Heat dissipation control method and system for stable operation of photovoltaic inverter

Abstract

The invention discloses a heat dissipation control method and system for stable operation of a photovoltaic inverter, and relates to the technical field of photovoltaic inverter control, and the method comprises the steps: collecting the multi-source data of heat dissipation equipment and elements in the use of the photovoltaic inverter; constructing a feature relation model and an LSTM neural network prediction model based on data acquisition; and through the constructed feature relation model and the LSTM neural network prediction model, analysis is carried out to compare different levels of heat dissipation power with the inverter efficiency improvement degree under different environment temperatures and power levels, a predicted temperature is formed, risk levels are divided, and the optimal heat dissipation level is determined. According to the method, the predicted temperature is graded by utilizing synchronous acquisition of multi-source data and constructing the feature relation model and the LSTM neural network prediction model, and the multi-mode heat dissipation strategy is formulated, so that the heat dissipation strategy is adaptively adjusted according to the state of the photovoltaic inverter to perform adaptive heat dissipation, and the heat dissipation efficiency is improved. And the heat dissipation effect of the photovoltaic inverter is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic inverter control, and particularly relates to a heat dissipation control method and system for the stable operation of a photovoltaic inverter. Background Art

[0002] A photovoltaic inverter is usually a commonly used device that converts low-voltage direct current generated by photovoltaic cells into alternating current meeting grid requirements. The photovoltaic inverter is a core component in a photovoltaic system and is one of the important system balances in a photovoltaic array system.

[0003] Internally, a photovoltaic inverter of the prior art usually has multiple components, including IGBTs and capacitors. When the IGBTs, capacitors, and other components are working, they usually generate heat. Therefore, it is necessary to use heat dissipation equipment to control the heat dissipation of the photovoltaic inverter. When the prior art controls the heat dissipation equipment for heat dissipation control, it usually uses a single heat dissipation mode to dissipate heat from the components, which is usually not convenient for dividing the risk levels of the predicted temperatures of the components, resulting in a single heat dissipation strategy. At the same time, affected by external dust, there are situations of insufficient heat dissipation and poor environmental adaptability. Summary of the Invention

[0004] The purpose of the present invention is to provide a heat dissipation control method and system for the stable operation of a photovoltaic inverter to solve the problems mentioned in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solution: A heat dissipation control method for the stable operation of a photovoltaic inverter, the method is applied to a photovoltaic inverter, and the photovoltaic inverter includes a heat dissipation device and components. The method includes the following specific steps:

[0006] Step 1: Collect multi-source data of the heat dissipation device and components during the use of the photovoltaic inverter, and perform it synchronously, and preprocess the collected data;

[0007] Step 2: Based on the data collection, construct a feature relationship model and an LSTM neural network prediction model;

[0008] Step 3: Through the constructed feature relationship model and LSTM neural network prediction model, analyze under different environmental temperatures and power levels, compare different levels of heat dissipation power and the degree of improvement in inverter efficiency, form a predicted temperature to divide risk levels and determine the optimal heat dissipation level, and according to the determined optimal heat dissipation level, control the hierarchical heat dissipation device to achieve the temperature reduction of the photovoltaic inverter;

[0009] Step 4: Develop a multi-modal heat dissipation strategy, and adaptively adjust the heat dissipation strategy according to the real-time operating state of the photovoltaic inverter and external environmental conditions;

[0010] Step 5: Make the operating environment of the PV inverter adaptable and self-maintaining, conduct health monitoring and life prediction on key components in the heat dissipation equipment, establish a remote monitoring platform, and achieve remote monitoring, operation, and maintenance of the PV inverter and the heat dissipation system;

[0011] Step 6: Record the data on the correlation between heat dissipation and inverter power within a set time, and conduct in-depth analysis to discover potential heat dissipation problems and optimization spaces.

[0012] Preferably, the multi-source data collected in Step 1 includes temperature data, environmental data, and electrical parameters. The multi-source data is collected by deploying high-precision thermocouples at the heat generation points, installing infrared thermal imaging cameras outside the heat dissipation equipment, and installing light intensity sensors, dust sensors, and current transformers.

