A method and system for light storage direct flexible control for net-zero carbon commercial complex

Abstract

The application discloses a light storage direct flexible control method and system for a net zero carbon commercial complex, which comprises collecting real-time data including human flow, photovoltaic power generation, energy storage state, lighting and elevator operation, predicting future human flow after data cleaning and trend component extraction. Based on the human flow prediction value, peak period is determined, and the energy consumption weight of the system is evaluated combined with real-time data to complete energy demand prediction. With the goal of minimizing photovoltaic curtailment, the optimal photovoltaic output instruction is calculated to form a power generation scheme. According to the energy storage state, the discharge power, execution time and duration of the future period are optimized to develop an energy storage scheduling strategy. According to the energy prediction and lighting data, the light brightness level and switching strategy are optimized in different regions to form a lighting adjustment scheme; combined with the human flow and elevator data, the elevator operation frequency and scheduling strategy are optimized. All schemes are summarized to generate unified collaborative control instructions and execute them. The method realizes efficient use of energy.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic and energy storage control technology, and in particular to a photovoltaic and energy storage direct-flexible control method and system for net-zero carbon commercial complexes. Background Technology

[0002] Currently, against the backdrop of global energy transition and sustainable development, commercial complexes, as major urban energy consumers, face the core challenge of dynamic coordination and precise control in energy management. These buildings integrate various energy-consuming devices such as photovoltaic power generation, energy storage systems, lighting, and elevators, exhibiting complex and interconnected operational characteristics, urgently requiring system-level coordinated response to improve overall energy efficiency. Currently, due to the lack of multi-source data fusion and dynamic analysis of real-time pedestrian flow, equipment status, and environmental parameters, the system struggles to accurately predict short-term energy supply and demand changes, leading to spatiotemporal imbalances between photovoltaic output and electricity load. This hinders the optimal scheduling of energy storage systems, easily resulting in localized power shortages or energy waste during peak periods, thus restricting the achievement of energy conservation and emission reduction goals and operational economics. To solve this problem, it is essential to rely on intelligent control technology and digital energy systems to achieve multi-device coordination and energy supply and demand balance, ultimately supporting the efficient, low-carbon, and stable operation of commercial complexes.

[0003] In one existing technology, an automated control system based on a fixed schedule and preset rules sets independent operating strategies for lighting, elevators, photovoltaic inverters, and energy storage devices: lighting is turned on and off by area and time period; elevators operate at fixed intervals; the photovoltaic system outputs at maximum power point tracking (MPPT); and energy storage devices charge during off-peak hours and discharge during peak hours. The system pre-divides operating modes such as "peak," "off-peak," and "low-peak" based on historical average pedestrian traffic data and switches modes at fixed times each day. There is no real-time data interaction between the subsystems; instructions are only issued by a central controller according to pre-programmed logic. This lack of online perception and dynamic response capabilities regarding instantaneous pedestrian traffic, equipment status, and power generation / consumption load leads to supply-demand imbalances and resource waste due to the reliance on static strategies and the independent operation of each subsystem, coupled with a lack of perception and coordinated response to real-time data changes.

[0004] Therefore, existing technologies cannot achieve energy-efficient utilization. Summary of the Invention

[0005] This invention provides a photovoltaic-storage-DC-flexible control method and system for net-zero carbon commercial complexes to achieve efficient energy utilization.

[0006] In a first aspect, to address the aforementioned technical problems, this invention provides a method for controlling the photovoltaic-storage-DC-flexible system in a net-zero carbon commercial complex, comprising:

[0007] Acquire real-time data including pedestrian flow, photovoltaic power generation, energy storage status, lighting data, and elevator operation data;

[0008] Based on the real-time data, data cleaning, trend component extraction, and future pedestrian flow prediction are performed to obtain the pedestrian flow prediction value.

[0009] Based on the predicted passenger flow, peak periods are determined, and combined with the real-time data, energy demand is predicted to obtain the predicted energy demand value.

[0010] Based on the energy demand forecast, the photovoltaic curtailment is minimized to obtain the optimal photovoltaic output command and generate a photovoltaic power generation scheme.

[0011] When the photovoltaic power generation scheme does not meet the predicted energy demand, the optimal discharge power, execution time and duration for the future period are calculated based on the energy storage status to obtain the energy storage scheduling scheme.

[0012] Based on the energy demand forecast, the photovoltaic power generation scheme, and the lighting data, the optimal brightness level and switching time of the lamps in each area are calculated to obtain a lighting adjustment scheme;

[0013] Based on the energy demand forecast, the passenger flow forecast, and the elevator operation data, the elevator operation frequency and scheduling strategy are optimized to obtain an elevator operation plan.

[0014] Based on the photovoltaic power generation scheme, the energy storage scheduling scheme, the lighting adjustment scheme, and the elevator operation scheme, a unified format of collaborative control instructions is generated and the operation control is executed.

[0015] In one optional implementation, the step of performing data cleaning, trend component extraction, and future pedestrian flow prediction based on the real-time data to obtain a pedestrian flow prediction value includes:

[0016] Based on the real-time data, data denoising is performed using db4 wavelet transform to obtain denoised data;

[0017] Based on the noise reduction data, the STL decomposition method is used to extract trends in pedestrian flow, photovoltaic power generation, lighting brightness, and elevator frequency to obtain trend data.

[0018] Based on the trend data, time series prediction modeling is performed using a long short-term memory network, and after iterative training, the predicted pedestrian flow value for future periods is output.

[0019] In one optional implementation, the step of determining peak periods based on the predicted pedestrian flow and combining it with the real-time data to predict energy demand, thereby obtaining a predicted energy demand value, includes:

[0020] When the predicted pedestrian flow is greater than the preset pedestrian flow threshold, it is determined to be a peak period. Combined with real-time data, the operating characteristics are extracted by the sliding window averaging method and the system energy consumption weight of the energy-consuming system is calculated by the analytic hierarchy process.

[0021] Based on the operational characteristics and the system energy consumption weights, the energy demand forecast is output by fusing the characteristics and performing regression prediction through a pre-established multilayer perceptron model.

[0022] In one optional implementation, the step of calculating the minimum photovoltaic curtailment based on the predicted energy demand to obtain the optimal photovoltaic output command and generate a photovoltaic power generation scheme includes:

[0023] Based on the energy demand forecast and the real-time data, with the objective function of minimizing the amount of curtailed solar power, a linear programming model is solved using the simplex method to output photovoltaic power output commands for each time period.

[0024] Acquire sunlight data, calculate the optimal tilt angle of photovoltaic modules using a solar position algorithm, and output the tilt angle adjustment parameters of photovoltaic modules;

[0025] The photovoltaic output command and the photovoltaic module tilt angle adjustment parameters are integrated to generate a photovoltaic power generation scheme.

[0026] In one optional implementation, when the photovoltaic power generation scheme does not meet the predicted energy demand, calculating the optimal discharge power, execution time, and duration for future periods based on the energy storage status to obtain an energy storage dispatch scheme includes:

[0027] The photovoltaic power generation capacity is extracted from the photovoltaic power generation scheme, and the difference between the predicted energy demand and the photovoltaic power generation capacity is calculated to obtain the energy gap.

[0028] Based on the energy shortage and the energy storage status, with the objective function of minimizing the amount of electricity purchased by the grid, a linear programming model is solved using the simplex method to output the optimal discharge power.

