Photovoltaic-edible mushroom collaborative intelligent agricultural industrial fusion management method and system

Abstract

This application relates to the field of agricultural product management technology, and particularly to a smart agricultural industrial integration management method and system for photovoltaic-edible fungi synergy. The proposed scheme involves: acquiring multi-source monitoring data from photovoltaic power generation and edible fungi production units to generate an input dataset for energy-production synergy analysis; identifying the physiological critical period of edible fungi based on manifold embedding and topology matching algorithms, and constructing a counterfactual manifold matching model under energy constraints to achieve verifiable determination of physiological state changes; dynamically evaluating fungal stability and energy consumption efficiency by combining stage perception and energy-production mapping empirical models; and generating a control scheme that balances energy utilization efficiency and fungal stability using a hierarchical multi-objective optimization algorithm, which is then output to the energy management terminal. This application can maintain optimal synergy between fungal growth stability and system energy efficiency under fluctuating energy environments, and is applicable to photovoltaic agriculture, smart edible fungi factories, and new energy-biological coupled agricultural scenarios.

Description

Technical Field

[0001] This application relates to the field of agricultural product management technology, and in particular to a smart agricultural industrial integration management method and system that combines photovoltaic and edible fungi. Background Technology

[0002] Existing photovoltaic-agricultural collaborative control technologies typically only optimize power generation forecasting, energy storage scheduling, or environmental regulation strategies on the energy side, lacking modeling of the correlation between energy disturbances and biological physiological states. This results in insufficient adaptability of the control results to the production process. When faced with photovoltaic fluctuations, existing methods cannot accurately identify key sensitive intervals in the physiological stages of edible fungi, easily leading to decreased growth stability and quality fluctuations. On the other hand, traditional energy consumption optimization methods often rely on fixed thresholds or empirical parameters for adjustment, making it difficult to maintain a dynamic balance between energy utilization efficiency and production stability under multi-source uncertain disturbances. Furthermore, existing smart agricultural control systems generally suffer from low algorithm layer coupling and poor interpretability of optimization results in multi-objective collaboration, failing to achieve structured verification and sustainable control of the fungal response process. Ultimately, this manifests as low system operating energy efficiency, increased production cycle fluctuations, and higher energy utilization redundancy.

[0003] To address the above issues, this application presents a smart agricultural industrial integration management and control method and system that integrates photovoltaic and edible fungi technologies. Summary of the Invention

[0004] The technical problem this application aims to solve is to address the shortcomings of existing technologies by providing a smart agricultural industrial integration management and control method and system for photovoltaic-edible fungi synergy. This method involves acquiring multi-source monitoring data from photovoltaic power generation and edible fungi production units to generate an input dataset for energy-production synergy analysis; identifying the physiological critical periods of edible fungi based on manifold embedding and topology matching algorithms, and constructing a counterfactual manifold matching model under energy constraints to achieve verifiable determination of physiological state changes; dynamically evaluating fungal stability and energy consumption efficiency by combining stage perception and energy-production mapping empirical models; and finally, using a hierarchical multi-objective optimization algorithm to generate a control scheme that balances energy utilization efficiency and fungal stability, which is then output to the energy management terminal.

[0005] To achieve the above objectives, this application provides the following technical solution:

[0006] A smart agricultural industrial integration management and control method for photovoltaic-edible fungi synergy, the method comprising:

[0007] Acquire photovoltaic data and monitoring data from edible fungi production units to generate an input dataset for power-production collaborative analysis;

[0008] Based on the input dataset, the current physiological critical period of the edible fungus is determined by the stage recognition algorithm, and whether the physiological safety boundary is triggered is determined by the preset empirical model.

[0009] If triggered, a control scheme that balances energy utilization efficiency and cell stability is generated through a multi-objective optimization algorithm, and the control scheme is simultaneously output to the energy management terminal.

[0010] The photovoltaic data includes the photovoltaic system's power generation, energy storage status, and electricity price changes. The monitoring data includes the temperature and humidity, carbon dioxide concentration, air velocity, substrate moisture content, growth stage, and energy consumption records of the edible fungi production unit.

[0011] The input dataset for generating the power-production collaborative analysis includes:

[0012] Identify irradiance level change events from the photovoltaic data that correspond to power generation level change markers, energy storage charging and discharging switching, and electricity price time-period step changes, and generate an equivalent photovoltaic disturbance event set.

[0013] Based on the energy consumption records of production units, the start-up and shutdown logs of environmental control, and the changes in the state of charge of energy storage, the allocable energy gap for each time slice in the equivalent photovoltaic disturbance event set is calculated, and the disturbance intensity, duration, and occurrence time in the equivalent photovoltaic disturbance event are filled in according to the allocable energy gap to obtain the photovoltaic power sequence.

[0014] The photovoltaic power sequence is subjected to multi-scale perturbation decomposition, and a time-stamped perturbation fingerprint is generated based on sudden drops, rebounds, short-period jitters, and intraday steps.

[0015] Based on historical batch data and reproductive stage results, response time-shifting rules are determined for the target edible fungi strain, wherein the response time-shifting rules are used to align the perturbation fingerprint to the analysis window that causes changes in the physiological state of the edible fungi;

[0016] Within the analysis window, causal alignment features are constructed by combining the rate of change of environmental parameters and the energy consumption slope, and an input dataset is generated.

[0017] Aligning the perturbation fingerprint to the analysis window that causes changes in the physiological state of edible fungi includes:

[0018] Cluster the rate of change of environmental parameters within the time period corresponding to each disturbance fingerprint to identify the time-delay distribution of different types of disturbances in response to temperature, humidity, carbon dioxide concentration and energy consumption.

[0019] Based on the time lag distribution and combined with the changing trends of physiological indicators during the critical period in historical batches, the time offset mapping relationship of the target edible fungi strain is calculated, wherein the time offset mapping relationship is used to determine the start time and decay period of the impact of the disturbance event.

[0020] Based on the time offset mapping relationship, the time axis of each disturbance type in the disturbance fingerprint is repositioned to obtain the analysis window.

[0021] The process of determining the current physiological critical period of edible fungi using a stage recognition algorithm includes:

[0022] Manifold embedding is performed on causal alignment features within a continuous alignment window to obtain a manifold structure that characterizes the feature distribution morphology;

[0023] Topological analysis is performed on the manifold structures of adjacent aligned windows to obtain structural descriptions of connectivity patterns, loop structures, density ridges, and migration paths for each manifold structure.

[0024] Based on the structural description, a topological matching relationship is constructed between counterfactual manifolds generated by counterfactual trajectories under energy constraints, and the minimum deformation mapping is calculated based on the topological matching relationship under the premise of maintaining connectivity and the number of loops.

[0025] When there is a topological instability between the minimum deformation mapping and the structural description, the current time period is determined to be a physiological critical period, and the corresponding phase transition type, perturbation fingerprint cluster and minimum witness set are output.

[0026] The process of embedding causal alignment features within a continuous alignment window into a manifold structure to characterize the feature distribution morphology includes:

[0027] Dimensionality reduction mapping is performed on the causal alignment features within each alignment window to project multidimensional information into a low-dimensional feature space. The multidimensional information includes at least one or more of the following: disturbance type, disturbance intensity, duration, energy gap ratio, energy storage participation, environmental parameter change rate, energy consumption slope, recovery slope, and alignment hysteresis.

[0028] In the low-dimensional feature space, a connected graph is constructed based on the local density and adjacency of causal alignment features;

[0029] The feature point set of the connected graph is embedded using a manifold learning algorithm to generate a low-dimensional representation for characterizing the photovoltaic perturbation-physiological response. During the manifold construction process, the connectivity weights between feature points are updated by a sliding time window based on the temporal order and energy constraints of the perturbation fingerprints corresponding to the feature point set.

[0030] Geometric morphology tracking is performed on the low-dimensional representation to extract evolutionary features and construct a temporal topological description of the manifold, generating the manifold structure.

[0031] Based on the structural description, a topological matching relationship is constructed between counterfactual manifolds generated by counterfactual trajectories under energy constraints, including:

[0032] Based on the photovoltaic energy supply status, energy storage capacity and operation control restrictions, multiple counterfactual trajectories that meet the energy constraints are generated. The counterfactual trajectories are alternative paths after perturbing the energy distribution and environmental variable time series without changing the external environmental control boundary.

