycle times in real time, and can accurately capture the abnormal states of lithium batteries (the over-discharge critical value with a single-cell voltage lower than 2.5 V or the overcharge risk higher than 3.65 V). Once a voltage anomaly is detected, the protection circuit is immediately triggered to cut off the charge and discharge circuit, blocking the development of dangerous working conditions at the physical level and reducing the incidence of safety accidents such as battery thermal runaway and short-circuit fires. When the battery protection circuit is activated, the system synchronously triggers the standby diesel generator to cut in to ensure the continuous power supply of key loads such as the building's fire protection system, emergency lighting, and data servers. The start logic of the diesel generator is designed to be "triggered immediately", and the response time is controlled within 10 s to avoid power outages caused by energy storage system failures (such as power outages in hospital operating rooms and data loss in computer rooms). After the voltage returns to normal, a 5-minute transitional power supply is maintained to provide a safety recovery buffer period for the lithium battery pack. After the single-cell voltage returns to the normal range (2.5 V - 3.65 V), the diesel generator continues to supply power for 5 minutes to prevent secondary anomalies in the battery state caused by grid or photovoltaic system fluctuations, and at the same time provides a stable voltage environment for the restart of the energy storage system to avoid hardware damage to key equipment caused by sudden power supply start and stop. At the same time, the linkage control between the standby diesel generator and the energy storage system forms a seamless switch of the "photovoltaic - energy storage - diesel" three-level power supply system. Under normal working conditions, photovoltaic power is preferentially used, the energy storage system is responsible for peak shaving and valley filling, and it automatically switches to diesel power generation in case of battery failure to ensure that key loads are continuously powered. After the battery is repaired, the diesel generator supplies power for 5 minutes, and then exits after the energy storage system is stably connected. When the single-cell battery voltage is abnormal, the battery management module synchronously sends an alarm signal containing the fault type, occurrence time, and battery number to the building energy management platform to help maintenance personnel quickly locate the fault point and carry out repairs.

[0054] In one of the technical solutions, the charge and discharge instruction set is transmitted to the bidirectional energy storage converter through an optical fiber ring network, and the transmission protocol adopts the IEC61850 GOOSE message format. Each instruction frame contains three groups of data: start timestamp, duration, and charge and discharge power value. The instruction execution cycle is synchronized with the time resolution of the dynamic programming algorithm at 15 minutes.

[0055] In the above technical solution, by adopting the optical fiber ring network and IEC61850 GOOSE message technology, the transmission reliability and execution accuracy of energy storage control instructions are significantly improved: the anti-electromagnetic interference characteristics and ring network self-healing mechanism of the optical fiber ring network ensure the stable transmission of instructions in a complex building environment, and the bit error rate is as low as 10 -12Below a certain level, it can avoid the out-of-control charge and discharge caused by communication failures; the sub-millisecond transmission delay and strict time synchronization mechanism of GOOSE messages enable the instruction frames (including timestamps, durations, and power values) to be completely aligned with the 15-minute time granularity of the dynamic programming algorithm, ensuring that the converter accurately performs charge and discharge operations at the preset time period boundaries and eliminating problems such as instruction misalignment or delay. It further enhances the system interoperability through a standardized communication architecture: the IEC61850 protocol supports plug-and-play of devices, solves the compatibility problem of multi-brand devices, and lays a foundation for future expansion of distributed energy access such as wind power and hydrogen energy; the unified instruction format reduces the operation and maintenance complexity and improves the fault troubleshooting efficiency. Combined with the high reliability of the fiber optic ring network, especially in extreme electromagnetic interference or equipment failure scenarios, it can ensure the safe operation of the lithium battery pack and the real-time response of critical instructions, providing core support for the intelligent and standardized upgrade of the building energy supply system.

[0056] In one of the technical solutions, the irradiance sensor is calibrated once every quarter, ensuring the accuracy and reliability of the irradiance measurement data during long-term operation: regular calibration can eliminate measurement deviations caused by dust accumulation on the sensor surface, component aging, or environmental factors (such as temperature drift), keeping the irradiance data within a high-precision range at all times, providing stable basic data for the photovoltaic power generation prediction model, avoiding the expansion of photovoltaic output prediction deviations caused by the decrease in sensor accuracy, and thus ensuring that the energy storage control decision-making model optimizes the charge and discharge strategy based on the real irradiance situation, ultimately improving the solar energy utilization rate.

[0057] The number of devices and the processing scale described here are used to simplify the description of the present invention. The application, modification, and variation of the building energy supply method based on solar photovoltaic energy storage of the present invention are obvious to those skilled in the art.