[0029] Historical load curves are obtained, and combined with the optimal discharge power, time period optimization allocation is performed through dynamic programming algorithm to output the start and stop time sequence of discharge periods;

[0030] Based on the optimal discharge power and the start-stop time sequence, an execution plan is generated through a time-slice round-robin scheduling algorithm, and an energy storage scheduling scheme containing discharge power, execution time, and duration is output.

[0031] In one optional implementation, the step of calculating the optimal brightness level and switching time of the luminaires in each area based on the predicted energy demand, the photovoltaic power generation scheme, and the lighting data to obtain a lighting adjustment scheme includes:

[0032] Ambient light data is acquired, and combined with the energy demand forecast, the photovoltaic power generation scheme, and the lighting data, the lighting brightness is graded using the Viterbi algorithm, and the optimal brightness level sequence and on / off time combination is output.

[0033] Based on the optimal brightness level sequence, a fuzzy PID controller is used to adjust the light intensity and output an anti-disturbance brightness correction value and a switching time offset.

[0034] Lighting energy consumption is extracted from the lighting data, and energy consumption deviation is compensated through a model predictive control algorithm to output a brightness compensation coefficient.

[0035] By integrating the optimal brightness level sequence, brightness correction value, brightness compensation coefficient, switching time combination, and switching time offset, a lighting adjustment scheme is generated.

[0036] In one optional implementation, the step of optimizing the elevator operating frequency and scheduling strategy based on the predicted energy demand, the predicted passenger flow, and the elevator operating data to obtain an elevator operation plan includes:

[0037] Based on the energy demand forecast, the passenger flow forecast, and the elevator operation data, the elevator scheduling decision is modeled using the Q-Learning algorithm, and the elevator operation frequency and dispatch priority sequence are output.

[0038] Based on the elevator operating frequency and dispatch priority sequence, the time window is optimized by Monte Carlo tree search to obtain the optimal operating time window;

[0039] The elevator load rate is extracted from the elevator operation data. When the elevator load rate exceeds the preset load threshold, the emergency floors are re-allocated using a dynamic programming algorithm to obtain the emergency dispatch floors and frequency compensation coefficients.

[0040] The elevator operation frequency, the elevator dispatch priority sequence, the emergency dispatch floor, and the frequency compensation coefficient are integrated to generate an elevator operation plan.

[0041] Secondly, the present invention provides a photovoltaic-storage-DC-flexible control system for net-zero carbon commercial complexes, comprising:

[0042] The data acquisition module is used to acquire real-time data including pedestrian flow, photovoltaic power generation, energy storage status, lighting data, and elevator operation data;

[0043] The pedestrian flow prediction module is used to perform data cleaning, trend component extraction, and future pedestrian flow prediction based on the real-time data to obtain the pedestrian flow prediction value.

[0044] The energy demand forecasting module is used to determine peak periods based on the predicted pedestrian flow and, in conjunction with the real-time data, to forecast energy demand and obtain the predicted energy demand value.

[0045] The photovoltaic power generation module is used to perform calculations to minimize photovoltaic curtailment based on the energy demand forecast, obtain the optimal photovoltaic output command, and generate a photovoltaic power generation scheme.

[0046] The energy storage scheduling module is used to calculate the optimal discharge power, execution time and duration for future periods based on the energy storage status when the photovoltaic power generation scheme does not meet the predicted energy demand;

[0047] The lighting adjustment module is used to calculate the optimal brightness level and switching time of the lamps in each area based on the energy demand forecast, the photovoltaic power generation scheme and the lighting data, so as to obtain the lighting adjustment scheme;

[0048] The elevator operation module is used to optimize the elevator operation frequency and scheduling strategy based on the energy demand forecast, the passenger flow forecast and the elevator operation data, so as to obtain an elevator operation plan.

[0049] The output control module is used to integrate and generate unified format collaborative control instructions and execute operation control based on the photovoltaic power generation scheme, the energy storage scheduling scheme, the lighting adjustment scheme and the elevator operation scheme.

[0050] Thirdly, the present invention also provides an electronic device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the photovoltaic-storage-DC-flexible control method for net-zero carbon commercial complexes as described in any one of the above.

[0051] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the photovoltaic-storage-DC-flexible control method for net-zero carbon commercial complexes as described above.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] (1) By integrating real-time data from multiple sources, wavelet transform and STL decomposition are used for data cleaning and trend extraction, and long short-term memory network is combined to perform high-precision traffic flow prediction, thereby improving the accuracy and timeliness of energy demand prediction and overcoming the shortcomings of traditional static strategies in responding to dynamic changes.

[0054] (2) Based on the energy demand forecast results, with the goal of minimizing photovoltaic curtailment, the photovoltaic output and module tilt angle are optimized by linear programming and solar position algorithm. At the same time, the intelligent scheduling discharge strategy is combined with energy storage status and dynamic programming algorithm to achieve the coordinated operation of photovoltaic and energy storage and the precise matching of energy supply and demand, thereby improving the utilization rate of renewable energy.

[0055] (3) Integrating lighting environment data and energy prediction information, the Viterbi algorithm and fuzzy PID controller are used to realize lighting brightness classification and anti-disturbance adjustment, and energy consumption compensation is carried out by relying on model predictive control. Furthermore, the scheduling strategy is optimized by combining elevator operation data and reinforcement learning algorithm to realize multi-device collaborative response and refined energy efficiency management.

[0056] (4) From data perception, prediction optimization to multi-system collaborative control, a closed-loop control system is formed. While improving the photovoltaic absorption rate and reducing the power grid purchase cost, it ensures the intelligent and flexible operation of load equipment such as lighting and elevators in commercial complexes, improves energy utilization efficiency and system operation economy, and supports the realization of net zero carbon goals. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the photovoltaic-storage-direct-flexible control method for net-zero carbon commercial complexes provided in the first embodiment of the present invention;

[0058] Figure 2 This is a schematic diagram of the photovoltaic-storage-DC-flexible control system for a net-zero carbon commercial complex provided in the second embodiment of the present invention. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] Reference Figure 1 The first embodiment of the present invention provides a method for controlling the photovoltaic, energy storage, direct current, and flexible energy transmission in a net-zero carbon commercial complex, comprising the following steps:

[0061] S11, acquire real-time data including pedestrian flow, photovoltaic power generation, energy storage status, lighting data, and elevator operation data;

[0062] S12, perform data cleaning, trend component extraction and future pedestrian flow prediction based on the real-time data to obtain the pedestrian flow prediction value;

[0063] S13, Based on the predicted passenger flow, determine the peak period and combine it with the real-time data to predict energy demand and obtain the predicted energy demand value.

[0064] S14. Based on the energy demand forecast, perform a calculation to minimize the amount of photovoltaic power curtailment, obtain the optimal photovoltaic output command, and generate a photovoltaic power generation scheme.

[0065] S15, when the photovoltaic power generation scheme does not meet the predicted energy demand, calculate the optimal discharge power, execution time and duration for the future period based on the energy storage status to obtain the energy storage scheduling scheme;

[0066] S16. Based on the energy demand forecast, the photovoltaic power generation scheme, and the lighting data, calculate the optimal brightness level and switching time of the lamps in each area to obtain a lighting adjustment scheme.

[0067] S17. Based on the energy demand forecast, the passenger flow forecast and the elevator operation data, optimize the elevator operation frequency and scheduling strategy to obtain the elevator operation plan.

[0068] S18. Based on the photovoltaic power generation scheme, the energy storage scheduling scheme, the lighting adjustment scheme, and the elevator operation scheme, a unified format collaborative control instruction is generated and the operation control is executed.