[0033] By embedding the causal alignment features corresponding to each counterfactual trajectory into a manifold, a counterfactual manifold is obtained to characterize the alternative energy-physiological response relationship.

[0034] Using the connectivity patterns and loop structures in the structural description as references, calculate the topology matching mapping between the real manifold and the counterfactual manifold while keeping connectivity and the number of loops constant.

[0035] The empirical model includes a stage-aware layer, a generative mapping layer, a multi-scale temporal fusion layer, and an output layer, wherein:

[0036] The stage perception layer is used to receive the causal alignment feature sequence generated based on historical batch data and the energy supply state parameters of the current batch. It generates a comprehensive feature vector to characterize the stage state by splicing parameters such as the perturbation intensity, duration, energy consumption slope and environmental response rate of the causal alignment features and combining the time-series encoding of photovoltaic energy fluctuation and energy storage release features.

[0037] The energy generation mapping layer includes multiple sets of nonlinear feature mapping units with shared weights, which are used to establish a dynamic correspondence between photovoltaic energy fluctuations, energy storage scheduling delays and the physiological responses of edible fungi. Based on the comprehensive feature vector, multidimensional implicit features are extracted to characterize the impact of energy supply changes on the metabolic stability of fungi.

[0038] The multi-scale temporal fusion layer includes a long-time response channel and a short-time response channel. The long-time response channel is used to capture energy consumption trends and physiological inertia changes in mycelium, while the short-time response channel is used to capture transient energy disturbances, environmental feedback, and stage transition boundary features. The outputs of the two channels are cross-fused through a gated residual mechanism and a phase attention mechanism to generate physiological response features.

[0039] The output layer is used to output the physiological stability score and safety boundary confidence level of the bacterial cells based on the physiological response characteristics, and to determine whether the physiological safety boundary is triggered by the temporal consistency of the physiological stability score, the fluctuation range of the confidence level, and the phase deviation between the physiological stability score and the energy consumption change.

[0040] The method of generating a control scheme that balances energy utilization efficiency and cell stability through a multi-objective optimization algorithm includes:

[0041] An energy utilization objective function and a physiological stability objective function are constructed using a multi-objective optimization algorithm. The energy utilization objective function is used to minimize energy consumption per unit output and photovoltaic power curtailment rate, while the physiological stability objective function is used to maximize cell physiological stability score and quality retention.

[0042] Based on constraints of energy supply and demand balance, energy storage capacity, and environmental regulation capability, a joint feasible domain is constructed.

[0043] The energy utilization objective function and the physiological stability objective function are solved in the joint feasible region by a hierarchical optimization strategy to obtain a control scheme. The hierarchical optimization strategy includes an upper-level optimization logic, a lower-level optimization logic, and a hybrid optimization logic. The upper-level optimization logic is used to determine the ratio of photovoltaic power generation and energy storage scheduling. The lower-level optimization logic is used to solve the environmental parameter setting and energy consumption allocation scheme. The hybrid optimization logic is used to coordinate the convergence between the two-level results.

[0044] A smart agricultural industrial integration control system for photovoltaic-edible fungi synergy, the system comprising:

[0045] The data acquisition unit is used to acquire and synchronize the power generation, energy storage status, electricity price changes of the photovoltaic system and the environmental monitoring data of the edible fungus production unit.

[0046] The collaborative analysis unit is used to identify equivalent disturbance events, extract causal alignment features, and determine physiological critical periods from the collected data, generating energy-biological collaborative analysis results.

[0047] The optimization decision-making unit is used to execute multi-objective optimization algorithms to solve the energy utilization objective function and the physiological stability objective function, and output an energy-life coordinated regulation scheme.

[0048] An execution control unit is used to receive the control scheme and drive the coordinated operation of photovoltaic power generation, energy storage management and edible fungus environmental control equipment.

[0049] Compared with the prior art, the beneficial effects of this application are:

[0050] This application establishes a topological mapping and counterfactual verification mechanism between photovoltaic energy disturbances and the physiological response of edible fungi, thereby achieving bidirectional dynamic synergy between energy utilization and fungal stability. This enables the production process to maintain steady-state operation and optimal energy efficiency under fluctuating energy environments, improving resource utilization and reducing the risk of physiological instability. Attached Figure Description

[0051] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0052] Figure 1 An exemplary application scenario diagram provided for an embodiment of this application;

[0053] Figure 2 This is a schematic diagram of the structure of the photovoltaic-edible fungi synergistic smart agriculture-industry integrated control system provided in the embodiments of this application;

[0054] Figure 3 A schematic flowchart illustrating the photovoltaic-edible fungi synergistic smart agriculture-industrial integration control method provided in this application embodiment;

[0055] Figure 4 A schematic diagram of the structure of the empirical model provided in the embodiments of this application. Detailed Implementation

[0056] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0057] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0058] The following description, using the example of a distributed photovoltaic-edible fungus co-production scenario within an integrated agricultural park, illustrates the applicability of the method described in this application to actual working conditions, so that those skilled in the art can understand the implementation path and expected technical effects.

[0059] During the cultivation of edible fungi, the edible fungi workshop operates in conjunction with the park's rooftop photovoltaic system, station-level energy storage, and time-of-use electricity pricing strategy. Photovoltaic output is high during sunny and hot afternoons but drops rapidly at night, while sudden changes in cloudy and rainy weather cause multiple short-cycle fluctuations within the day. At the same time, the tolerance windows of mycelial expansion, primordia differentiation, and fruiting body enlargement to temperature, humidity, evaporation intensity, and environmental disturbances are significantly different, and there is a non-negligible physiological response time lag.

[0060] Traditional practices typically rely on environmental setpoints and operational experience to make incremental adjustments to ventilation, humidification, and dehumidification based on changes in electricity prices or energy consumption indicators, only resorting to energy storage when necessary. This strategy, which focuses on numerical thresholds, fixed timetables, and minimizing local energy consumption, often encounters two common industry challenges when photovoltaic output experiences frequent random disturbances, energy storage dispatch suffers from response delays, and there are significant batch-to-batch differences in substrates / bacterial strains:

[0061] Firstly, fluctuations on the energy side are directly projected onto the environmental side, leading to the failure of primordia or quality fluctuations during the critical period.

[0062] Secondly, in pursuit of short-term energy efficiency, the steady state in subsequent stages is sacrificed, making it difficult to guarantee batch consistency and quality.

[0063] The method in this application does not rely on the fixed hardware structure of the workshop, but is aimed at the universal scenario of the coupling of renewable energy uncertainty and physiological time delay. It performs data layer modeling by causal alignment between energy disturbance and physiological response, and then identifies the real critical period at the topological level. Finally, it achieves coordinated regulation of energy and physiology through multi-objective optimization.

[0064] In one specific implementation, given the limited grid-connected capacity of the industrial park and its participation in demand response, the edible mushroom workshop adopts a multi-room parallel operation. Firstly, starting with incomplete metering data from the photovoltaic side, changes in irradiance levels implied by power generation level changes, energy storage charging / discharging switching, and electricity price jumps are identified as equivalent photovoltaic disturbance events. Then, by combining energy consumption records and control logs to complete the intensity, duration, and occurrence time, a photovoltaic power time series suitable for subsequent analysis is obtained. Subsequently, this time series undergoes multi-scale disturbance decomposition to generate time-stamped disturbance fingerprints. Based on historical batch and stage results, response time-shift rules are learned for the target fungal species, adaptively aligning energy-side disturbances to the window most likely to trigger physiological changes. The causal alignment features constructed in this way compress asynchronous energy-environment-physiological processes, which are difficult to compare directly, into a computable input dataset within the same time reference frame.

[0065] Furthermore, this application does not use a threshold to determine whether it is critical, but instead observes the geometric and topological evolution of the feature distribution on the time axis by embedding the causal features of the continuously aligned windows into a manifold.

[0066] When the growth state of edible fungi is in the non-critical stage, the embedded manifold maintains a stable continuity of connectivity patterns and density ridges between adjacent windows. Once the coupling relationship between energy perturbation and physiological response undergoes structural distortion, such as the original channel breaking, loop disappearing, or having to undergo folding / tearing to match the counterfactual manifold generated under the same energy constraint, it can be determined that the physiological critical period has been entered.