[0058] Although the embodiments of the present invention have been disclosed as above, it is not limited to the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the field, additional modifications can be easily made. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to the specific details and the illustrated and described examples here.

Claims

1. A building energy supply method based on solar photovoltaic energy storage, characterized in that, Including: Laying a monocrystalline silicon photovoltaic module array on the surface of the building roof. The photovoltaic module array is respectively connected to a lithium battery pack and the internal AC distribution box of the building through a bidirectional energy storage inverter, and the output end of the AC distribution box is connected to the building electrical equipment; Setting a meteorological data acquisition module on the building roof. The meteorological data acquisition module includes an irradiance sensor and a temperature sensor that respectively measure the irradiance value on the surface of the photovoltaic module and the temperature value of the backplane of the photovoltaic module; Installing a load monitoring module on each power supply branch of the building AC distribution box. The load monitoring module collects the real-time power consumption of the air conditioning system, lighting equipment, and office equipment through a current crossbar; Setting a server of the building energy management platform in the building. In the building energy management platform, a photovoltaic power generation prediction model, a building load prediction model, and an energy storage control decision model are configured. The photovoltaic power generation prediction model receives the real-time measurement data of the meteorological data acquisition module, and combines the predicted irradiance value and the predicted ambient temperature value for the next 24 hours provided by the weather forecast to establish a photovoltaic output prediction curve based on time series; The building load prediction model generates a building load power demand curve that matches the time resolution of the photovoltaic power generation prediction model by analyzing the historical data of the operating power of the air conditioning system, the opening time rule data of the lighting equipment, and the power consumption period distribution data of the office equipment provided by the load monitoring module, and combining the date type input by the built-in working day / holiday mode selector; Among them, the energy storage control decision model generates a time domain difference matrix according to the calculated photovoltaic output prediction curve and the building load power demand curve, and according to the charge and discharge period distribution of the time domain difference matrix, uses the dynamic programming algorithm to calculate the charge and discharge power distribution plan of the lithium battery pack within the next 24 hours, and converts it into a time series charge and discharge instruction set to control the bidirectional energy storage inverter; 2. The building energy supply method based on solar photovoltaic energy storage according to claim 1, characterized in that When the deviation between the actual irradiance value and the predicted value monitored by the meteorological data acquisition module in real time exceeds ±15%, the real-time correction mode of the energy storage control decision model is triggered. The correction mode dynamically adjusts the charge and discharge power of the lithium battery pack according to the difference between the current actual photovoltaic output and the real-time demand of the building load, and at the same time feeds back the deviation data to the photovoltaic power generation prediction model for rolling correction; 3. The building energy supply method based on solar photovoltaic energy storage according to claim 1, characterized in that, The specific implementation steps for the photovoltaic power generation prediction model to establish a photovoltaic output prediction curve based on time series are as follows: The meteorological data acquisition module uploads the irradiance measurement value and the temperature value of the backplane of the photovoltaic module in real time, and the meteorological bureau API interface provides the predicted irradiance data and the predicted ambient temperature data for the next 24 hours in real time; Using the sliding window algorithm to fuse the real-time uploaded irradiance data and the predicted irradiance data provided by the meteorological bureau to generate an irradiance prediction curve: Interpolate the 1-hour resolution irradiance prediction data provided by the meteorological bureau API to a 15-minute time granularity. Within each 15-minute time period, generate an irradiance prediction curve G(t) based on the irradiance prediction values provided by the meteorological bureau; calculate the deviation rate δ between the average value G1 of the real-time uploaded irradiance data for the first 10 minutes and the irradiance prediction data G2 for the corresponding period, and adjust the irradiance prediction curve G(t): When the absolute value of δ ≤ 15%, the actual irradiance prediction curve G’(t) = G(t); When the absolute value of δ > 15%, the actual irradiance prediction curve G’(t) = G(t) × (1 + 0.7 × δ / 100); Calculate the dust deposition coefficient γ on the surface of the photovoltaic module based on the deviation between the real-time uploaded irradiance data and the theoretical irradiance value; ； Correct the efficiency ε of the photovoltaic module according to the real-time backplane temperature data: ; Generate a photovoltaic power output prediction curve P(t) at a 15-minute time step: P(t) = ε × (1 - γ) × G’(t) × A; G L is the theoretical irradiance value, G L = G0 × cos(θ), where G0 is the theoretical solar irradiance that can be received on the surface of the photovoltaic module under ideal clear weather conditions without clouds and atmospheric pollution, W / m 2 ; θ is the solar incidence angle; ε0 is the nominal conversion efficiency of the photovoltaic module under standard test conditions; △T is the difference between the backplane temperature value of the photovoltaic module and the standard test temperature, and the standard test temperature = 25°C; A is the effective sunlight-receiving area of the photovoltaic module, m 2 .