[0069] In step S11, real-time data including pedestrian flow, photovoltaic power generation, energy storage status, lighting data, and elevator operation data are acquired.

[0070] Specifically, pedestrian flow data is collected by visual sensors and infrared counters installed at entrances and main passages, including instantaneous pedestrian flow, cumulative pedestrian flow and its temporal distribution; photovoltaic power generation data comes from the output monitoring module of the photovoltaic inverter, including real-time power generation, daily cumulative power generation and module temperature; energy storage status data is provided by the battery management system, including battery state of charge, health status, current charging and discharging power and remaining available capacity; lighting data is obtained through the intelligent lighting controller, including the on / off status of lights in each area, real-time brightness level, energy consumption current and running time; elevator operation data is collected by sensors built into the elevator control cabinet, including elevator start / stop status, current floor, direction of travel, passenger capacity and energy consumption.

[0071] The real-time data acquired in this step provides the raw input for subsequent data cleaning, trend extraction, and multi-module collaborative control, forming the foundation for the dynamic response and precise regulation of the entire photovoltaic-storage-direct-drive-flexible control system. The comprehensiveness and real-time nature of data acquisition directly affect the accuracy and timeliness of population flow forecasting, energy demand assessment, and equipment scheduling strategies, thus determining whether the system can achieve efficient energy utilization under the net-zero carbon target in complex operating environments.

[0072] In step S12, data cleaning, trend component extraction, and future pedestrian flow prediction are performed based on the real-time data to obtain the pedestrian flow prediction value.

[0073] In one specific implementation, the step of performing data cleaning, trend component extraction, and future pedestrian flow prediction based on the real-time data to obtain a pedestrian flow prediction value includes:

[0074] Based on the real-time data, data denoising is performed using db4 wavelet transform to obtain denoised data;

[0075] Based on the noise reduction data, the STL decomposition method is used to extract trends in pedestrian flow, photovoltaic power generation, lighting brightness, and elevator frequency to obtain trend data.

[0076] Based on the trend data, time series prediction modeling is performed using a long short-term memory network, and after iterative training, the predicted pedestrian flow value for future periods is output.

[0077] Specifically, the data denoising process employs the db4 wavelet transform algorithm, which converts the time-domain signal to the frequency domain for processing through mathematical transformation. The process is as follows: First, the input signal is convolved with a pre-determined low-pass decomposition filter. The weighted average of the signal at each point is calculated. The weights are determined by the coefficients of the low-pass decomposition filter, which are fixed values ​​pre-calculated based on the db4 wavelet basis function, representing the weighted proportion of the contribution of different positions to the current point. During convolution, the filter coefficients, as weights, are multiplied by the corresponding values ​​in the signal and summed to obtain the weighted average of that point, thus obtaining the low-frequency component. Simultaneously, convolution is performed with a high-pass decomposition filter to capture subtle changes in the signal, obtaining the high-frequency component. The convolution result is downsampled, retaining half of the data points. The above process is repeated for the obtained low-frequency component to achieve multi-level decomposition. Subsequently, the noise threshold is calculated based on the statistical distribution characteristics of the high-frequency coefficients in the finest layer. Specifically, the median absolute value of all high-frequency coefficients in this layer is first calculated and used as a benchmark for estimating the noise level. Then, the final threshold is calculated based on the proportional relationship between robust statistics and this benchmark. Soft thresholding is applied to all high-frequency coefficients, setting coefficients with absolute values ​​less than the noise threshold to zero and shrinking coefficients greater than the threshold according to a certain rule: subtracting the noise threshold from the absolute value of the coefficient while maintaining its original sign. This results in a new, shrunken coefficient value. Finally, the signal is reconstructed through a reverse process. The processed coefficients are first upsampled and then convolved with a reconstruction filter. This reconstruction filter, paired with the decomposition filter, uses a set of fixed coefficients predetermined by the selected db4 wavelet basis function. Its function is to re-merge the coefficients processed in the sub-bands to accurately reconstruct the original, noise-removed signal. This process is repeated step-by-step to obtain the final denoised signal.

[0078] Trend component extraction employs a seasonal trend decomposition method, which iteratively calculates and separates the various components of a time series. First, the sliding window size is determined through periodic analysis of historical data to ensure coverage of major cyclical patterns. In the inner loop, locally weighted regression is used to smooth the data. This method typically employs quadratic or cubic polynomials for local fitting, calculating a weighted fit value for each point in the time series. The weights are assigned according to a bell-shaped function, which assigns higher weights to adjacent points and decreasing weights to points further away. The width of the smoothing window is pre-set based on the cyclical characteristics of the time series to ensure coverage of major trend change patterns. A least-squares polynomial function is used to fit the function, with data points closer to the current point receiving higher weights. The trend component is subtracted from the original data to obtain the detrended series. This series is then folded according to the period length, and locally weighted regression is performed again for the data at each phase point to obtain the seasonal component. This process is repeated until convergence. In the outer loop, robustness weights are calculated based on the residual size to reduce the impact of outliers. The entire decomposition process is repeated until stability is achieved.

[0079] Multivariate time series prediction is based on a Long Short-Term Memory (LSTM) network model, a deep recurrent neural network specifically designed for processing sequential data. The model input is a preprocessed multivariate time series, with each time step containing data across multiple dimensions, including pedestrian flow, photovoltaic power, lighting, and elevators. The network core consists of multiple stacked LTM layers, each containing an input gate, a forget gate, and an output gate to learn long-term dependencies in the data. Finally, a fully connected output layer yields the predicted pedestrian flow for future time periods. The model is trained using supervised learning, employing a large amount of historically collected real-time data as the training set and historical pedestrian flow data as the target label. During training, gradients are calculated using the backpropagation algorithm, and an adaptive moment estimation optimizer is used to adjust network parameters to minimize the mean squared error loss between predicted and true values. Multiple iterations of training are performed until the model converges, achieving accurate prediction capabilities. This model processes sequential data through a special gating mechanism. The input gate controls the inflow of new information, using the sigmoid function to determine which information needs updating and the tanh function to generate candidate values. The forget gate determines which historical information needs to be retained or discarded. The output gate controls the output of the current state. During model training, gradients are calculated using backpropagation, and parameters are optimized using gradient descent to minimize prediction error. The loss function uses mean squared error to measure the difference between predicted and true values. The optimizer adjusts the parameter update step size using an adaptive learning rate to ensure the stability of the training process.

[0080] The data denoising process eliminates measurement errors and random interference, the trend extraction process captures the essential patterns of data change, and the prediction model establishes accurate mapping relationships based on historical data. The final output of the predicted pedestrian flow serves as the basis for subsequent energy scheduling and equipment control decisions, and its accuracy directly impacts the overall system's operational efficiency. This data processing method based on mathematical principles provides a reliable technical guarantee for energy optimization in commercial complexes.

[0081] In step S13, peak periods are determined based on the predicted pedestrian flow, and energy demand is predicted in conjunction with the real-time data to obtain the predicted energy demand value.

[0082] In one specific implementation, the step of determining peak periods based on the predicted pedestrian flow and combining it with the real-time data to predict energy demand, thereby obtaining a predicted energy demand value, includes:

[0083] When the predicted pedestrian flow is greater than the preset pedestrian flow threshold, it is determined to be a peak period. Combined with real-time data, the operating characteristics are extracted by the sliding window averaging method and the system energy consumption weight of the energy-consuming system is calculated by the analytic hierarchy process.