[0067] Understandably, the topological phase transition recognition mechanism stems from the redefinition of criticality in this application:

[0068] It's not that the environmental quantity of edible fungi exceeds the upper or lower limit, but rather that under the combined effect of photovoltaic constraints and physiological time delay, the stable strategy space experiences structural collapse or no longer becomes homeomorphic with the counterfactual feasible space.

[0069] It should be emphasized that the commonality among the applicable operating conditions of the method in this application is that:

[0070] Uncertainty on the energy side is unavoidable, and physiological responses exhibit time dependence and phase shifts. Traditional control methods driven by empirical thresholds or static formulations are insufficient to provide structural-level safety guarantees. This application transforms the randomness on the photovoltaic side into a computable risk characterization and uses hierarchical optimization to obtain a synergistic solution for photovoltaic scheduling, energy storage release, and environmental trajectories. This allows for early perception of critical periods and interpretable control of the energy-energy coupling relationship without altering the existing hardware architecture.

[0071] In this embodiment, the rate of critical period instability events can be reduced, energy efficiency per unit yield can be improved, and curtailment of solar power can be reduced. At the same time, an evidence package that can be used for auditing and attribution can be provided to meet the requirements of smart agriculture industrial integration for explainable, verifiable and transferable control.

[0072] refer to Figure 1 , Figure 1 This is an exemplary application scenario diagram provided for an embodiment of this application.

[0073] like Figure 1 As shown in the embodiment, the photovoltaic-edible fungus synergistic smart agriculture industrial integration management and control system is applied to a smart agricultural park with distributed photovoltaic power supply and edible fungus production unit.

[0074] The application scenarios mainly include four core components: photovoltaic arrays, energy management units, edible fungi cultivation rooms, and processors.

[0075] In a specific embodiment not shown in the figure, the photovoltaic array is installed on the workshop roof or open space in the park to generate solar power and supply power to the energy management unit. The energy management unit is used to schedule and monitor the photovoltaic power generation, energy storage status, and interaction with the external power grid, forming real-time operating data on the energy side. The edible fungus cultivation room is equipped with environmental control equipment and a multi-point sensor network to collect physiological and production-related data such as temperature, humidity, carbon dioxide concentration, and energy consumption. The processor, as the core control unit of the system, is used to receive data input from the photovoltaic side and the production side, execute energy-life synergy analysis, critical period identification, and multi-objective optimization decision-making logic, and output corresponding control commands to the energy management unit and environmental control device to achieve adaptive and coordinated control of energy distribution and physiological stability.

[0076] refer to Figure 2 , Figure 2 This is a schematic diagram of the structure of the photovoltaic-edible fungi synergistic smart agriculture-industrial integration control system provided in the embodiments of this application.

[0077] In one example, the system includes:

[0078] The data acquisition unit is used to acquire and synchronize the power generation, energy storage status, electricity price changes of the photovoltaic system and the environmental monitoring data of the edible fungus production unit.

[0079] The collaborative analysis unit is used to identify equivalent disturbance events, extract causal alignment features, and determine physiological critical periods from the collected data, generating energy-biological collaborative analysis results.

[0080] The optimization decision-making unit is used to execute multi-objective optimization algorithms to solve the energy utilization objective function and the physiological stability objective function, and output an energy-life coordinated regulation scheme.

[0081] An execution control unit is used to receive the control scheme and drive the coordinated operation of photovoltaic power generation, energy storage management and edible fungus environmental control equipment.

[0082] Next, with reference to the accompanying drawings, the intelligent agricultural and industrial integration management and control method of photovoltaic-edible fungi synergy provided in the embodiments of this application will be further described. Figure 3 The methods shown include:

[0083] S1: Acquire photovoltaic data and monitoring data from edible fungi production units to generate an input dataset for power-production collaborative analysis;

[0084] In this embodiment, photovoltaic data sources include the power generation of the photovoltaic array, irradiance, and energy storage state of charge. Simultaneously, data on temperature and humidity, carbon dioxide concentration, energy consumption records, air velocity, and substrate moisture content in the edible fungus cultivation workshop are collected. The data is uniformly collected by an independent data acquisition device and synchronized in time. After perturbation identification and time-series completion, an input dataset is formed. This allows for the estimation of power generation sequences and energy supply and demand relationships through equivalent photovoltaic perturbation events, even in the absence of continuous power measurements, thereby achieving time-series alignment between the energy side and the physiological side.

[0085] Those skilled in the art will understand that photovoltaic data and monitoring data can be flexibly configured according to specific equipment types or sensing conditions, as long as they can reflect the coupling state between energy supply and physiological environment to a minimum, and this application does not impose any limitations.

[0086] S2: Based on the input dataset, determine the current physiological critical period of the edible fungus using a stage recognition algorithm;

[0087] In this embodiment, the stage recognition algorithm performs manifold embedding on causal alignment features to extract the topological changes in the feature distribution within a continuous time window. When the manifold exhibits connectivity mode breaks, density ridge collapses, or topological instability between adjacent windows, it is considered to have entered a physiological critical period.

[0088] Unlike traditional threshold-based methods, this application's method does not rely on fixed environmental upper and lower limits, but rather on the impact of energy disturbances on the physiological response structure of bacteria. This approach allows for the early identification of risk windows before significant environmental changes are caused by photovoltaic fluctuations or energy storage delays, enabling predictive intervention at critical physiological stages.

[0089] S3: Determine whether the physiological safety boundary is triggered based on a preset experience model;

[0090] In this embodiment, the empirical model comprises a stage-aware layer, an energy-generating mapping layer, a multi-scale temporal fusion layer, and an output layer. The model receives the causal alignment features and energy state parameters of the current batch, and outputs a bacterial physiological stability score and a safety boundary confidence level after multi-layer mapping. When the score deviates continuously or the confidence level collapses, the physiological safety boundary is triggered. By fusing photovoltaic fluctuation frequency and bacterial response inertia, adaptive modeling for different bacterial species and substrate conditions is achieved, avoiding the problem of previous empirical formulas being unable to be transferred.

[0091] Those skilled in the art will understand that the empirical model can be implemented in various forms, as long as it can characterize the dynamic impact of energy disturbance on physiological homeostasis, and this application will not elaborate on this.

[0092] S4: If triggered, a control scheme that balances energy utilization efficiency and cell stability is generated through a multi-objective optimization algorithm, and the control scheme is output to the energy management terminal simultaneously.

[0093] In this embodiment, the multi-objective optimization algorithm takes maximizing photovoltaic energy utilization and optimizing the bacterial physiological stability score as dual objectives, and solves the problem by combining energy supply and demand balance, energy storage capacity, and environmental regulation constraints. The upper-level optimization determines the scheduling ratio of photovoltaic and energy storage, while the lower-level optimization calculates ventilation, humidification, and energy consumption allocation parameters, ultimately forming a comprehensive regulation scheme including temperature and humidity setting trajectories and energy storage release strategies. Under the conditions of random fluctuations in photovoltaic power and conflicts with environmental inertia, it can achieve a dynamic balance between energy utilization efficiency and bacterial physiological stability, avoiding energy waste or physiological fluctuations caused by traditional single-objective optimization.

[0094] Before delving into the specific technical details of the steps, the embodiments of this application need to be emphasized again.

[0095] In a typical photovoltaic-biological co-production system, the relationship between energy and physiology is not a stable functional mapping, but rather a feedback relationship with dynamic time delays. Energy-side disturbances are often sudden and unpredictable, while the physiological response is limited by metabolic inertia and microclimate regulation delays. This time misalignment frequently leads traditional control strategies to face the dilemma of responding too late or too excessively. Furthermore, due to the highly nonlinear characteristics of the edible fungi production environment, even small deviations in its microenvironment can trigger significant fluctuations in macroscopic yield and quality at specific stages. The existence of this sensitive window is the theoretical basis for the physiological critical period concept proposed in this application.

[0096] To address the aforementioned issues, the control method proposed in this embodiment no longer uses static setpoints as the basis for regulation. Instead, it captures the system's state evolution trajectory under different energy inputs by reconstructing the causal manifold relationship between energy disturbances and physiological states. When the causal manifold relationship experiences local tearing or mapping instability in the topological structure, it signifies that the bacterial state has entered a critical region highly sensitive to disturbances from the stable domain. Within this logical framework, whether something is critical no longer depends on exceeding the limits of a single quantity such as temperature, humidity, or carbon dioxide concentration, but is defined by the continuity of the overall system response structure.