4. The building energy supply method based on solar photovoltaic energy storage according to claim 3, wherein The specific steps for the building load prediction model to generate the building load power demand curve are as follows: Obtain historical data from the load monitoring module and preprocess it: Record the time sequence matrix of the switching times and power levels of lighting equipment in each area every day; Statistically analyze the probability distribution of the daily usage duration of office equipment; Model based on the basic load component: For air-conditioning load: (t)=P 空调 (0)×[1 + 0.08×△ 空调 T(t)+0.05(ρ(t)-0.2)]; For lighting loads: For the load of office equipment: ; Correct the building load power according to the date type: P 负荷 (t)=[P 空调 (t)+P 照明 (t)+P 办公 (t)]×σ(D); Generate the building load power demand curve: Discretize P 负荷 (t) at a 15-minute time granularity to generate a 24-hour building load power demand curve; Among them, P 空调 (0) is the reference power of the air conditioning system, in kW, and is taken as the average value under the condition of 25 °C in historical data; △ 空调 T(t) is the difference between the real-time outdoor temperature and the set temperature; ρ(t) is the real-time personnel density, person / m 2 ; N i is the quantity of the i-th type of lighting device; P i is the rated power of a single lighting device of the i-th type, in kW; S i (t) is the switching state function of the i-th type of lighting equipment, where 0 represents off and 1 represents on, and it is obtained from the time series matrix of the switching time and power level of the lighting equipment in each area every day; B is the ambient light intensity measured in real time inside the building, in lux; M j is the quantity of the j-th type of office equipment; Pstandby is the standby power of the equipment, in kW; Pfull load is the full-load power of the equipment, in kW; F(t) is the probability of equipment usage in a time period, which is obtained from the probability distribution of the daily usage duration of office equipment, and takes values from 0 to 1; σ(D) is the date correction coefficient, taking 1.0 on weekdays, 0.75 on weekends, and 0.5 on legal holidays.

5. The building energy supply method based on solar photovoltaic energy storage according to claim 4, wherein The method for obtaining the time domain difference matrix is as follows: The energy storage control decision model aligns the photovoltaic power output prediction curve and the load power demand curve along the time axis, calculates the power difference at each time node, and generates the time domain difference matrix. The positive value periods in the time domain difference matrix are the energy storage charging periods, and the negative value periods are the energy storage discharging periods.

6. The building energy supply method based on solar photovoltaic energy storage according to claim 5, characterized in that, The objective function of the dynamic programming algorithm is as follows: ; The constraints are as follows: ; Among them, S(t) is the percentage of the remaining battery capacity of the lithium battery pack in the total capacity; C(t) is the time-of-use electricity price of the power grid at time t, in yuan / kWh; H(t) is the electricity purchased from the power grid during period t, in kW; H(t) = max(0, ; β is the battery state of charge equalization weight factor, taking 0.05; P ch (t) is the charging power in the t time period, kW; P di (t) is the discharge power in the t period, in kW; K(t) is the power of photovoltaic power generation exceeding the load demand at time t, in kW; L(t) is the power of the load demand exceeding the photovoltaic power generation at time t, in kW; P ch-max is the maximum charging power of the lithium battery pack, in kW; P dis-max is the maximum discharge power of the lithium battery pack, in kW; E 总 Total capacity of the lithium battery pack, kWh; △t = 15 minutes.

7. The building energy supply method based on solar photovoltaic energy storage according to claim 6, characterized in that, The lithium battery pack is equipped with a battery management module to monitor the voltage of each single battery, the temperature of the battery pack, and the number of charge and discharge cycles in real time. When it is detected that the voltage of any single battery is less than 2.5V or the voltage of any single battery is greater than 3.65V, the battery protection circuit is triggered to cut off the charge and discharge circuit, and an alarm signal is sent to the building energy management platform to start the standby diesel generator set to supply power to the building's critical loads. The diesel generator set shuts down after maintaining 5 minutes of transitional power supply when the voltage of the single battery returns to normal.

8. The building energy supply method based on solar photovoltaic energy storage according to claim 7, characterized in that, The charge and discharge instruction set is transmitted to the bidirectional energy storage converter through the fiber optic ring network. The transmission protocol adopts the IEC61850 GOOSE message format. Each instruction frame contains three groups of data: the start timestamp, the duration, and the charge and discharge power value. The instruction execution period is synchronized with the time resolution of the dynamic programming algorithm at 15 minutes.

9. The building energy supply method based on solar photovoltaic energy storage according to claim 8, characterized in that, The irradiance sensor is calibrated once every quarter.

Images