[0084] Based on the operational characteristics and the system energy consumption weights, the energy demand forecast is output by fusing the characteristics and performing regression prediction through a pre-established multilayer perceptron model.

[0085] Specifically, a statistical discrimination method is used in the peak period determination phase. The system compares the predicted pedestrian flow with a threshold determined through historical data analysis. This threshold is obtained by calculating the percentile of historical pedestrian flow data, specifically using the 85th percentile as the discrimination standard. When the predicted pedestrian flow exceeds this threshold, the system determines that it has entered peak period operation. This determination process can be expressed as: if the predicted pedestrian flow is greater than the 85th percentile of historical pedestrian flow, it is marked as a peak period; otherwise, it is an off-peak period.

[0086] Based on peak period determination, the system initiates operational feature extraction and energy consumption weight calculation. Operational feature extraction employs a sliding window averaging method, with the window size determined based on the dynamic response characteristics of the system's main equipment, taking integer multiples of the load variation cycle. Within the window, the system calculates the statistical characteristics of each operational parameter (including photovoltaic power generation, energy storage state of charge, lighting power, and elevator frequency), including the arithmetic mean, variance, and rate of change. These statistics characterize the system's average operational level, fluctuation characteristics, and changing trends, collectively forming the operational feature vector.

[0087] The energy consumption weighting calculation employs the Analytic Hierarchy Process (AHP), establishing a hierarchical model comprising an objective layer, a criterion layer, and a scheme layer. This model aims to decompose the complex energy consumption weighting decision-making problem into clear structural layers. The objective layer defines the overall goal of the weighting calculation, the criterion layer contains characteristic indicators affecting energy consumption, and the scheme layer corresponds to each specific energy-consuming system. This structure provides a logical framework and comparative basis for subsequent pairwise comparisons between systems, construction of the judgment matrix, and weight calculation. First, the judgment matrix is ​​constructed, and the relative importance of each criterion is determined through pairwise comparisons, using a 1-9 scale for quantitative evaluation. Then, the largest eigenvalue and corresponding eigenvector of the judgment matrix are calculated, and the eigenvector is normalized to obtain the weight vector. Finally, a consistency check is performed to ensure that the logical consistency of the judgment matrix meets the requirements. Through this series of calculations, the accurate weight values ​​for each energy-consuming system are obtained.

[0088] The extracted operational feature vector and system energy consumption weight values ​​are used as inputs. A pre-trained multilayer perceptron model is used for feature fusion and regression prediction. The training process of this model is supervised learning, using a large number of historical operational feature vectors and corresponding actual system energy consumption values ​​as training datasets. During training, the model calculates predicted values ​​through forward propagation and uses the backpropagation algorithm to iteratively adjust the internal connection weight parameters based on the error between the predicted and true values. An adaptive optimizer minimizes the mean squared error loss function, ultimately enabling the model to learn the complex mapping relationship between input features and energy demand. The model consists of one input layer, three hidden layers, and one output layer. The input layer receives the concatenation result of the operational feature vector and the energy consumption weight vector. The hidden layers use the ReLU activation function, and the output of each layer is calculated through forward propagation. The model updates the weight parameters through the backpropagation algorithm, uses the mean squared error function as the loss function, and employs an adaptive moment estimation algorithm as the optimizer. After sufficient training, the model can accurately predict the energy demand value for future periods. The predicted energy demand reflects the comprehensive energy consumption demand of various energy-consuming devices under specific population flow conditions.

[0089] Peak period determination is based on statistical theory, feature extraction employs digital signal processing methods, weight calculation utilizes decision analysis mathematical tools, and the final prediction is based on a neural network model. This multi-method fusion mathematical framework can accurately capture system operating characteristics and provide a reliable predictive basis for energy dispatching.

[0090] In step S14, based on the predicted energy demand, the photovoltaic curtailment is minimized to obtain the optimal photovoltaic output command and generate a photovoltaic power generation scheme.

[0091] In one specific implementation, the step of performing a calculation to minimize photovoltaic curtailment based on the predicted energy demand, obtaining the optimal photovoltaic output command, and generating a photovoltaic power generation scheme includes:

[0092] Based on the energy demand forecast and the real-time data, with the objective function of minimizing the amount of curtailed solar power, a linear programming model is solved using the simplex method to output photovoltaic power output commands for each time period.

[0093] Acquire sunlight data, calculate the optimal tilt angle of photovoltaic modules using a solar position algorithm, and output the tilt angle adjustment parameters of photovoltaic modules;

[0094] The photovoltaic output command and the photovoltaic module tilt angle adjustment parameters are integrated to generate a photovoltaic power generation scheme.

[0095] Specifically, a linear programming model for minimizing photovoltaic (PV) curtailment is first established. Using a 24-hour optimization period, the PV output command for each time period is used as the decision variable. The objective function is set to minimize the total curtailment, which is the sum of the differences between the maximum potential PV power generation and the actual output command. Constraints include: the output command for each time period must not exceed the maximum potential PV power generation for that time period; the output command must meet a certain proportion of the predicted energy demand for that time period (this proportion is determined based on historical operating data); and the rate of change of the output command between adjacent time periods must not exceed the maximum ramp-up rate of the PV inverter. The linear programming problem is solved using the simplex method. Starting from an initial feasible solution, the algorithm iteratively moves between vertices of the feasible region, selecting the direction that decreases the objective function value in each iteration. Finally, the optimal solution that minimizes curtailment is found, and the PV output command for each time period is output.

[0096] Simultaneously, the system calculates the optimal tilt angle of the photovoltaic modules using a solar position algorithm. Based on the latitude and longitude coordinates of the installation site and the current date and time, this algorithm first calculates the precise position of the sun in the sky, including the solar altitude angle and azimuth angle. The calculation process is based on the celestial coordinate system and the laws of Earth's rotation and revolution, determining its angle relative to the Earth's horizontal plane using astronomical parameters such as the solar declination angle and hour angle. Subsequently, the algorithm establishes a geometric relationship model between the photovoltaic module's receiving surface and the incident sunlight, aiming to maximize the total amount of direct solar radiation received by the module surface during the calculation period. Through iterative optimization calculations, it solves for the corresponding optimal tilt angle value and outputs the photovoltaic module tilt angle adjustment parameters.

[0097] Finally, the photovoltaic output commands for each time period are integrated with the photovoltaic module tilt angle adjustment parameters. The output commands are arranged in chronological order to form an output command sequence, and the tilt angle parameters are converted into motor control signals. The combination of the two constitutes a complete photovoltaic power generation scheme, which includes both output planning in the time dimension and module attitude optimization in the spatial dimension.

[0098] Mathematical optimization methods ensured an optimal match between photovoltaic (PV) power generation and electricity demand, maximizing PV capacity while effectively preventing curtailment. Precise tilt angle adjustment further improved PV power generation efficiency, providing a reliable clean energy supply solution for the low-carbon operation of commercial complexes.

[0099] In step S15, when the photovoltaic power generation scheme does not meet the predicted energy demand, the optimal discharge power, execution time and duration for the future period are calculated based on the energy storage status to obtain the energy storage scheduling scheme.

[0100] In one specific implementation, when the photovoltaic power generation scheme does not meet the predicted energy demand, calculating the optimal discharge power, execution time, and duration for the future period based on the energy storage status to obtain an energy storage dispatch scheme includes:

[0101] The photovoltaic power generation capacity is extracted from the photovoltaic power generation scheme, and the difference between the predicted energy demand and the photovoltaic power generation capacity is calculated to obtain the energy gap.