[0097] On the other hand, on the energy side, traditional optimization often focuses on photovoltaic utilization or energy storage balance, while ignoring the potential for such optimization to induce the accumulation of disturbances on the physiological side. This embodiment introduces a multi-objective optimization framework, treating physiological stability and energy efficiency as parallel objectives. When the power curve on the photovoltaic power generation side suddenly drops, the optimization algorithm not only considers energy balance constraints but also calculates the disturbance propagation path and time-shifting effect in the physiological manifold in real time, thereby generating a control scheme that balances short-term energy efficiency and long-term stability.

[0098] It should be noted that the energy-storage synergy logic adopted in this application does not rely on specific sensing hardware or environmental structures. As long as the most basic energy input, energy storage status, and environmental feedback data can be obtained, the manifold embedding and topology identification calculations can be completed. This feature allows this application to be flexibly deployed in agricultural parks, factory workshops, or experimental modules of different sizes without requiring significant modifications to existing facilities. Through the above approach, this embodiment, while retaining the original energy management mechanism of the photovoltaic system, achieves an integrated closed loop of energy disturbance perception, physiological state prediction, and optimized control, thereby improving the stability and interpretability of the synergy system from a fundamental perspective.

[0099] Next, we will further elaborate on the technical aspects of the input dataset in this application.

[0100] It is understandable that the design of the input dataset in this application is not merely a matter of organizing and stitching together sensor data, but rather a means of establishing a computable causal alignment framework between the energy and physiological sides. By structurally reconstructing the time dependency between photovoltaic-side perturbation behavior and bacterial physiological responses, the coupled dynamics that cannot be directly measured can be presented in data form.

[0101] It is understandable that the photovoltaic-physiological synergistic system exhibits a distinct dual-timescale characteristic:

[0102] The changes in photovoltaic power exhibit high-frequency fluctuations, while the physiological processes of edible fungi have a delayed response and long inertia. If processed directly with a uniform time step, the features on both sides will be distorted and mismatched in the model input, resulting in the causal direction of energy perturbation and physiological feedback being masked.

[0103] This application, based on the principle of system identification, treats continuous energy fluctuations as a series of equivalent disturbance events, using their occurrence time, duration, and disturbance intensity as the smallest structural unit. It then utilizes energy consumption logs and control records to infer the energy gap and reconstruct the power distribution from incomplete time-series data. This event-based approach redefines random fluctuations as a finite number of discrete disturbances with physical semantics, solving the problems of clock misalignment and signal sparsity.

[0104] Furthermore, by introducing response time-shifting rules based on historical batches, each perturbation event is projected onto the timeline to the window most likely to trigger physiological changes, thereby generating causal alignment features. The logical basis of this transformation lies in:

[0105] If there exists an alignment method that ensures a stable correlation between the rate of environmental change and the energy consumption slope before and after the disturbance across different batches, then the time offset corresponding to this alignment is the inherent response time delay of the system. In other words, this application completes the time reference system correction between energy disturbance and physiological state during the input construction stage, enabling the data itself to possess causal comparability and topological stability.

[0106] In one example, the generation of the input dataset for power generation synergistic analysis includes:

[0107] S1.1: Identify irradiance level change events corresponding to power generation level change markers and energy storage charge / discharge switching from the photovoltaic data, and generate an equivalent photovoltaic disturbance event set;

[0108] Specifically, to transform discontinuous or limited metering information into event-based inputs usable for subsequent alignment and modeling, joint analysis of time-series records from both the photovoltaic (PV) and energy storage sides is required. PV-side irradiance level changes can be triggered by signals such as inverter status words, active power step changes at the grid connection point, and very short-time voltage-reactive power compensation switching. Energy storage-side charge / discharge switching can be indicated by the direction of state-of-charge changes, DC-side power sign flipping, and converter operation mode logs. By establishing an indication bit sequence under a unified clock and merging the time spans between adjacent valid indications, candidate irradiance level change segments are obtained. These segments are expressed with a start time, end time, and level label, providing boundaries for subsequent completion of intensity and duration parameters.

[0109] S1.2: Based on the energy consumption records of the production unit, the start-up and shutdown logs of the environmental control and the changes in the state of charge of the energy storage, calculate the allocatable energy gap of each time slice in the equivalent photovoltaic disturbance event set, and fill in the disturbance intensity, duration and occurrence time in the equivalent photovoltaic disturbance event according to the allocatable energy gap to obtain the photovoltaic power sequence;

[0110] Specifically, to reconstruct continuous power time series under incomplete metering conditions, the energy conservation principle is used to constrain the allocatable energy within event segments. Energy consumption trajectories on the workshop side are summarized on a time-slice basis, including start / stop logs and rated power of sub-loads such as ventilation, humidification, dehumidification, circulating pumps, and lighting. Combined with short-term energy consumption metering or meter pulse statistics, lower and upper limits of energy consumption are obtained. The energy storage state-of-charge differential is then converted into DC equivalent power, superimposed on the aforementioned energy consumption intervals, and correlated with the measured power at the grid connection point or the estimated irradiance level to obtain the allocatable energy gap for each time slice. This allocatable energy gap is used to backfill the power trajectory around the start and end of the event, ensuring that the event intensity and duration are no longer determined solely by the state word, but rather by the consistency constraints of energy inflow and outflow.

[0111] In this embodiment, power backfilling adopts a combination of piecewise linear and segment smoothing: linear interpolation is used to quickly approximate the intensity within the event, and spline smoothing is used to avoid artificial breakpoints at the event boundary; when there are low confidence marker events, the principle of minimizing energy gap and the direction of energy storage charge change are used for consistency screening to eliminate candidates that contradict the energy flow direction.

[0112] S1.3: Perform multi-scale perturbation decomposition on the photovoltaic power sequence, and generate a time-stamped perturbation fingerprint based on sudden drops, rebounds, short-period jitters and intraday steps;

[0113] Specifically, the goal of multi-scale perturbation decomposition is to represent power changes of different time scales and forms with a unified fingerprint entry, facilitating cross-day and cross-batch comparisons. The photovoltaic power sequence first undergoes trend-fluctuation separation to extract intraday baselines, then performs windowed search on the residual signal to identify candidate locations for sudden drops, rebounds, short-period jitters, and steps. Each candidate is categorized according to amplitude, duration, slope variation, and neighborhood symmetry, generating a fingerprint entry that includes type, start and end times, amplitude, recovery slope, and neighborhood background.

[0114] In this embodiment, fingerprint generation includes:

[0115] When backfilling occurs in the power sequence, fingerprint extraction employs a strict threshold in segments with high backfill confidence and a relaxed threshold with attached confidence labels in segments with low confidence, facilitating sample weighting during subsequent response time-shift learning. The fingerprint database supports cross-day inheritance, enabling the use of intraday step templates from the previous day for rapid matching on the next day, accelerating generation and improving consistency. Short-period jitter fingerprints filter out noisy high-frequency spikes through minimum period constraints, retaining energy perturbations that have potential impacts on physiological processes.

[0116] S1.4: Based on historical batch data and reproductive stage results, determine the response time-shifting rule for the target edible fungus species, wherein the response time-shifting rule is used to align the perturbation fingerprint to the analysis window that causes changes in the physiological state of the edible fungus;

[0117] Specifically, the response time-shift rule is established based on the temporal distribution of physiological changes induced by isomorphic perturbations in different batches. By comparing and learning from the triples of fingerprint-environmental change rate-stage results in historical batches, the time shift and duration of influence of different fingerprint types at different reproductive stages are determined. The response time-shift rule is not a fixed constant, but a mapping relationship conditioned on stage, matrix, and control strength, allowing for dynamic correction with a small number of calibration samples when a new batch is started. The output of the response time-shift rule is a set of alignment window parameters for each fingerprint type, including the offset center, window width, and confidence weight, used to project energy-side events to the time period most likely to induce physiological changes.

[0118] In one example, aligning the perturbation fingerprint to the analysis window that causes changes in the physiological state of edible fungi includes:

[0119] Cluster the rate of change of environmental parameters within the time period corresponding to each disturbance fingerprint to identify the time-delay distribution of different types of disturbances in response to temperature, humidity, carbon dioxide concentration and energy consumption.