[0013] Preferably, the data preprocessing in Step 1 includes data synchronization and alignment, noise filtering and outlier processing, and feature engineering. The data synchronization and alignment uses the timestamp interpolation method to interpolate low-frequency data along the high-frequency time axis, and realizes multi-channel data synchronous acquisition through FPGA hardware. The noise filtering and outlier processing uses filtering algorithms, including adaptive Kalman filtering and wavelet transform denoising. The feature engineering includes spatial features and temporal features. The formula of the timestamp interpolation method is as follows:

[0014] Where, refers to the interpolated temperature value, refers to the current interpolation target time point, refers to the timestamp of the previous frame of data of the thermal imager, refers to the timestamp of the next frame of data of the thermal imager, refers to the thermal imager at the temperature value collected at the moment, refers to the thermal imager at the temperature value collected at the moment.

[0015] Preferably, the operation of the LSTM neural network prediction model in Step 2 includes an input layer and model inference. The input layer includes a historical temperature sequence, an environmental parameter matrix, and an electrical load feature. The model inference is used by deploying a lightweight LSTM model on an edge computing unit to predict the temperature change curve and potential hot spot coordinates of each heat generation point within the next five minutes. The construction steps of the LSTM neural network prediction model are as follows:

[0016] S1: Define and preprocess the input data, input data dimension, and then use data fusion to align the low-frequency thermal imager data to the high-frequency time axis through time interpolation to form a multi-dimensional tensor with a unified timestamp;

[0017] S2: Design the LSTM model architecture, set the network topology structure, and implement and embed environmental conditions using the attention mechanism;

[0018] S3: Train and optimize the L model, design the loss function, formulate the training strategy and perform hardware acceleration. The training strategy includes data augmentation and training parameters;

[0019] S4: Edge deployment and lightweighting, optimize the Jetson Nano deployment through model compression technology;

[0020] S5: Model verification and iteration, perform offline metric verification, and establish an online self-learning mechanism.

[0021] Preferably, the risk levels divided by the predicted temperature in step three include the first temperature level, the second temperature level, and the third temperature level. The temperatures of the first temperature level, the second temperature level, and the third temperature level are less than sixty degrees Celsius, between sixty and eighty degrees Celsius, and greater than eighty degrees Celsius respectively. The first temperature level is the normal temperature range, only passive cooling is required, and it is set to determine the optimal cooling level. The second temperature level requires active intervention of the cooling device, and the third temperature level requires the cooling device to start emergency composite cooling.

[0022] Preferably, the multi-modal cooling strategy formulated in step four includes a passive cooling mode, an active cooling mode, and a composite cooling mode, and they respectively correspond to the risk levels divided by the predicted temperature.

[0023] Preferably, the passive cooling mode uses a graphene composite material to wrap the components, absorb instantaneous heat, and utilize air convection. The active cooling mode uses the dynamic brushless DC fan speed regulation and airflow path optimization of the cooling device for heat dissipation. The dynamic brushless DC fan speed regulation is based on the PID parameter self-tuning of the predicted temperature curve and accelerates in advance according to the predicted demand to prevent temperature overshoot. The formula included in the active cooling mode is as follows:

[0024] where, refers to the proportionality coefficient, refers to the integral coefficient, refers to the differential coefficient, refers to the temperature deviation, refers to the integral of the temperature deviation, refers to the rate of change of the temperature deviation.

[0025] Preferably, the composite heat dissipation mode uses a semiconductor refrigeration sheet of a heat dissipation device and a liquid-cooled microchannel for heat dissipation, and an over-temperature protection linkage is set at the same time. The semiconductor refrigeration sheet deploys a TEC array in the predicted hot spot area, and the cold end temperature is controlled to below 50 °C through PID control. The liquid-cooled microchannel drives an ethylene glycol solution to flow through the microchannel embedded with power devices by starting a micro magnetic drive pump to perform liquid-cooled heat dissipation on the components.

[0026] Preferably, the use environment adaptation and self-maintenance of the photovoltaic inverter in step five are achieved by setting a dust self-cleaning process and anti-condensation control. The dust self-cleaning process includes electrostatic adsorption and reverse pulse cleaning. The operation mode of electrostatic adsorption is to energize the high-voltage electrostatic grid at the air inlet for 30 seconds to adsorb dust particles in the air. The operation mode of reverse pulse cleaning is to trigger a reverse purge of compressed air pulses for 3 seconds when the differential pressure sensor detects a 20% increase in the filter resistance.