[0102] Based on the energy shortage and the energy storage status, with the objective function of minimizing the amount of electricity purchased by the grid, a linear programming model is solved using the simplex method to output the optimal discharge power.

[0103] Historical load curves are obtained, and combined with the optimal discharge power, time period optimization allocation is performed through dynamic programming algorithm to output the start and stop time sequence of discharge periods;

[0104] Based on the optimal discharge power and the start-stop time sequence, an execution plan is generated through a time-slice round-robin scheduling algorithm, and an energy storage scheduling scheme containing discharge power, execution time, and duration is output.

[0105] Specifically, the energy gap is calculated first. The system extracts the photovoltaic power generation sequence for each time period from the photovoltaic power generation scheme, and performs element-by-element subtraction with the corresponding energy demand forecast sequence to obtain the energy gap sequence for each time period. A positive gap indicates that there is a power shortage during that period, requiring the energy storage system to supplement the power supply.

[0106] Subsequently, a linear programming model is established with the objective of minimizing the amount of electricity purchased from the grid. The decision variable is the discharge power of the energy storage system in each time period, and the objective function is to minimize the total amount of electricity purchased from the grid, which is the sum of the differences between the energy shortage and the energy storage discharge power in each time period. The constraints include: the discharge power in each time period does not exceed the maximum allowable discharge power of the energy storage system, the total discharge does not exceed the current available capacity of the energy storage system, and the rate of change of discharge power does not exceed the maximum ramp rate of the energy storage system. The optimization problem is solved using the simplex method. Starting from the initial basic feasible solution, through basis transformation and iterative calculation, the vertex solution that optimizes the objective function is found, and the optimal discharge power sequence for each time period is output.

[0107] Next, a dynamic programming algorithm is used for time-segment optimization allocation. A day is divided into several equal-length time segments. The state is defined as the time segment index and the cumulative discharge amount. To construct a finite state space, the continuous variable of cumulative discharge amount is discretized into multiple pre-defined power levels. The number and intervals of these levels are predetermined based on the total capacity of the energy storage system and the scheduling accuracy requirements. The minimum grid purchase amount represented by the state value is calculated recursively: starting from the last time segment, the solution proceeds backwards. The value of each state is determined by the immediate purchase amount generated by the decision (discharge power) in the current time segment, plus the optimal value in the new state after the transition in subsequent time segments. By traversing and comparing all possible decisions, the path that minimizes the total purchase amount is selected, and the state value represents the minimum grid purchase amount in that state. Starting from the initial state, the optimal value for each state is calculated step by step according to the state transition equation. Specifically, when the system is in a certain time period index and its corresponding cumulative discharge state, if a specific discharge power decision is made, the electricity released by that decision will be added to the current cumulative discharge, thus forming a new cumulative discharge state corresponding to the next time period index. The system state then transitions from the current time period and discharge state to the new discharge state of the next time period, recording the optimal decision path. By backtracking, the globally optimal discharge time period allocation scheme is found, and the start and stop time sequence of the discharge time periods is output. This sequence clearly identifies the specific time period range within which the energy storage system needs to discharge.

[0108] Finally, a time-slice round-robin scheduling algorithm is used to generate the execution plan. Taking the start and stop time sequence of the discharge period as input, each discharge period is divided into several time slices of equal length. The algorithm's application here aims to achieve fine-grained allocation and smooth transition of power commands. When allocating discharge power commands to each time slice, it relies on the optimal discharge power sequence obtained from upper-level optimization calculations and adheres to the constraint that the power change rate does not exceed the maximum allowable ramp rate of the energy storage system. Specifically, a first-order low-pass digital filter processes the original power command sequence. The cutoff frequency of this filter is pre-set according to the actual physical response characteristics of the energy storage system, thereby filtering out high-frequency abrupt changes in the commands and generating a smoothly changing power command execution plan, ultimately ensuring that the energy storage device can respond smoothly. Based on the optimal discharge power sequence, corresponding discharge power commands are allocated to each time slice, while considering the response characteristics of the energy storage system to ensure a smooth transition of power commands. The final output is a detailed energy storage scheduling scheme containing the discharge power value, execution time, and duration of each time slice.

[0109] This solution effectively compensates for the shortcomings of photovoltaic power generation by ensuring that the energy storage system discharges at the optimal power level during the most needed periods, while minimizing the purchase of electricity from the grid, thus achieving efficient energy utilization.

[0110] In step S16, based on the predicted energy demand, the photovoltaic power generation scheme, and the lighting data, the optimal brightness level and switching time of the lamps in each area are calculated to obtain a lighting adjustment scheme.

[0111] In one specific implementation, the step of calculating the optimal brightness level and switching time of lamps in each area based on the predicted energy demand, the photovoltaic power generation scheme, and the lighting data to obtain a lighting adjustment scheme includes:

[0112] Ambient light data is acquired, and combined with the energy demand forecast, the photovoltaic power generation scheme, and the lighting data, the lighting brightness is graded using the Viterbi algorithm, and the optimal brightness level sequence and on / off time combination is output.

[0113] Based on the optimal brightness level sequence, a fuzzy PID controller is used to adjust the light intensity and output an anti-disturbance brightness correction value and a switching time offset.

[0114] Lighting energy consumption is extracted from the lighting data, and energy consumption deviation is compensated through a model predictive control algorithm to output a brightness compensation coefficient.

[0115] By integrating the optimal brightness level sequence, brightness correction value, brightness compensation coefficient, switching time combination, and switching time offset, a lighting adjustment scheme is generated.

[0116] Specifically, the system first acquires real-time ambient illuminance data collected by an ambient light sensor, which represents the ambient brightness level in lux. Combined with energy supply and demand information provided by energy demand forecasts, power generation capacity data for each time period in the photovoltaic power generation scheme, and the current brightness level and operating status of luminaires in each area contained in the lighting data, a basic dataset for lighting control is constructed.

[0117] The Viterbi algorithm is used to calculate lighting brightness levels. This algorithm models the lighting control process as a Hidden Markov Model (HMM), where the hidden states are the possible brightness levels for each area, and the observation sequence consists of ambient illuminance, energy supply-demand ratio, and pedestrian traffic. The algorithm first calculates the initial state probability, i.e., the initial distribution probability of each brightness level. Then, it calculates the state transition probability, representing the likelihood of transitions between brightness levels. Next, it calculates the emission probability, representing the probability of generating a corresponding observation at a specific brightness level. The initial state probability is calculated by statistically analyzing the frequency of each brightness level occurring at the initial moment in historical data; the state transition probability is determined by statistically analyzing the frequency of transitions from one brightness level to another in historical data and calculating the transition frequency; the emission probability is estimated by statistically analyzing the distribution of various observations (such as ambient illuminance and pedestrian traffic) at a specific brightness level. The maximum probability path for each state at each moment is calculated using dynamic programming, and the path backtracking pointer is recorded. Finally, by backtracking, the algorithm finds the state sequence with the highest probability and outputs the optimal brightness level sequence for each area in future time periods, along with the corresponding on / off time combinations.