[0120] Based on the time lag distribution and combined with the changing trends of physiological indicators during the critical period in historical batches, the time offset mapping relationship of the target edible fungi strain is calculated, wherein the time offset mapping relationship is used to determine the start time and decay period of the impact of the disturbance event.

[0121] Based on the time offset mapping relationship, the time axis of each disturbance type in the disturbance fingerprint is repositioned to obtain the analysis window.

[0122] Specifically, for each time period corresponding to a perturbation fingerprint, the instantaneous rate of change of environmental parameters (including temperature, humidity, carbon dioxide concentration, and air velocity) is extracted, and a time difference sequence is constructed with the time of perturbation occurrence as a reference. To eliminate the influence of differences in perturbation amplitude and background, the rate of change is normalized and resampled on a time scale to obtain a unified perturbation-response sample set. The perturbation-response sample set is then clustered using a distance metric based on dynamic time warping to form several typical time-delay response patterns. Each pattern corresponds to the response characteristics of a certain type of energy perturbation to the physiological environment. The clustering results not only reveal the distribution patterns of different perturbation types in terms of time delay but also provide a statistical basis for subsequently establishing species-specific time-off mapping relationships.

[0123] Furthermore, after obtaining the time-lag distribution, and combining the changing trends of physiological critical indicators (such as changes in mycelial respiration rate, delayed primordia formation, or abnormal fruiting body water potential) recorded in historical batches, the occurrence times of physiological effects corresponding to different perturbation types are calculated. The calculation process employs a probability-weighted time-off inference logic:

[0124] Using the peak moment of the environmental response as the central point, a time offset is fitted according to the probability distribution of physiological index changes in batch samples, thus obtaining the offset mapping relationship between perturbation and physiological response. This offset mapping relationship describes the sensitive time domain of different bacterial species to specific perturbation morphologies, including not only the onset time of the response but also the response decay period, used to characterize the persistence and recovery features of the perturbation's impact. Unlike traditional fixed-delay models, this application uses statistical inference to adaptively adjust the time offset according to bacterial species, reproductive stage, and energy consumption pattern, avoiding the problem that a single delay parameter cannot reflect complex physiological processes.

[0125] Furthermore, based on the aforementioned time offset mapping relationship, each perturbation type is repositioned along the time axis to form an analysis window that matches the physiological response. The repositioning process uses the time of perturbation occurrence as the reference point, the corresponding offset interval as the window center, and sets the window width according to the decay period. After time axis repositioning, similar perturbations in different batches are aligned to time periods with similar physiological responses, thereby ensuring the comparability and structural consistency of subsequent causal feature construction and manifold embedding analysis.

[0126] Through the aforementioned processing, the time reference of the input data is transformed from physical sampling time to physiological effective time, enabling the model to maintain a stable characterization of the relationship between energy disturbance and physiological state under different production conditions, and to achieve early identification and dynamic perception of physiological critical periods.

[0127] S1.5: Within the analysis window, combine the rate of change of environmental parameters and the energy consumption slope to construct causal alignment features and generate the input dataset;

[0128] Specifically, after the alignment window is established, the environmental parameter change rate sequence and energy consumption slope sequence are calculated within the window at a fixed step size, and then spliced ​​with the type, intensity, duration, and recovery slope of the corresponding disturbance fingerprint entry to form an event-response aligned feature slice.

[0129] Next, we will further elaborate on the technical content of the method of this application regarding the physiological critical period.

[0130] It is understood that the physiological critical period in this application specifically refers to the time period during which the ability of edible fungi to maintain a stable physiological state against external disturbances enters a sensitive range under photovoltaic energy constraints and the conditions of the reproductive stage. During this time period, the set of feasible regulatory strategies for maintaining stability undergoes structural contraction, breakage, or loss of homeomorphic correspondence with the set of feasible alternative strategies.

[0131] In other words, the physiological critical period is not defined by changes in the growth stage of edible fungi, but by structural changes that occur in the time-aligned reference frame in the causal chain of energy disturbance-microclimate response-physiological state.

[0132] Those skilled in the art will understand that, even if the nominal environmental values ​​remain compliant during this period, physiological stability may still be compromised due to topological degradation of the strategy space. Therefore, placing the criterion at the structural layer rather than the threshold layer is more in line with the actual working conditions of collaborative production.

[0133] In one example, determining the current physiological critical period of the edible fungus using a stage identification algorithm includes:

[0134] S2.1: Perform manifold embedding on the causal alignment features within the continuous alignment window to obtain a manifold structure that characterizes the feature distribution morphology;

[0135] Specifically, to ensure that energy perturbations and physiological responses form measurable structural objects within the same reference frame, a low-dimensional representation is constructed for the causal alignment features within each alignment window. The feature terms include fields such as perturbation type, perturbation intensity, duration, energy gap percentage, energy storage participation, environmental parameter change rate, energy consumption slope, recovery slope, and alignment lag. First, dimensional unification, distribution stretching, and outlier annotation are performed. Then, temporal location encoding is introduced to preserve the temporal semantics within the window. Subsequently, adjacency relationships are established at both the same-batch and cross-batch levels. A connected graph is generated based on local density and nearest-neighbor connectivity criteria. Finally, a manifold learning method that preserves both local geometry and the global skeleton embeds the high-dimensional features into the low-dimensional space.

[0136] In this embodiment, the embedding process adopts a two-stage strategy: in the first stage, an initial connected graph is constructed within a batch based on a fixed number of neighbors and a density threshold condition to obtain low-dimensional embeddings within the batch; in the second stage, the embedding results of different batches are aligned and spliced ​​based on the perturbation fingerprint similarity of the alignment window between batches to form an embedding sequence on a continuous time axis.

[0137] In one example, the manifold embedding of causal alignment features within a continuous alignment window to obtain a manifold structure characterizing the feature distribution morphology includes:

[0138] Dimensionality reduction mapping is performed on the causal alignment features within each alignment window to project multidimensional information into a low-dimensional feature space. The multidimensional information includes at least one or more of the following: disturbance type, disturbance intensity, duration, energy gap ratio, energy storage participation, environmental parameter change rate, energy consumption slope, recovery slope, and alignment hysteresis.

[0139] In the low-dimensional feature space, a connected graph is constructed based on the local density and adjacency of causal alignment features;

[0140] The feature point set of the connected graph is embedded using a manifold learning algorithm to generate a low-dimensional representation for characterizing the photovoltaic perturbation-physiological response. During the manifold construction process, the connectivity weights between feature points are updated by a sliding time window based on the temporal order and energy constraints of the perturbation fingerprints corresponding to the feature point set.

[0141] Geometric morphology tracking is performed on the low-dimensional representation to extract evolutionary features and construct a temporal topological description of the manifold, generating the manifold structure.

[0142] Specifically, to ensure that the multidimensional causal alignment features retain the dynamic correlation characteristics of physiological processes in the low-dimensional space, structural preprocessing of the input high-dimensional dataset is required. The feature entries within each alignment window include perturbation intensity, duration, energy gap percentage, energy storage participation, environmental parameter change rate, energy consumption slope, recovery slope, and alignment lag. First, normalization and range constraints ensure comparability of parameters in the amplitude space. Then, temporal position encoding is introduced to parameterize the temporal sequence of perturbations, distinguishing the different roles of leading and lagging perturbations in physiological responses. Temporal position encoding uses the center time of the sliding window as a reference and employs linear interpolation to embed it into the feature vector, ensuring continuity between adjacent windows on the time axis. The essential purpose here is to map the originally asynchronous perturbation information to a unified temporal reference frame, thereby forming a continuous phase trajectory after temporal embedding, which can be used for subsequent geometric analysis.

[0143] In this embodiment, manifold embedding is achieved by constructing a connected graph. The connected graph uses the similarity and energy constraint relationships between causally aligned features as the weighting basis, no longer relying solely on Euclidean distance, but introducing energy consumption difference and energy storage response consistency as auxiliary weighting factors. The weight calculation employs a hybrid form of dynamic time-warped distance and energy dissimilarity, and is updated in real-time within a sliding time window, ensuring that the connectivity strength of similar samples adjusts accordingly when the photovoltaic energy supply state undergoes abrupt changes. This effectively avoids the geometric tearing problem that occurs in traditional dimensionality reduction algorithms during periods of energy instability, ensuring the structural continuity of the manifold under energy perturbations. Simultaneously, to reduce overfitting caused by feature space sparsity, local kernel density estimation is used to suppress isolated point weights, ensuring that the connected graph retains only statistically significant feature neighborhoods.