[0027] A heat dissipation control system for the stable operation of a photovoltaic inverter, comprising:

[0028] A data acquisition and preprocessing module, which is used to acquire multi-source data of heat dissipation devices and components during the use of the photovoltaic inverter, and perform the acquisition synchronously, and preprocess the acquired data;

[0029] A model construction module, which is used to construct a feature relationship model and an LSTM neural network prediction model based on the data acquisition;

[0030] An analysis module, which is used to analyze and compare the heat dissipation power at different levels and the improvement degree of the inverter efficiency under different environmental temperatures and power levels, form a predicted temperature to divide risk levels and determine the optimal heat dissipation level;

[0031] A strategy formulation module, which is used to formulate a multi-modal heat dissipation strategy and adaptively adjust the heat dissipation strategy according to the real-time operating state of the photovoltaic inverter and external environmental conditions;

[0032] A monitoring module, which is used to adapt the use environment of the photovoltaic inverter to self-maintenance, and perform health monitoring and life prediction on key components in the heat dissipation device, establish a remote monitoring platform, and realize remote monitoring and operation and maintenance of the photovoltaic inverter and the heat dissipation system;

[0033] A data storage and analysis module, which is used to record the correlation data between heat dissipation and inverter power within a set time, and conduct in-depth analysis to discover potential heat dissipation problems and optimization space.

[0034] The technical effects and advantages of the present invention:

[0035] (1) The present invention constructs a feature relationship model and an LSTM neural network prediction model to classify the predicted temperature by using synchronous acquisition of multi-source data and performing preprocessing, and formulates a multi-modal heat dissipation strategy, thereby facilitating the adaptation and adjustment of the heat dissipation strategy according to the state of the photovoltaic inverter for adaptive heat dissipation, and further improving the heat dissipation effect of the photovoltaic inverter;

[0036] (2) The present invention enables the use environment of the photovoltaic inverter to be self-adaptive and self-maintaining. By detecting dust at the air inlet of the photovoltaic inverter and cleaning the dust, and at the same time, health monitoring and life prediction are performed on key components in the heat dissipation device, and a remote monitoring platform is established, which is beneficial to improving the environmental adaptability of the photovoltaic inverter and facilitating remote monitoring of the components in the heat dissipation device for convenient and timely operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a schematic flow chart of the heat dissipation control method for the photovoltaic transformer of the present invention.

[0038] Figure 2 It is a schematic flow chart of the construction of the LSTM neural network prediction model of the present invention.

[0039] Figure 3 It is a block diagram of the heat dissipation control system for the photovoltaic transformer of the present invention.

[0040] Figure 4 It is a diagram of the network topology structure relationship of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0042] Embodiment 1

[0043] The present invention provides as Figures 1-4A heat dissipation control method for stable operation of a photovoltaic inverter is shown, including that the method is applied to the photovoltaic inverter, which includes a heat dissipation device and components. The heat dissipation device includes a radiator, fins, a brushless DC fan, a liquid-cooled microchannel, and a thermoelectric cooler. The components include the back of the IGBT chip, DC bus capacitors, and inductance coils, all of which are key heat sources. Patch-type PT100 platinum resistors, infrared thermopile sensors, and fiber Bragg grating temperature sensors are respectively installed on them to collect the junction temperature of the IGBT module chip, the temperature of the DC bus capacitor housing, and the temperature of the filter inductance coil, and are respectively used to monitor the instantaneous temperature rise caused by switching losses, evaluate the aging and thermal stress of electrolytic capacitors, and detect the eddy current heating caused by high-frequency harmonics. An infrared thermal imager, a DS18B20 digital temperature sensor, and a Hall effect sensor are respectively installed on the heat dissipation device, and the installation positions are respectively facing the radiator fin area, embedded in the inner wall of the microchannel pipeline, and the axis of the brushless DC fan motor, which are respectively used to collect the surface temperature field of the radiator, the inlet and outlet temperatures of the liquid-cooled microchannel, and the fan speed, and are respectively used to capture the non-uniformity of temperature distribution, calculate the heat exchange efficiency of the liquid-cooled system, and verify the response accuracy of PID control. In addition, a laser particle sensor is installed inside the air inlet grille of the radiator to collect the dust concentration at the air inlet to trigger the judgment of the self-cleaning threshold. A capacitive humidity sensor is installed on the top of the photovoltaic inverter housing to collect environmental humidity data for inputting the anti-condensation control strategy. A Hall voltage is installed on the AC output busbar to collect the output power data of the inverter to correlate the relationship between the load and the heat generation. The method includes the following specific steps:

[0044] Step 1: Collect multi-source data of the heat dissipation device and components during the use of the photovoltaic inverter. The collected multi-source data includes temperature data, environmental data, and electrical parameters. The multi-source data is collected by deploying high-precision thermocouples at the heat sources, installing an infrared forming camera outside the heat dissipation device, and installing a light intensity sensor, a dust sensor, and a current transformer. The dust sensor is beneficial to detecting the particulate concentration at the air inlet using the laser scattering principle, and the current transformer is beneficial to collecting the load current waveform and calculating the correlation between the harmonic components and the heat generation, and they are carried out synchronously, and the collected data is preprocessed;

[0045] Step 2: Based on data collection, construct a feature relationship model and an LSTM neural network prediction model. The feature relationship model is constructed to reflect the relationship between the heat dissipation requirements and efficiency changes of the inverter under different ambient temperatures and power levels. The construction method of the feature relationship model is to determine the key features (photovoltaic inverter power and heat generation, heat dissipation equipment parameters, ambient temperature, and internal temperature distribution of the photovoltaic inverter), and then based on the principles of thermodynamics, a mathematical model of inverter heat dissipation can be established. In the model, the inverter power, heat dissipation method, radiator area, fan speed, etc. are used as input variables, and the internal temperature of the inverter, heat dissipation efficiency, etc. are used as output variables. Through theoretical analysis or experimental data, a relationship formula between the input variables and output variables is established. According to actual needs, the model is optimized, and then through actual experiments, the accuracy and reliability of the model are verified. The construction steps of the LSTM neural network prediction model are as follows:

[0046] S1: Define and preprocess the input data. The input data dimension includes the IGBT junction temperature sequence, the surface temperature field of the radiator, the harmonic distortion rate of the load current, and the ambient temperature and humidity. Then, data fusion is used to align the low-frequency thermal imager data to the high-frequency time axis through time interpolation to form a multi-dimensional tensor with a unified timestamp. Then, PCA is applied to the thermal imager pixel data to retain the principal components with 95% energy proportion and compress them into an 8-dimensional vector.

[0047] S2: Design the LSTM model architecture, set the network topology structure, as shown in Figure 4 The network topology structure includes an input layer, a spatial attention layer, a bidirectional LSTM layer, a temporal attention layer, and a fully connected output layer, which are used to receive 30 minutes of historical data, dynamically weight the importance of the PCA principal components of the thermal imager, capture the long-term and short-term temperature evolution laws, focus on key time points, and predict the temperatures of 6 heat generation points in the next 5 minutes, respectively. The attention mechanism is used to achieve and embed environmental conditions. The attention mechanism implementation includes spatial attention and temporal attention. Spatial attention is used for thermal imager feature enhancement, and temporal attention is used for key time focusing. The environmental condition embedding maps the ambient temperature and humidity to a 4-dimensional vector through the Embedding layer and concatenates it with the LSTM hidden state to enhance the model's adaptability to outdoor scenarios.

[0048] S3: Train and optimize the L model, design the loss function, and formulate the training strategy and perform hardware acceleration. The training strategy includes data augmentation and training parameters. The design of the loss function requires dual-objective optimization and an over-temperature penalty term, while constraining the absolute temperature error and the temperature rise rate error. Coefficient = 0.7, = 0.3, and for samples with predicted temperatures exceeding 85°C, the loss weight is increased by 3 times. Data augmentation is achieved by adding Gaussian noise to simulate sensor errors and randomly occluding local areas of the thermal imager to simulate dust occlusion;

[0049] S4: Edge deployment and lightweighting. Through model compression techniques, optimize the deployment on Jetson Nano. Model compression techniques include knowledge distillation, channel pruning, and quantization-aware training. The implementation methods are respectively using the trained large model to guide the training of the lightweight Student model, removing neuron connections with a contribution of <5% in the LSTM layer, and converting the weights from FP32 to INT8. The Jetson Nano deployment optimization uses the TensorRT engine to convert the model into the TRT format, reducing the inference latency from 50ms to 18ms. At the same time, memory management adopts double-buffer and memory pool technologies to avoid frequent memory allocation;