[0118] A fuzzy PID controller is used for light intensity adjustment calculation. This controller uses the optimal brightness level sequence as the setpoint and the real-time ambient illuminance as the process variable. First, the brightness deviation value is calculated, which is the difference between the setpoint and the process variable. Then, the deviation change rate is calculated, which is the difference between the current deviation and the deviation at the previous moment. The precise deviation value and deviation change rate are converted into fuzzy quantities using a membership function. This membership function adopts a trigonometric function form, and its parameters are pre-set based on the actual distribution range of brightness deviation and change rate in historical operating data. The precise input value is divided into multiple fuzzy linguistic variable levels: "negative large," "negative small," "zero," "positive small," and "positive large." The rule base upon which the fuzzy inference system is based consists of multiple empirical rules in the form of "if-then," such as "if the brightness deviation is positive and the deviation change rate is positive, then the output adjustment amount is positive." The rule base covers all possible combinations of fuzzy variable levels to ensure the completeness of the control. The calculation of the switching time offset involves analyzing recent trends in ambient illuminance, using a sliding window averaging method to determine its rate of change, and then linearly compensating for a preset baseline switching time based on this rate. This achieves dynamic synchronization between the lighting strategy and changes in natural light, and the result is input into the fuzzy inference system. The fuzzy inference system performs inference based on a predefined fuzzy rule base, with rules stating that "if the deviation is large and the rate of change of the deviation is large, then the output adjustment amount is large." The centroid method is used to defuzzify the data, converting the fuzzy output into a precise brightness correction value. Simultaneously, the switching time offset is calculated based on the ambient light change trend to ensure that lighting adjustments are synchronized with environmental changes.

[0119] Historical lighting energy consumption sequences are extracted from lighting data, and energy consumption deviation compensation is performed using a model predictive control algorithm. A state-space model of the lighting system is established, which is a simplified discrete linear dynamic model used to characterize the dynamic response relationship of the lighting system's energy consumption to changes in brightness level. The model's state variables reflect the internal operating status of the system, the control input is the brightness level command of the lamps, and the output is the corresponding real-time energy consumption. Key parameters in the model are obtained through system identification methods, namely, by collecting a large amount of brightness level adjustment sequences and corresponding energy consumption response data from historical operation, and using mathematical methods such as least squares to estimate and fit the model parameters, thereby establishing a mathematical model that can accurately predict the dynamic changes in lighting energy consumption, with brightness level as the control input and energy consumption as the output. In each control cycle, the energy consumption trajectory for future periods is predicted based on the current system state. An optimization problem is constructed, with the objective function being to minimize the deviation between predicted energy consumption and expected energy consumption while satisfying lighting comfort constraints. By solving this optimization problem, a brightness compensation coefficient sequence is obtained. This coefficient is used to adjust the brightness command to ensure that the actual energy consumption meets the energy allocation requirements.

[0120] Finally, the processing results are integrated, and the optimal brightness level sequence is added element-wise with the brightness correction value to obtain the adjusted brightness command. This adjusted command is then multiplied by the brightness compensation coefficient to complete the energy consumption calibration. Combining the switching time combination and the switching time offset, a lighting adjustment scheme is generated, containing parameters such as the specific brightness level values ​​of the luminaires in each area, the precise switching time points, and the duration. This scheme ensures optimal energy consumption configuration while meeting lighting requirements through rigorous mathematical calculations.

[0121] In step S17, based on the predicted energy demand, the predicted passenger flow, and the elevator operation data, the elevator operation frequency and scheduling strategy are optimized to obtain an elevator operation plan.

[0122] In one specific implementation, the step of optimizing the elevator operating frequency and scheduling strategy based on the predicted energy demand, the predicted passenger flow, and the elevator operating data to obtain an elevator operation plan includes:

[0123] Based on the energy demand forecast, the passenger flow forecast, and the elevator operation data, the elevator scheduling decision is modeled using the Q-Learning algorithm, and the elevator operation frequency and dispatch priority sequence are output.

[0124] Based on the elevator operating frequency and dispatch priority sequence, the time window is optimized by Monte Carlo tree search to obtain the optimal operating time window;

[0125] The elevator load rate is extracted from the elevator operation data. When the elevator load rate exceeds the preset load threshold, the emergency floors are re-allocated using a dynamic programming algorithm to obtain the emergency dispatch floors and frequency compensation coefficients.

[0126] The elevator operation frequency, the elevator dispatch priority sequence, the emergency dispatch floor, and the frequency compensation coefficient are integrated to generate an elevator operation plan.

[0127] Specifically, a Q-Learning algorithm is first used to establish an elevator scheduling decision model. The elevator system is modeled as a Markov decision process, where the state space consists of dimensions such as the number of people waiting on each floor, the current position of the elevator, the direction of travel, the number of people in the car, and the energy demand level. The action space includes operational commands such as elevator acceleration, deceleration, stopping, and skipping floors. The reward function is designed as a weighted summation, including an energy consumption term (proportional to the square of acceleration), a waiting time penalty term, an overload penalty term, and an energy efficiency reward term. The specific weight coefficients of each term are not fixed values, but are tuned through a large amount of actual operating data, aiming to find the optimal balance between different performance indicators such as energy consumption, efficiency, and comfort. The determination of the weights relies on the analysis of historical operating data, and is repeatedly adjusted and verified in a simulation environment to finally obtain a set of weighting coefficients that optimize the overall performance of the elevator system. The Q-value is updated using a temporal difference method and iteratively calculated using the Bellman equation. During training, the learning rate uses a preset initial value and gradually decays according to a predetermined strategy based on the training progress to balance convergence speed and stability. The discount factor is set to a predefined value close to one to account for the impact of long-term rewards. The action selection strategy adopts an ε-greedy method, setting the exploration probability to a predetermined value in the early stages of training to ensure sufficient exploration of the environment, and then gradually reducing the probability according to the plan to shift to utilizing learned experience. The entire training process requires a pre-set number of cycles until the Q-value update rate tends to stabilize and the strategy no longer changes significantly, at which point convergence is considered complete. After sufficient training, the system converges to the optimal strategy, outputting the elevator operation frequency sequence and dispatch priority sequence.

[0128] Subsequently, the Monte Carlo tree search algorithm was used for time window optimization. A search tree was constructed, where each node represents a time window partitioning scheme. The root node of the search tree represents the initial or current time window partitioning state; each branch represents a possible partitioning operation, such as splitting an existing time window into two or adjusting its boundaries; child nodes represent the new partitioning scheme obtained after performing the operation; leaf nodes represent a complete, indivisible time window partitioning scheme, which contains a complete and specific partitioning arrangement of the elevator running time. In the selection phase, the upper confidence interval formula was used to calculate the exploration value of each node, balancing exploration and utilization. In the expansion phase, new time window partitioning schemes were generated to reasonably divide the running time. In the simulation phase, the quality of the schemes was evaluated through random sampling, and the average waiting time and energy consumption indicators were calculated. In the backtracking phase, the number of node visits and the cumulative reward value were updated. After sufficient iterative search, the time window division scheme with the optimal comprehensive evaluation index is selected. This comprehensive evaluation index is a predefined quantitative standard that integrates multiple key performance indicators calculated in the simulation phase, such as the average waiting time, the total energy consumption of the elevator system, and the longest waiting time for passengers. It is then fused into a single comprehensive evaluation value through a weighted summation function to comprehensively measure and compare the overall advantages and disadvantages of different time window division schemes, thereby obtaining the optimal operating time window.