[0144] Furthermore, to capture the dynamic evolution characteristics of the manifold within different time windows, geometric morphology tracking is performed on the low-dimensional representation. This process achieves continuous mapping of the manifold structure by calculating the rate of change of the principal direction field, the migration path of density ridges, and the trajectory of cluster centers between adjacent windows. Specifically, the low-dimensional point sets of aligned windows t and t+1 are established with a temporal correspondence through nearest neighbor matching, and a geometric evolution vector field is generated using the rate of change of curvature and the local density change trend. When the distribution of the evolution vector shows obvious direction reversal, loop breakage, or density contraction, it indicates that a potential phase transition has occurred in the physiological state. This geometric morphology tracking not only provides a computable index of topological changes but also enables the manifold to naturally express the dynamic process from energy perturbation to physiological transition. The final manifold structure is not just a static mapping of feature distribution but a dynamic geometric object with temporal continuity, energy consistency, and topological analyzability, providing a high-resolution structured basis for the identification of physiological critical periods.

[0145] S2.2: Perform topological analysis on the manifold structure of adjacent aligned windows to obtain the structural description of the connectivity mode, loop structure, density ridge and migration path corresponding to each manifold structure;

[0146] Specifically, to continuously track the morphological changes in the distribution of causal features over time, topological features are extracted and temporally compared for the manifold structures of adjacent aligned windows. Connectivity patterns are obtained through threshold control maps and cluster detection to determine the number of clusters, cluster size, and shortest bridging paths between clusters; loop structures are generated through skeleton extraction and loop detection to determine the number of loops, loop length distribution, and key bridging edges; density ridges are extracted using kernel density contour lines to obtain main ridge segments and bifurcation points; migration paths are described by the temporal evolution of cluster center trajectories and high-density channels to depict morphological migration across windows.

[0147] In this embodiment, structural analysis employs a window alignment method for sliding computation. This involves extracting structural elements from the manifold structures of aligned windows t and t+1, and establishing element correspondences using stable graph matching criteria. To reduce element jitter caused by sporadic noise, a persistence rule is set: an element is confirmed to enter the structural description only if it remains present in at least two consecutive windows; if an element appears only in a single window, it is marked as a transient event for subsequent consistency checks but does not enter the backbone description. For migration paths, path segments with consistent directions are generated by tracking the overlap between cluster centers and high-density channels in adjacent windows, and the path strength and path update count are recorded to subsequently determine whether channels break or fork.

[0148] S2.3: Based on the structural description, construct the topological matching relationship between counterfactual manifolds generated by counterfactual trajectories under energy constraints, and calculate the minimum deformation mapping under the premise of maintaining connectivity and the number of loops according to the topological matching relationship;

[0149] Specifically, to verify whether the feasible policy space under real-world operation maintains structural consistency with the alternative policy space, a set of counterfactual trajectories is generated under the same energy constraint, and their causal features are embedded to obtain a counterfactual manifold. The matching relationship is based on the fundamental constraint of maintaining connectivity and the number of loops. Under this constraint, the minimum deformation path that maps the counterfactual manifold to the real manifold is sought. Deformation allows for fine-tuning of node positions and correction of edge weights, but disconnecting existing connected channels or adding new loops is prohibited to ensure structural homogeneity.

[0150] In one example, based on the structural description, a topological matching relationship is constructed between counterfactual manifolds generated by counterfactual trajectories under energy constraints, including:

[0151] Based on the photovoltaic energy supply status, energy storage capacity and operation control restrictions, multiple counterfactual trajectories that meet the energy constraints are generated. The counterfactual trajectories are alternative paths after perturbing the energy distribution and environmental variable time series without changing the external environmental control boundary.

[0152] By embedding the causal alignment features corresponding to each counterfactual trajectory into a manifold, a counterfactual manifold is obtained to characterize the alternative energy-physiological response relationship.

[0153] Using the connectivity patterns and loop structures in the structural description as references, calculate the topology matching mapping between the real manifold and the counterfactual manifold while keeping connectivity and the number of loops constant.

[0154] Specifically, firstly, based on the photovoltaic energy supply status, energy storage capacity, and production operation control constraints, multiple counterfactual trajectories satisfying energy constraints are generated. The generation of counterfactual trajectories is not a simple random perturbation, but rather a result of joint constraints based on the energy balance equation and environmental control boundary conditions. By applying small, reversible perturbations to parameters such as energy allocation ratio, control delay, and environmental response rate without changing the upper and lower limits of the control system output, multiple theoretically achievable but not actually occurring alternative energy-physiological response paths are generated. These paths represent the potential evolutionary possibilities of the system under different energy allocation strategies. The generation algorithm must ensure that each counterfactual trajectory holds true in the sense of energy conservation, i.e., the balance between photovoltaic power generation, energy storage discharge, and load consumption satisfies the constraint equations. Through this process, a family of trajectories reflecting the potential feasible state space is constructed, providing a reference object for subsequent topology matching.

[0155] In this embodiment, after generating counterfactual trajectories, the causal alignment features corresponding to each trajectory are processed by the same manifold embedding algorithm to form a counterfactual manifold. To ensure the comparability of the two types of manifolds, the embedding parameters must be consistent, especially in terms of neighborhood radius, density threshold, and time window size, so that the counterfactual manifold and the real manifold can be regarded as equipotential forms under different energy configurations in low-dimensional space. At this time, the counterfactual manifold and the real manifold should ideally maintain a topological homeomorphism in terms of geometric morphology, that is, have the same number of connected clusters, number of loops, and density skeleton morphology. If the counterfactual manifold breaks, collapses, or the number of loops changes abruptly under a certain perturbation condition, it means that the alternative strategy under the energy constraint can no longer maintain the same physiological stable region structure, thus it can be inferred that the system is in a physiological critical state during that period.

[0156] Furthermore, using the structural description of the real manifold as a reference, topological matching calculations are performed on the counterfactual manifold. The matching process is based on minimum deformation mapping under the constraints of maintaining connectivity and the number of loops. The deformation field from the counterfactual manifold to the real manifold is solved by minimizing the mapping cost function (composed of node correspondence error, edge weight offset, and skeleton curvature). If matching can be completed within a finite deformation range, it indicates that the feasible strategy space of the system still has topological stability under the current energy perturbation; if topological operations (such as edge breaking, folding, and reconnection) are required to achieve matching, it indicates that the stable structure of the system has been destroyed. The matching process introduces energy constraint weights, so that the feasibility of the mapping is affected not only by the geometric shape but also by the energy accessibility, thus ensuring that the analysis results have physical meaning. Finally, the calculated matching error distribution and deformation intensity image can directly reflect the type of phase transition that occurs during the critical period, such as strategy space collapse (cluster merging caused by energy supply interruption) and feedback delay instability (sudden increase in the number of loops).

[0157] S2.4: When there is a topological consistency instability between the minimum deformation mapping and the structural description, determine that the current time period is a physiological critical period, and output the corresponding phase transition type, perturbation fingerprint cluster and minimum witness set;

[0158] Specifically, topological consistency instability refers to the inability of the real manifold and the counterfactual manifold to complete registration under the minimum deformation constraint by maintaining connectivity and no increasing or decreasing loops, or by introducing continuous folding, tearing, bridging disappearance and other operations to achieve matching. Once such a situation occurs, it means that the feasible strategy space for maintaining physiological stability within the given energy constraint has undergone structural degradation, and the current time period is determined to have entered the physiological critical period.

[0159] refer to Figure 4 , Figure 4 A schematic diagram of the structure of the empirical model provided in the embodiments of this application.

[0160] In one example, the empirical model includes a stage-aware layer, a generative mapping layer, a multi-scale temporal fusion layer, and an output layer, wherein:

[0161] The stage perception layer is used to receive the causal alignment feature sequence generated based on historical batch data and the energy supply state parameters of the current batch. It generates a comprehensive feature vector to characterize the stage state by splicing parameters such as the perturbation intensity, duration, energy consumption slope and environmental response rate of the causal alignment features and combining the time-series encoding of photovoltaic energy fluctuation and energy storage release features.