[0050] S5: In the way of model verification and iteration, conduct offline metric verification. The offline verification metrics include the temperature prediction MAE, the F1 of the temperature rise inflection point detection, and the recall rate of over-temperature warning. And establish an online self-learning mechanism. The online self-learning mechanism includes incremental learning and drift detection. In incremental learning, new data is mixed into the original dataset at a ratio of 5% every week to fine-tune the model. Drift detection is to monitor the KL divergence of the prediction error. If it exceeds the threshold, trigger a full-scale retraining;

[0051] Step 3: Through the constructed feature relationship model and the LSTM neural network prediction model, the operation of the LSTM neural network prediction model includes an input layer and model inference. The input layer includes the historical temperature sequence, the environmental parameter matrix, and the electrical load characteristics. The usage method of model inference is to deploy a lightweight LSTM model on the edge computing unit to predict the temperature change curve and potential hot spot coordinates of each heat generation point within the next five minutes, analyze under different environmental temperatures and power levels, compare the improvement degrees of different levels of heat dissipation power and inverter efficiency, form a risk level by dividing the predicted temperature, and determine the optimal heat dissipation level. According to the determined optimal heat dissipation level, control the hierarchical heat dissipation device to achieve the cooling of the photovoltaic inverter. The risk level division of the predicted temperature includes the first temperature level, the second temperature level, and the third temperature level. The temperatures of the first temperature level, the second temperature level, and the third temperature level are respectively less than sixty degrees Celsius, between sixty and eighty degrees Celsius, and greater than eighty degrees Celsius. The first temperature level is the normal temperature range, only passive cooling is required, which is set as determining the optimal heat dissipation level. The second temperature level requires active intervention of the heat dissipation device, and the third temperature level requires the heat dissipation device to start emergency combined cooling;

[0052] Step 4: Develop a multi-modal heat dissipation strategy. According to the real-time operating status of the photovoltaic inverter and external environmental conditions, adaptively adjust the heat dissipation strategy. The multi-modal heat dissipation strategy includes passive heat dissipation mode, active heat dissipation mode, and composite heat dissipation mode, and each corresponds to a risk level divided by predicted temperature;

[0053] Step 5: Make the usage environment of the photovoltaic inverter self-adaptive and self-maintaining. The self-adaptation and self-maintenance of the usage environment of the photovoltaic inverter are achieved by setting a dust self-cleaning process and anti-condensation control. The dust self-cleaning process includes electrostatic adsorption and reverse pulse cleaning. The operation method of electrostatic adsorption is to energize the high-voltage electrostatic grid at the air inlet for 30 seconds to adsorb dust particles in the air. The operation method of reverse pulse cleaning is to trigger a reverse purge of compressed air pulses for 3 seconds when the differential pressure sensor detects a 20% increase in the filter resistance, and conduct health monitoring and life prediction on key components in the heat dissipation equipment, establish a remote monitoring platform, and realize remote monitoring and operation and maintenance of the photovoltaic inverter and the heat dissipation system;

[0054] Step 6: Record the data on the correlation between heat dissipation and inversion power within a set time, and conduct in-depth analysis to discover potential heat dissipation problems and optimization space.

[0055] Specifically, the data preprocessing in Step 1 includes data synchronization and alignment, noise filtering and outlier processing, and feature engineering. Data synchronization and alignment adopt the timestamp interpolation method, interpolating low-frequency data according to the high-frequency time axis, and realizing multi-channel data synchronous acquisition through FPGA hardware, which helps to solve the problem of different sampling frequencies of different sensors. Noise filtering and outlier processing adopt filtering algorithms, including adaptive Kalman filtering and wavelet transform denoising. For temperature sensor data, the process noise covariance and observation noise covariance are dynamically adjusted. Wavelet transform denoising is used to process the random noise in the thermal imager image and retain the edge features of the temperature field. Feature engineering includes spatial features and temporal features. Spatial features are used to extract the maximum temperature difference from thermal imager data and calculate the entropy of the surface temperature gradient distribution of the radiator. Temporal features are used to calculate the IGBT temperature rise rate through a sliding window and extract the harmonic distortion rate of the load current as a heat generation correlation factor. The formula of the timestamp interpolation method is as follows:

[0056] Among them, refers to the interpolated temperature value, which is used to align low-frequency thermal imager data to the high-frequency time axis to achieve multi-sensor data synchronization. refers to the current interpolation target time point, which is used to represent the system unified clock timestamp. refers to the timestamp of the previous frame of data of the thermal imager. refers to the timestamp of the next frame of data of the thermal imager. refers to the thermal imager at The temperature value collected at a moment, refers to the temperature value collected by the thermal imager at a moment, which is used to calculate the linear change trend between two low-frequency sampling points.