[0129] Simultaneously, elevator load rate is monitored in real time. When the load rate exceeds the threshold determined through statistical analysis of historical operating data, an emergency dispatch mechanism is activated. A dynamic programming algorithm is used to reallocate the floor service order: the state is defined as the current elevator position and remaining service demand, the stage as a time node, and the decision as the next service floor. To apply the dynamic programming algorithm, the continuous or high-dimensional state space needs to be discretized. The current elevator position state is discretized by dividing continuous floor positions into several predefined, continuous floor intervals; the remaining service demand state is quantified by dividing the total demand into multiple pre-set discrete demand levels according to a quantity range. This discretization strategy transforms the original continuous or high-dimensional state space into a set of finite discrete state points, thus constructing a discrete state space suitable for dynamic programming solutions. State transition equations and recursive relationships are established. The value function is designed to minimize the total service time. Specifically, in the recursive calculation, the value function is derived from the final state back to the initial state, and the value of each state is defined as the minimum time required to complete all remaining services starting from that state and following the optimal decision. Its calculation comprehensively considers the direct service time consumption, the time consumption corresponding to state transitions caused by decisions, and the optimal value of subsequent states, thus decomposing the global goal of minimizing the total service time into subproblems that can be solved step by step recursively. It also considers energy consumption constraints and maximum waiting time limits. By solving the Bellman equation in reverse order, the optimal service sequence is obtained using a value iteration method, outputting the emergency dispatch floor and the corresponding frequency compensation coefficients.

[0130] Finally, the optimization results were integrated: the elevator operating frequency sequence and frequency compensation coefficient were weighted and fused to generate an adjusted operating frequency plan; the elevator dispatch priority sequence and emergency dispatch floors were merged to generate the final dispatch strategy; and combined with the optimal operating time window, a complete elevator operation plan was formed. This plan improves the operating efficiency and service quality of the elevator system while ensuring that energy constraints are met, providing a reliable technical guarantee for energy-saving operation of commercial complexes.

[0131] In step S18, based on the photovoltaic power generation scheme, the energy storage scheduling scheme, the lighting adjustment scheme, and the elevator operation scheme, a unified format collaborative control instruction is generated and the operation control is executed.

[0132] Specifically, the system first receives the output schemes from each subsystem: the photovoltaic power generation scheme includes a sequence of photovoltaic output commands and module tilt adjustment parameters; the energy storage scheduling scheme includes a discharge power sequence, execution time, and duration; the lighting adjustment scheme includes a sequence of brightness levels for each area and switching time parameters; and the elevator operation scheme includes a sequence of operating frequencies and dispatch priorities. These schemes use a unified time base and are discretized at fixed time intervals.

[0133] System coordinated optimization is achieved through a multi-objective optimization algorithm. An objective function is established, encompassing multiple optimization objectives: minimizing total energy consumption, maximizing photovoltaic (PV) absorption, and optimizing user comfort. Inter-system coupling constraints are set, including power balance constraints, equipment operation constraints, and response time constraints. A weighted summation method is used to transform the multi-objective problem into a single-objective optimization problem. The optimal coordination scheme is then solved using gradient descent. This weighted summation method assigns a pre-defined weight coefficient to each optimization objective based on system operating preferences, combining the multiple objectives of minimizing total energy consumption, maximizing PV absorption, and optimizing user comfort into a unified comprehensive objective function. Subsequently, this comprehensive objective function is used as the optimization object of gradient descent. By calculating the gradient of the function relative to each control variable and iteratively updating the variable values ​​along the gradient descent direction, the optimal coordination scheme that minimizes the comprehensive objective function is finally found.

[0134] The design incorporates an instruction generation mechanism that converts optimization results into executable instructions for the devices. This mechanism first maps the control parameter values ​​output by the optimization algorithm to standardized data formats recognizable by specific devices, according to their corresponding physical meanings and dimensions. Then, based on the communication protocol specifications provided by different device manufacturers, the formatted data is encapsulated into corresponding instruction frame structures. These structures include fields for the target device address, function code, specific control parameter values, and verification information. Finally, a series of control instructions containing precise operating commands that can be directly issued to field actuators are generated. Photovoltaic system instructions include output setpoints and tilt adjustment amounts; energy storage system instructions include charging / discharging power and timing control parameters; lighting system instructions include brightness level values ​​and switch control signals; and elevator system instructions include operating frequency values ​​and priority setting parameters. All instructions utilize a unified data structure and communication protocol.

[0135] This collaborative control process, through rigorous mathematical optimization and real-time monitoring, ensures that all subsystems operate in a coordinated manner under unified scheduling, meeting energy optimization goals while guaranteeing system safety and stability, ultimately maximizing the overall energy efficiency of the commercial complex.

[0136] Reference Figure 2 The second embodiment of the present invention provides a photovoltaic-storage-DC-flexible control system for net-zero carbon commercial complexes, comprising:

[0137] The data acquisition module is used to acquire real-time data including pedestrian flow, photovoltaic power generation, energy storage status, lighting data, and elevator operation data;

[0138] The pedestrian flow prediction module is used to perform data cleaning, trend component extraction, and future pedestrian flow prediction based on the real-time data to obtain the pedestrian flow prediction value.

[0139] The energy demand forecasting module is used to determine peak periods based on the predicted pedestrian flow and, in conjunction with the real-time data, to forecast energy demand and obtain the predicted energy demand value.

[0140] The photovoltaic power generation module is used to perform calculations to minimize photovoltaic curtailment based on the energy demand forecast, obtain the optimal photovoltaic output command, and generate a photovoltaic power generation scheme.

[0141] The energy storage scheduling module is used to calculate the optimal discharge power, execution time and duration for future periods based on the energy storage status when the photovoltaic power generation scheme does not meet the predicted energy demand value, thereby obtaining an energy storage scheduling scheme.

[0142] The lighting adjustment module is used to calculate the optimal brightness level and switching time of the lamps in each area based on the energy demand forecast, the photovoltaic power generation scheme and the lighting data, so as to obtain the lighting adjustment scheme;

[0143] The elevator operation module is used to optimize the elevator operation frequency and scheduling strategy based on the energy demand forecast, the passenger flow forecast and the elevator operation data, so as to obtain the elevator operation plan.

[0144] The output control module is used to integrate and generate unified format collaborative control instructions and execute operation control based on the photovoltaic power generation scheme, the energy storage scheduling scheme, the lighting adjustment scheme and the elevator operation scheme.

[0145] It should be noted that the photovoltaic-storage-direct-flexible control device for net-zero carbon commercial complexes provided in this embodiment of the invention is used to execute all the process steps of the photovoltaic-storage-direct-flexible control method for net-zero carbon commercial complexes described in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0146] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a photovoltaic-storage-DC-flexible control program for net-zero carbon commercial complexes. When the processor executes the computer program, it implements the steps in the various embodiments of the photovoltaic-storage-DC-flexible control method for net-zero carbon commercial complexes described above, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as a photovoltaic-storage-DC-flexible control module for a net-zero carbon commercial complex.

[0147] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0148] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0149] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0150] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0151] If the modules / units integrated into the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0152] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0153] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for controlling the photovoltaic-storage-DC-flexible system in a net-zero carbon commercial complex, characterized in that, include: Acquire real-time data including pedestrian flow, photovoltaic power generation, energy storage status, lighting data, and elevator operation data; Based on the real-time data, data cleaning, trend component extraction, and future pedestrian flow prediction are performed to obtain the pedestrian flow prediction value. Based on the predicted passenger flow, peak periods are determined, and combined with the real-time data, energy demand is predicted to obtain the predicted energy demand value. Based on the energy demand forecast, the photovoltaic curtailment is minimized to obtain the optimal photovoltaic output command and generate a photovoltaic power generation scheme. When the photovoltaic power generation scheme does not meet the predicted energy demand, the optimal discharge power, execution time and duration for the future period are calculated based on the energy storage status to obtain the energy storage scheduling scheme. Based on the energy demand forecast, the photovoltaic power generation scheme, and the lighting data, the optimal brightness level and switching time of the lamps in each area are calculated to obtain a lighting adjustment scheme; Based on the energy demand forecast, the passenger flow forecast, and the elevator operation data, the elevator operation frequency and scheduling strategy are optimized to obtain an elevator operation plan. Based on the photovoltaic power generation scheme, the energy storage scheduling scheme, the lighting adjustment scheme, and the elevator operation scheme, a unified format of collaborative control instructions is generated and the operation control is executed.