[0162] The energy generation mapping layer includes multiple sets of nonlinear feature mapping units with shared weights, including nonlinear feature mapping unit one to nonlinear feature mapping unit N, which are used to establish a dynamic correspondence between photovoltaic energy fluctuations, energy storage scheduling delays and the physiological responses of edible fungi. Based on the comprehensive feature vector, multidimensional implicit features are extracted to characterize the impact of energy supply changes on the metabolic stability of fungi.

[0163] The multi-scale temporal fusion layer includes a long-time response channel and a short-time response channel. The long-time response channel is used to capture energy consumption trends and physiological inertia changes in mycelium, while the short-time response channel is used to capture transient energy disturbances, environmental feedback, and stage transition boundary features. The outputs of the two channels are cross-fused through a gated residual mechanism and a phase attention mechanism to generate physiological response features.

[0164] The output layer is used to output the physiological stability score and safety boundary confidence level of the bacterial cells based on the physiological response characteristics, and to determine whether the physiological safety boundary is triggered by the temporal consistency of the physiological stability score, the fluctuation range of the confidence level, and the phase deviation between the physiological stability score and the energy consumption change.

[0165] It should be noted that the structure of the empirical model described in this embodiment is not limited to a specific network form. Long short-term memory networks, graph convolutional networks, or self-attention mechanism networks can all replace the corresponding modules, as long as a dynamic causal mapping relationship can be established between energy perturbation and physiological response, the same technical effect can be achieved.

[0166] In one example, the generation of a control scheme that balances energy utilization efficiency and cell stability using a multi-objective optimization algorithm includes:

[0167] S3.1: Construct an energy utilization objective function and a physiological stability objective function through a multi-objective optimization algorithm, wherein the energy utilization objective function is used to minimize the energy consumption per unit output and the photovoltaic power curtailment rate, and the physiological stability objective function is used to maximize the bacterial cell physiological stability score and quality retention.

[0168] Specifically, to ensure the control scheme is both energy- and physiologically feasible and measurable, the objective function is parameterized using observable operational quantities and phased key performance indicators. The energy utilization objective function revolves around two indicators: energy consumption per unit output and curtailment rate. Energy consumption per unit output is calculated by time-aligned, normalized cumulative statistics of energy consumption and measurable output over a rolling time domain. The curtailment rate is measured by the proportion of unused allocable energy in the equivalent photovoltaic disturbance event library. Both are accumulated in segments using phase boundary points as nodes, making the objective sensitive to phase switching and load transfer. The physiological stability objective function extracts physiological stability scores and quality retention from the empirical model output as core metrics. It incorporates the temporal consistency of the scores, the phase difference of the scores to energy disturbances, and the persistence of fluctuations near phase boundaries into weighting terms to prevent the optimizer from exploiting short-term shocks to improve long-term scores.

[0169] S3.2: Based on constraints of energy supply and demand balance, energy storage capacity, and environmental regulation capability, construct a joint feasible domain;

[0170] Specifically, the joint feasible region uses energy conservation and equipment capacity as boundaries, establishing a set of soft and hard constraints for each time slice within the rolling time domain. Energy supply and demand balance constraints are checked by matching photovoltaic allocable energy, grid-connected power upper and lower bounds, and load demand, and the constraints of the demand response period are incorporated with priority labels, allowing the feasible region to automatically shrink when demand response is triggered. Energy storage capacity constraints are linked to the upper and lower limits of state of charge, charge / discharge rates, and lockout intervals, and a minimum hold time for charge / discharge switching is added to avoid ineffective fluctuations from frequent reverse operations; simultaneously, a loss model is introduced as an equivalent cost, automatically degrading overcharging and discharging during optimization. Environmental control capability constraints cover the upper and lower bounds of the set trajectory for temperature and humidity channels, slope constraints, dead zone width, and actuator duty cycle upper limits. For the CO2 channel and ventilation volume, maximum ramp rate and hysteresis control band are added to the lower limits of hygiene and safety regulations to ensure the trajectory can be followed by the actual execution chain in a closed loop.

[0171] In this embodiment, the feasible region construction maintains consistency with the temporal benchmark of the causal alignment features. Specifically, at the alignment window granularity, the allowable fluctuation band of the environmental trajectory and the reachable set of the actuator are defined, and the time slice corresponding to the minimum witness set in the critical period evidence package is set as the contraction band, ensuring that the optimization search avoids crossing structurally vulnerable regions. To improve solution efficiency, the feasible region is partitioned:

[0172] Use a loose band during non-sensitive periods and tighten constraints in critical neighborhoods. At the same time, provide equivalent softening rules for each type of constraint, allowing minor out-of-bounds violations within an acceptable cost range. Record the scope and duration of the out-of-bounds violation with a penalty item to facilitate subsequent auditing and playback review.

[0173] S3.3: The energy utilization objective function and the physiological stability objective function are solved in the joint feasible region through a hierarchical optimization strategy to obtain the control scheme. The hierarchical optimization strategy includes upper-level optimization logic, lower-level optimization logic and hybrid optimization logic. The upper-level optimization logic is used to determine the ratio of photovoltaic power generation and energy storage scheduling. The lower-level optimization logic is used to solve the environmental parameter setting and energy consumption allocation scheme. The hybrid optimization logic is used to coordinate the convergence between the two-level results.

[0174] Specifically, the hierarchical optimization strategy adopts a solution process of "upper-layer energy allocation - lower-layer environmental trajectory - inter-layer coordination". Under the joint feasible region projection, the upper-layer optimization logic searches for the allocation path of photovoltaic direct supply, energy storage charging and discharging, and grid connection interaction with the photovoltaic power curve and energy storage state of charge as variables. The goal is to minimize the cost of curtailment and equivalent energy consumption and smooth the power time series so that the lower layer can obtain an energy trajectory with a slope and amplitude that can be followed. The lower-layer optimization logic solves for the temperature and humidity setting trajectory, ventilation and humidification rhythm, and energy consumption allocation under the given energy allocation of the upper layer. It performs rolling optimization with the temporal consistency of physiological stability scores and phase deviation within the alignment window as the core constraints. In the critical neighborhood, it prioritizes finding trajectory families that maintain the integrity of connectivity channels and density ridges, and performs topology fence checks on candidate families to prevent the introduction of short-term strong interferences that may cause channel breakage.

[0175] In this embodiment, inter-layer coordination employs a hybrid optimization logic for convergence management and weight adaptation. Specifically, after each iteration between upper and lower layers, a fast counterfactual replay is performed on candidate solutions to evaluate their robustness in the critical neighborhood using empirical models and alignment features. If score rollback or increased topological fence risk occurs, a penalty is sent back to the upper layer to encourage energy allocation to shift towards lower-risk solutions. If the lower layer experiences execution chain saturation or limited ramping, unreachable information is sent back to the upper layer, requiring a reallocation of charging / discharging power and grid-connected interaction. To improve solution efficiency, the upper layer uses historical best solutions and the early morning solution of the day as a warm start, while the lower layer uses neighborhood splines of the upper layer's allocation as the starting trajectory. Within each sliding window, three to five candidate trajectories with different emphases are retained to facilitate rapid switching and maintain diversity.