[0057] Specifically, the passive heat dissipation mode uses a graphene composite material to wrap the components, absorb instantaneous heat, and utilize air convection. The air convection can use a micro stepping motor to adjust the inclination angle of the radiator fins to improve the air convection efficiency. The active heat dissipation mode uses the dynamic brushless DC fan speed regulation and air flow path optimization of the heat dissipation device for heat dissipation, which is beneficial to controlling the deflection angle of the deflector of the heat dissipation device according to the CFD simulation results to eliminate local eddies. The dynamic brushless DC fan speed regulation is based on the PID parameter self-tuning of the predicted temperature curve and accelerates in advance according to the predicted demand to prevent temperature overshoot. The formula included in the active heat dissipation mode is as follows:

[0058] Among them, refers to the proportional coefficient, which is used to determine the response intensity to the current temperature deviation. The larger the value, the more radical the fan speed adjustment. refers to the integral coefficient, which is used to eliminate the steady-state error, compensate for the long-term temperature offset, and prevent the temperature from rising cumulatively. refers to the differential coefficient, which is used to suppress temperature fluctuations. By predicting the future temperature change trend, the fan speed is adjusted in advance to avoid overshoot. refers to the temperature deviation. refers to the integral of the temperature deviation, which is used to accumulate the sum of historical temperature deviations and eliminate continuous small deviations. refers to the rate of change of the temperature deviation, which is used to reflect the temperature change trend.

[0059] More specifically, the composite heat dissipation mode uses a semiconductor refrigeration chip and a liquid-cooled microchannel of the heat dissipation device for heat dissipation, and at the same time sets up an over-temperature protection linkage, which is beneficial to triggering the load reduction operation when the temperature still does not drop within the set time. The semiconductor refrigeration chip deploys a TEC array in the predicted hot spot area, and controls the cold end temperature below fifty degrees Celsius through PID control. The liquid-cooled microchannel starts a micro magnetic drive pump to drive the ethylene glycol solution to flow through the microchannel embedded with power devices to perform liquid-cooled heat dissipation on the components.

[0060] A heat dissipation control system for the stable operation of a photovoltaic inverter, including:

[0061] A data acquisition and preprocessing module, which is used to collect multi-source data of the heat dissipation device and components during the use of the photovoltaic inverter, and perform it synchronously, and preprocess the collected data;

[0062] A model construction module, which is used to construct a feature relationship model and an LSTM neural network prediction model based on data collection;

[0063] An analysis module, which is used to analyze and compare the heat dissipation power of different levels and the improvement degree of inverter efficiency under different environmental temperatures and power levels, form a predicted temperature to divide risk levels, and determine the optimal heat dissipation level;

[0064] A strategy formulation module, which is used to formulate a multimodal heat dissipation strategy and adaptively adjust the heat dissipation strategy according to the real-time operating status of the photovoltaic inverter and external environmental conditions;

[0065] A monitoring module, which is used to adaptively adjust and self-maintain the usage environment of the photovoltaic inverter, conduct health monitoring and life prediction on key components in the heat dissipation device, establish a remote monitoring platform, and realize remote monitoring and operation and maintenance of the photovoltaic inverter and the heat dissipation system;

[0066] A data storage and analysis module, which is used to record the correlation data between heat dissipation and inverter power within a set time, and conduct in-depth analysis to discover potential heat dissipation problems and optimization spaces.

[0067] Embodiment 2

[0068] A heat dissipation control method for stable operation of a photovoltaic inverter, including the following specific steps:

[0069] Step 1: Collect multi-source data of the heat dissipation device and components during the use of the photovoltaic inverter. The collected multi-source data includes temperature data, environmental data, and electrical parameters. The multi-source data is collected by deploying high-precision thermocouples at the heat generation points, installing infrared forming cameras outside the heat dissipation device, and installing light intensity sensors and current transformers, and is carried out synchronously, and the collected data is preprocessed;

[0070] Step 2: Based on the data collection, construct a feature relationship model, which is used to reflect the relationship between the heat dissipation requirements and efficiency changes of the inverter under different environmental temperatures and power levels;

[0071] Step 3: Through the constructed feature relationship model, control the hierarchical heat dissipation device to cool down the photovoltaic inverter;

[0072] Step 4: Record the correlation data between heat dissipation and inverter power within a set time, and conduct in-depth analysis to discover potential heat dissipation problems and optimization spaces.