2. The photovoltaic-storage-DC-flexible control method for net-zero carbon commercial complexes according to claim 1, characterized in that, The step of performing data cleaning, trend component extraction, and future pedestrian flow prediction based on the real-time data to obtain the pedestrian flow prediction value includes: Based on the real-time data, data denoising is performed using db4 wavelet transform to obtain denoised data; Based on the noise reduction data, the STL decomposition method is used to extract trends in pedestrian flow, photovoltaic power generation, lighting brightness, and elevator frequency to obtain trend data. Based on the trend data, time series prediction modeling is performed using a long short-term memory network, and after iterative training, the predicted pedestrian flow value for future periods is output.

3. The photovoltaic-storage-DC-flexible control method for net-zero carbon commercial complexes according to claim 1, characterized in that, The step of determining peak periods based on the predicted pedestrian flow and combining it with real-time data to predict energy demand, thereby obtaining a predicted energy demand value, includes: When the predicted pedestrian flow is greater than the preset pedestrian flow threshold, it is determined to be a peak period. Combined with real-time data, the operating characteristics are extracted by the sliding window averaging method and the system energy consumption weight of the energy-consuming system is calculated by the analytic hierarchy process. Based on the operational characteristics and the system energy consumption weights, the energy demand forecast is output by fusing the characteristics and performing regression prediction through a pre-established multilayer perceptron model.

4. The photovoltaic-storage-DC-flexible control method for net-zero carbon commercial complexes according to claim 1, characterized in that, The step of minimizing photovoltaic curtailment based on the predicted energy demand, obtaining the optimal photovoltaic output command, and generating a photovoltaic power generation scheme includes: Based on the energy demand forecast and the real-time data, with the objective function of minimizing the amount of curtailed solar power, a linear programming model is solved using the simplex method to output photovoltaic power output commands for each time period. Acquire sunlight data, calculate the optimal tilt angle of photovoltaic modules using a solar position algorithm, and output the tilt angle adjustment parameters of photovoltaic modules; The photovoltaic output command and the photovoltaic module tilt angle adjustment parameters are integrated to generate a photovoltaic power generation scheme.

5. The photovoltaic-storage-DC-flexible control method for net-zero carbon commercial complexes according to claim 1, characterized in that, When the photovoltaic power generation scheme does not meet the predicted energy demand, the optimal discharge power, execution time, and duration for the future period are calculated based on the energy storage status to obtain an energy storage dispatch scheme, including: The photovoltaic power generation capacity is extracted from the photovoltaic power generation scheme, and the difference between the predicted energy demand and the photovoltaic power generation capacity is calculated to obtain the energy gap. Based on the energy shortage and the energy storage status, with the objective function of minimizing the amount of electricity purchased by the grid, a linear programming model is solved using the simplex method to output the optimal discharge power. Historical load curves are obtained, and combined with the optimal discharge power, time period optimization allocation is performed through dynamic programming algorithm to output the start and stop time sequence of discharge periods; Based on the optimal discharge power and the start-stop time sequence, an execution plan is generated through a time-slice round-robin scheduling algorithm, and an energy storage scheduling scheme containing discharge power, execution time, and duration is output.

6. The photovoltaic-storage-DC-flexible control method for net-zero carbon commercial complexes according to claim 1, characterized in that, The step involves calculating the optimal brightness level and switching time of lights in each area based on the predicted energy demand, the photovoltaic power generation scheme, and the lighting data to obtain a lighting adjustment scheme, including: Ambient light data is acquired, and combined with the energy demand forecast, the photovoltaic power generation scheme, and the lighting data, the lighting brightness is graded using the Viterbi algorithm, and the optimal brightness level sequence and on / off time combination is output. Based on the optimal brightness level sequence, a fuzzy PID controller is used to adjust the light intensity and output an anti-disturbance brightness correction value and a switching time offset. Lighting energy consumption is extracted from the lighting data, and energy consumption deviation is compensated through a model predictive control algorithm to output a brightness compensation coefficient. By integrating the optimal brightness level sequence, brightness correction value, brightness compensation coefficient, and switching time offset of the switching period, a lighting adjustment scheme is generated.

7. The photovoltaic-storage-DC-flexible control method for net-zero carbon commercial complexes according to claim 1, characterized in that, The process of optimizing elevator operating frequency and scheduling strategies based on the predicted energy demand, predicted passenger flow, and elevator operation data to obtain an elevator operation plan includes: Based on the energy demand forecast, the passenger flow forecast, and the elevator operation data, the elevator scheduling decision is modeled using the Q-Learning algorithm, and the elevator operation frequency and dispatch priority sequence are output. Based on the elevator operating frequency and dispatch priority sequence, the time window is optimized by Monte Carlo tree search to obtain the optimal operating time window; The elevator load rate is extracted from the elevator operation data. When the elevator load rate exceeds the preset load threshold, the emergency floors are re-allocated using a dynamic programming algorithm to obtain the emergency dispatch floors and frequency compensation coefficients. The elevator operation frequency, the elevator dispatch priority sequence, the emergency dispatch floor, and the frequency compensation coefficient are integrated to generate an elevator operation plan.

8. A photovoltaic-storage-DC-flexible control system for net-zero carbon commercial complexes, characterized in that, include: The data acquisition module is used to acquire real-time data including pedestrian flow, photovoltaic power generation, energy storage status, lighting data, and elevator operation data; The pedestrian flow prediction module is used to perform data cleaning, trend component extraction, and future pedestrian flow prediction based on the real-time data to obtain the pedestrian flow prediction value. The energy demand forecasting module is used to determine peak periods based on the predicted pedestrian flow and, in conjunction with the real-time data, to forecast energy demand and obtain the predicted energy demand value. The photovoltaic power generation module is used to perform calculations to minimize photovoltaic curtailment based on the energy demand forecast, obtain the optimal photovoltaic output command, and generate a photovoltaic power generation scheme. The energy storage scheduling module is used to calculate the optimal discharge power, execution time and duration for future periods based on the energy storage status when the photovoltaic power generation scheme does not meet the predicted energy demand; The lighting adjustment module is used to calculate the optimal brightness level and switching time of the lamps in each area based on the energy demand forecast, the photovoltaic power generation scheme and the lighting data, so as to obtain the lighting adjustment scheme; The elevator operation module is used to optimize the elevator operation frequency and scheduling strategy based on the energy demand forecast, the passenger flow forecast and the elevator operation data, so as to obtain an elevator operation plan. The output control module is used to integrate and generate unified format collaborative control instructions and execute operation control based on the photovoltaic power generation scheme, the energy storage scheduling scheme, the lighting adjustment scheme and the elevator operation scheme.

9. An electronic device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the photovoltaic-storage-DC-flexible control method for a net-zero carbon commercial complex as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the photovoltaic-storage-DC-flexible control method for a net-zero carbon commercial complex as described in any one of claims 1 to 7.

Images