[0176] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A smart agricultural industrial integration management and control method for photovoltaic-edible fungi synergy, characterized in that, The method includes: Acquire photovoltaic data and monitoring data from edible fungi production units to generate an input dataset for power-production collaborative analysis; Based on the input dataset, the current physiological critical period of the edible fungus is determined by the stage recognition algorithm, and whether the physiological safety boundary is triggered is determined by the preset empirical model. If triggered, a control scheme that balances energy utilization efficiency and bacterial physiological stability is generated through a multi-objective optimization algorithm, and the control scheme is synchronously output to the energy management terminal. The input dataset for generating the power-production collaborative analysis includes: Identify irradiance level change events corresponding to power generation level change markers and energy storage charge / discharge switching from the photovoltaic data, and generate an equivalent photovoltaic disturbance event set. Based on the energy consumption records of production units, the start-up and shutdown logs of environmental control, and the changes in the state of charge of energy storage, the allocable energy gap for each time slice in the equivalent photovoltaic disturbance event set is calculated, and the disturbance intensity, duration, and occurrence time in the equivalent photovoltaic disturbance event are filled in according to the allocable energy gap to obtain the photovoltaic power sequence. The photovoltaic power sequence is subjected to multi-scale perturbation decomposition, and a time-stamped perturbation fingerprint is generated based on sudden drops, rebounds, short-period jitters, and intraday steps. Based on historical batch data and reproductive stage results, response time-shifting rules are determined for the target edible fungi strain, wherein the response time-shifting rules are used to align the perturbation fingerprint to the analysis window that causes changes in the physiological state of the edible fungi; Within the analysis window, causal alignment features are constructed by combining the rate of change of environmental parameters and the energy consumption slope, and an input dataset is generated. Aligning the perturbation fingerprint to the analysis window that causes changes in the physiological state of edible fungi includes: Cluster the rate of change of environmental parameters within the time period corresponding to each disturbance fingerprint to identify the time-delay distribution of the response of different disturbance types to temperature, humidity, carbon dioxide concentration and energy consumption; Based on the time lag distribution and combined with the changing trends of critical period physiological indicators in historical batch data, the time offset mapping relationship of the target edible fungi strain is calculated, wherein the time offset mapping relationship is used to determine the start time and decay period of the impact of the disturbance event. Based on the time offset mapping relationship, the time axis of each disturbance type in the disturbance fingerprint is repositioned to obtain the analysis window.

2. The intelligent agricultural industrial integration management and control method for photovoltaic-edible fungi synergy according to claim 1, characterized in that, The photovoltaic data includes the power generation and energy storage status of the photovoltaic system, and the monitoring data includes the temperature and humidity, carbon dioxide concentration, air velocity, substrate moisture content, growth stage, and energy consumption records of the edible fungi production unit.

3. The intelligent agricultural industrial integration management and control method for photovoltaic-edible fungi synergy according to claim 1, characterized in that, The process of determining the current physiological critical period of edible fungi using a stage recognition algorithm includes: Manifold embedding is performed on causal alignment features within a continuous alignment window to obtain a manifold structure that characterizes the feature distribution morphology; Topological analysis is performed on the manifold structures of adjacent aligned windows to obtain structural descriptions of connectivity patterns, loop structures, density ridges, and migration paths for each manifold structure. Based on the structural description, a topological matching relationship is constructed between counterfactual manifolds generated by counterfactual trajectories under energy constraints, and the minimum deformation mapping is calculated based on the topological matching relationship under the premise of maintaining connectivity and the number of loops. When there is a topological consistency instability between the minimum deformation mapping and the structural description, the current time period is determined to be a physiological critical period, and the corresponding phase transition type, perturbation fingerprint cluster and minimum witness set are output. Based on the structural description, a topological matching relationship is constructed between counterfactual manifolds generated by counterfactual trajectories under energy constraints, including: Based on the photovoltaic energy supply status, energy storage capacity and operation control restrictions, multiple counterfactual trajectories that meet the energy constraints are generated. The counterfactual trajectories are alternative paths after perturbing the energy distribution and environmental variable time series without changing the external environmental control boundary. By embedding the causal alignment features corresponding to each counterfactual trajectory into a manifold, a counterfactual manifold is obtained to characterize the alternative energy-physiological response relationship. Using the connectivity patterns and loop structures in the structural description as references, calculate the topology matching mapping between the real manifold and the counterfactual manifold while keeping connectivity and the number of loops constant.

4. The intelligent agricultural industrial integration control method for photovoltaic-edible fungi synergy according to claim 3, characterized in that, The process of embedding causal alignment features within a continuous alignment window into a manifold structure to characterize the feature distribution morphology includes: The causal alignment features within each alignment window are subjected to dimensionality reduction mapping, and multidimensional information is projected onto a low-dimensional feature space. The multidimensional information includes at least one or more of the following: disturbance type, disturbance intensity, duration, energy gap ratio, energy storage participation, environmental parameter change rate, energy consumption slope, recovery slope, and alignment hysteresis. In the low-dimensional feature space, a connected graph is constructed based on the local density and adjacency of causal alignment features; The feature point set of the connected graph is embedded using a manifold learning algorithm to generate a low-dimensional representation for characterizing the photovoltaic perturbation-physiological response. During the manifold structure construction process, the connectivity weights between feature points are updated by a sliding time window based on the temporal order and energy constraints of the perturbation fingerprints corresponding to the feature point set. Geometric morphology tracking is performed on the low-dimensional representation to extract evolutionary features and construct a temporal topological description of the manifold, generating the manifold structure.

5. The intelligent agricultural industrial integration management and control method for photovoltaic-edible fungi synergy according to claim 1, characterized in that, The empirical model includes a stage-aware layer, a generative mapping layer, a multi-scale temporal fusion layer, and an output layer, wherein: The stage perception layer is used to receive the causal alignment feature sequence generated based on historical batch data and the energy supply state parameters of the current batch. It generates a comprehensive feature vector to characterize the stage state by splicing parameters such as the perturbation intensity, duration, energy consumption slope and environmental response rate of the causal alignment features and combining the time-series encoding of photovoltaic energy fluctuation and energy storage release features. The energy generation mapping layer includes multiple sets of nonlinear feature mapping units with shared weights, which are used to establish a dynamic correspondence between photovoltaic energy fluctuations, energy storage scheduling delays and the physiological responses of edible fungi. Based on the comprehensive feature vector, multidimensional implicit features are extracted to characterize the impact of energy supply changes on the metabolic stability of fungi. The multi-scale temporal fusion layer includes a long-time response channel and a short-time response channel. The long-time response channel is used to capture energy consumption trends and physiological inertia changes in mycelium, while the short-time response channel is used to capture transient energy disturbances, environmental feedback, and stage transition boundary features. The outputs of the two channels are cross-fused through a gated residual mechanism and a phase attention mechanism to generate physiological response features. The output layer is used to output the physiological stability score and safety boundary confidence level of the bacterial cells based on the physiological response characteristics, and to determine whether the physiological safety boundary is triggered by the temporal consistency of the physiological stability score, the fluctuation range of the confidence level, and the phase deviation between the physiological stability score and the energy consumption change.

6. The intelligent agricultural industrial integration control method for photovoltaic-edible fungi synergy according to claim 5, characterized in that, The method of generating a regulatory scheme that balances energy utilization efficiency and bacterial physiological stability through a multi-objective optimization algorithm includes: An energy utilization objective function and a physiological stability objective function are constructed using a multi-objective optimization algorithm. The energy utilization objective function is used to minimize energy consumption per unit output and photovoltaic power curtailment rate, while the physiological stability objective function is used to maximize cell physiological stability score and quality retention. Based on constraints of energy supply and demand balance, energy storage capacity, and environmental regulation capability, a joint feasible domain is constructed. The energy utilization objective function and the physiological stability objective function are solved within the joint feasible region by a hierarchical optimization strategy to obtain a control scheme. The hierarchical optimization strategy includes an upper-level optimization logic, a lower-level optimization logic, and a hybrid optimization logic. The upper-level optimization logic is used to determine the ratio of photovoltaic power generation and energy storage scheduling. The lower-level optimization logic is used to solve the environmental parameter setting and energy consumption allocation scheme. The hybrid optimization logic is used to coordinate the convergence between the two-level results.

7. A photovoltaic-edible fungi synergistic smart agriculture-industrial integration control system, used to implement the photovoltaic-edible fungi synergistic smart agriculture-industrial integration control method as described in any one of claims 1-6, characterized in that, The system includes: The data acquisition unit is used to acquire and synchronize the power generation, energy storage status, electricity price changes of the photovoltaic system and the environmental monitoring data of the edible fungus production unit. The collaborative analysis unit is used to identify equivalent photovoltaic disturbance events, extract causal alignment features, and determine physiological critical periods from the collected data, generating power-production collaborative analysis results. The optimization decision unit is used to execute a multi-objective optimization algorithm to generate a regulation scheme that takes into account both energy utilization efficiency and bacterial physiological stability, and outputs the regulation scheme to the execution control unit; An execution control unit is used to receive the control scheme and drive the coordinated operation of photovoltaic power generation, energy storage management and edible fungus environmental control equipment.

Images