[0073] The difference between this embodiment and the first embodiment is that by constructing a single feature relationship model, the optimal heat dissipation level is determined, and heat dissipation is carried out according to the optimal heat dissipation level. However, in terms of the heat dissipation level, only the optimal can be determined, and hierarchical operations cannot be performed, so the heat dissipation strategy is single.

[0074] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A heat dissipation control method for stable operation of a photovoltaic inverter, characterized in that: The method is applied to a photovoltaic inverter, the photovoltaic inverter includes a heat dissipation device and components, and the method includes the following specific steps: Step 1: Collect multi-source data of heat dissipation equipment and components used in photovoltaic inverters simultaneously, and pre-process the collected data; Step 2: Based on data collection, construct feature relationship model and LSTM neural network prediction model; Step 3: By constructing a feature relationship model and LSTM neural network prediction model, we analyze and compare the different levels of heat dissipation power and inverter efficiency improvement under different ambient temperatures and power levels, form a predicted temperature risk level and determine the optimal heat dissipation level. According to the determined optimal heat dissipation level, we control the graded heat dissipation equipment to achieve photovoltaic inverter cooling; Step 4: Develop a multi-modal heat dissipation strategy and adaptively adjust the heat dissipation strategy according to the real-time operating status of the PV inverter and external environmental conditions; Step 5: Make the use environment of the photovoltaic inverter adaptive and self-maintaining, and conduct health monitoring and life prediction for key components in the heat dissipation equipment, establish a remote monitoring platform, and realize remote monitoring and operation and maintenance of the photovoltaic inverter and heat dissipation system; Step 6: Record the data on the correlation between heat dissipation and inverter power within the set time, and conduct in-depth analysis to identify potential heat dissipation problems and optimization space.

2. A heat dissipation control method for stable operation of a photovoltaic inverter according to claim 1, characterized in that: The multi-source data collected in step one includes temperature data, environmental data and electrical parameters. The multi-source data is collected by deploying high-precision thermocouples at heat points, installing infrared molding cameras outside heat dissipation equipment, and installing light intensity sensors, dust sensors and current transformers.

3. A heat dissipation control method for stable operation of a photovoltaic inverter according to claim 1, characterized in that: The data preprocessing in step 1 includes data synchronization and alignment, noise filtering and outlier processing and feature engineering. The data synchronization and alignment adopts a timestamp interpolation method, interpolating low-frequency data according to a high-frequency time axis, and realizing multi-channel data synchronous acquisition through FPGA hardware. The noise filtering and outlier processing adopt a filtering algorithm, and the filtering algorithm includes adaptive Kalman filtering and wavelet transform denoising. The feature engineering includes spatial features and time series features. The formula of the timestamp interpolation method is as follows: in, refers to the interpolated temperature value, refers to the current interpolation target time point, Refers to the timestamp of the previous frame of thermal imager data. Refers to the timestamp of the next frame of data from the thermal imager. It means that the thermal imager The temperature value collected at all times, It means that the thermal imager The temperature value collected at all times.

4. A heat dissipation control method for stable operation of a photovoltaic inverter according to claim 1, characterized in that: The operation of the LSTM neural network prediction model in step 2 includes an input layer and model reasoning. The input layer includes a historical temperature sequence, an environmental parameter matrix, and electrical load characteristics. The model reasoning is used by deploying a lightweight LSTM model on an edge computing unit to predict the temperature change curve and potential hot spot coordinates of each heating point within the next five minutes. The steps for constructing the LSTM neural network prediction model are as follows: S1: Define and preprocess the input data, input data dimensions, and then use data fusion to align the low-frequency thermal imager data to the high-frequency time axis through time interpolation to form a multi-dimensional tensor with a unified timestamp; S2: LSTM model architecture design, setting network topology, and using attention mechanism implementation and environmental condition embedding; S3: L model training and optimization, loss function design, and formulation of training strategies and hardware acceleration. The training strategies include data enhancement and training parameters. S4: Edge deployment and lightweight, using model compression technology to optimize JetsonNano deployment; S5: Model verification and iteration, offline indicator verification, and establishment of online self-learning mechanism.

5. A heat dissipation control method for stable operation of a photovoltaic inverter according to claim 1, characterized in that: The predicted temperature risk level in step three includes a first temperature level, a second temperature level and a third temperature level. The temperatures of the first temperature level, the second temperature level and the third temperature level are respectively less than sixty degrees Celsius, between sixty and eighty degrees Celsius and greater than eighty degrees Celsius. The first temperature level is a normal temperature range and only requires passive heat dissipation, which is set to determine the optimal heat dissipation level. The second temperature level requires active intervention of the heat dissipation device. The third temperature level requires the heat dissipation device to start emergency composite heat dissipation.

6. A heat dissipation control method for stable operation of a photovoltaic inverter according to claim 1, characterized in that: The multi-modal heat dissipation strategy formulated in the step 4 includes a passive heat dissipation mode, an active heat dissipation mode and a composite heat dissipation mode, and each corresponds to the predicted temperature risk level.

7. A heat dissipation control method for stable operation of a photovoltaic inverter according to claim 6, characterized in that: The passive heat dissipation mode uses graphene composite materials to wrap components to absorb instantaneous heat and utilize air convection. The active heat dissipation mode uses dynamic brushless DC fan speed regulation and airflow path optimization of the heat dissipation device for heat dissipation. The dynamic brushless DC fan speed regulation is based on PID parameter self-tuning of the predicted temperature curve, and accelerates in advance according to the predicted demand to prevent temperature overshoot. The formula included in the active heat dissipation mode is as follows: in, is the proportionality coefficient, is the integration coefficient, is the differential coefficient, is the temperature deviation, is the integral of the temperature deviation, It refers to the rate of change of temperature deviation.

8. A heat dissipation control method for stable operation of a photovoltaic inverter according to claim 6, characterized in that: The composite heat dissipation mode adopts the semiconductor refrigeration plate and liquid cooling microchannel of the heat dissipation equipment to dissipate heat, and sets up over-temperature protection linkage at the same time. The semiconductor refrigeration plate deploys the TEC array in the predicted hot spot area, and controls the cold end temperature to below fifty degrees Celsius through PID. The liquid cooling microchannel starts the micro magnetic drive pump to drive the ethylene glycol solution to flow through the microchannel embedded with the power device to perform liquid cooling on the component.

9. A heat dissipation control method for stable operation of a photovoltaic inverter according to claim 1, characterized in that: The use environment adaptation and self-maintenance of the photovoltaic inverter in step five are achieved by setting a dust self-cleaning process and anti-condensation control. The dust self-cleaning process includes electrostatic adsorption and reverse pulse cleaning. The operation mode of the electrostatic adsorption is to energize the high-voltage electrostatic grid at the air inlet for thirty seconds to adsorb dust particles in the air. The operation mode of the reverse pulse cleaning is to trigger a compressed air pulse reverse purge for three seconds when the pressure difference sensor detects that the filter resistance increases by 20%.

10. A heat dissipation control system for stable operation of a photovoltaic inverter, characterized in that: include: A data acquisition and preprocessing module, which is used to collect multi-source data of heat dissipation equipment and components used in photovoltaic inverters, and to preprocess the collected data synchronously; A model building module, wherein the model building module is used to build a feature relationship model and an LSTM neural network prediction model based on data collection; An analysis module, which is used to analyze and compare different levels of heat dissipation power and inverter efficiency improvement under different ambient temperatures and power levels, form a predicted temperature classification risk level and determine the optimal heat dissipation level; A strategy formulation module, which is used to formulate a multi-modal heat dissipation strategy and adaptively adjust the heat dissipation strategy according to the real-time operating status of the photovoltaic inverter and external environmental conditions; A monitoring module, which is used to make the photovoltaic inverter adapt to the use environment and self-maintain, and to perform health monitoring and life prediction on key components in the heat dissipation equipment, establish a remote monitoring platform, and realize remote monitoring and operation and maintenance of the photovoltaic inverter and the heat dissipation system; The data storage and analysis module is used to record the heat dissipation and inverter power correlation data within a set time, and conduct in-depth analysis to find potential heat dissipation problems and optimization space.

Images