Design of an Improved IoT-Based PV-Powered Soil Remote Monitoring System with Low Data Acquisition Failure Rate

Abstract

To enable remote and automatic monitoring of the farmland soil information, this paper has developed a soil monitoring system based on the Internet of Things (IoT), which mainly involves the development of a gateway server node, wireless sensor nodes, a remote monitoring platform, and photovoltaic (PV) modules. The Raspberry Pi 5-based gateway server periodically sends data acquisition commands to wireless sensor nodes via LoRa, receives soil data returned by sensor nodes, and stores them in a MySQL database. Using a remote monitoring platform, Internet users can monitor real-time and historical soil data stored in the database. The STM32F103C8T6-based wireless sensor node receives data acquisition commands from the gateway server, uses soil temperature and humidity sensors as well as a pH sensor to collect soil status, and then sends sensor data back to the gateway server via LoRa. The system is powered by both PV energy and batteries, which enhances the endurance capability. Experimental results show that the designed system works well in remotely monitoring soil information. Using the proposed query attempt dynamic adjustment (QADA) method, the wireless sensor node dynamically adjusts the number of query attempts, which reduces the data acquisition failure rate from 21–25% to no more than 0.33%. Using the obtained qualitative relationship that the data acquisition delay varies inversely with the LoRa transfer rate, the data acquisition delay can be reduced to less than 67 ms.

1. Introduction

Agricultural soil status monitoring serves as the core foundation for the sustainable development of modern agriculture. Precisely acquiring key parameter information, such as soil temperature, humidity, and pH value, is of great significance for understanding crop growth environments and optimizing agricultural resource management [1,2]. Therefore, developing intelligent agricultural soil monitoring systems is an important technological path for realizing smart agriculture [3], as well as an inevitable requirement for safeguarding food security and agricultural ecological environment.

With the rapid development of modern information technology, intelligent agricultural monitoring systems have been increasingly used [4]. The characteristics of the existing works, such as the contributions, are summarized in

Table 1. Summary of the related works.

References	Contributions	Open Issues or Limits
[5]	A wireless underground sensor network is proposed to monitor soil moisture using a multi-hop routing protocol.	The maximum round-trip delay was 190 s.
[6]	Wireless sensor nodes are used to acquire and transmit georeferenced soil information.	It presents high levels of error in measuring nutrients.
[7]	A soil leachate monitoring system is designed to evaluate phosphorus and nitrogen in agricultural runoff.	Wired signal transmission and on-site analysis are required.
[8]	A soil moisture monitoring system for cotton fields is designed using electromagnetic induction technology, and the Kalman filter algorithm is employed to enhance the model precision.	The precision of the global model and its applicability across broader areas still require extensive research.
[9]	Passive ultra-high frequency (UHF) radio frequency identification (RFID) is used for sensor nodes to enable continuous soil monitoring.	The read range only exceeds 2 m, and no external sensors are used.
[10]	The work designs a remote monitoring system for agricultural environmental information, including soil, weather and crops, and introduces a fuzzy PID controller to optimize water management.	The controller practical implementation faces challenges such as complexity and computing resources.
[11]	A heterogeneous sensing system for soil moisture mapping is proposed, which contains wireless sensor networks for large-scale continuous detection, mobile robots for in situ monitoring, and portable probes for flexible sampling.	Environmental factors such as soil salinity, greatly affect radio frequency signal transmission.
[12]	A cost-efficient linear robotic camera system with one degree of freedom is designed for automatic plant photography utilizing Wi-Fi communication.	The wireless communication performance is not investigated.
[13]	The work uses sensor nodes to detect agricultural information, including soil moisture and temperature, and send sensor data to the gateway node via Wi-Fi.	The communication distance of Wi-Fi limits the coverage range of the monitoring system.
[14]	The Internet of Things (IoT) and the Transformer model are employed to monitor agricultural soil moisture, temperature, and air humidity, as well as predict temperature and air humidity data.	The application faces the challenge of a lack of communication infrastructure such as the Internet and Wi-Fi.
[15,16]	The works use the Arduino-based sensor nodes to acquire agricultural information including soil moisture and send data to a cloud server via Wi-Fi.	Soil pH is not monitored [15] and the method [16] is only evaluated in a short period of time.
[17]	The work designs printed RFID sensor tags to collect soil parameters, uses a UHF reader on a robot to collect sensor data, and sends data to a web server via Wi-Fi and Bluetooth.	The reading distance of the tags buried in the soil is 70 cm, and a longer reading range is expected.
[18]	A low-cost agriculture monitoring system is designed to collect information such as soil moisture, which contains Arduino-based hardware using Wi-Fi, and server and phone applications.	The communication distance of the used Wi-Fi protocol is limited.
[19]	Two monitoring systems are developed based on Wi-Fi and global systems for mobile communication (GSM) to collect crop, soil, and outdoor information.	Soil pH is not monitored, and environmental changes are not used to improve prediction accuracy.
[20]	The work designs a monitoring system with sensor nodes connecting with the gateway via LoRa, and makes spatio-temporal estimation of soil moisture and temperature.	The outliers have a significant impact on the estimated value.
[21]	The work designs an agricultural irrigation management system, where sensors transmit data such as soil moisture to the gateway via LoRa.	The inconsistent power supply in rural regions can cause disruptions to the system operation.
[22]	The work develops a farm monitoring system, where soil moisture probes communicate with a base station via radio link, the base station sends data to the gateway via LoRa, and the gateway connects with the server via GSM.	The communication performance, such as data transmission success rate and delays, is not investigated.
[23]	The work develops a monitoring system for woody crops, where Arduino-based sensor nodes for soil parameters communicate with the gateway via LoRa, and the gateway sends data to an FTP server via a router.	The sensor accuracy for soil water content is low, and the communication between the sensor and the gateway is unidirectional.
[24]	An optical wireless sensor network is designed to monitor changes in soil composition, and the gateway nodes forward sensor data to a cloud server.	Its limitations lie in the use of artificial samples and the relatively small sample size.
[25]	The work studies the power consumption of an agriculture monitoring system, where sensor nodes for soil moisture connect to the gateway via Bluetooth, and the gateway sends data to a cloud server via GSM.	The Bluetooth communication range and battery lifetimes are limited.
[26]	The work develops sensor nodes to monitor soil moisture, matric potential, and hydraulic conductivity, and sends data to a cloud server via GSM.	Wireless communication performance is not given enough attention.

.

In addition to the respective strengths and limitations of the aforementioned literature, they also have the following common limitations. Most nodes exchange information using Wi-Fi or Bluetooth technologies, which limits the communication range. Most sensor and gateway nodes are powered by battery packs, which limits the endurance time. Data acquisition failure rate and delay are seldom studied, which can deteriorate system performance. To address the aforementioned issues, this paper designs an IoT-based PV-powered soil remote monitoring system with low data acquisition failure rate and low data acquisition delay. Main contributions are presented as follows. First, a complete hardware system with an IoT architecture has been independently developed, where the Raspberry Pi 5-based gateway server node can collect soil information from multiple STM32F103C8T6-based wireless sensor nodes via the long-range, low-power LoRa protocol. Second, a comprehensive software system has been independently designed, which enables the hardware system to automatically collect and store soil temperature, humidity, and pH value information, while providing a remote access platform for Internet users to view real-time and historical data. Third, a novel QADA method is proposed to enable wireless sensor nodes to dynamically adjust the number of query attempts, reducing the data acquisition failure rate from 21–25% to 0.33% or less. Using the obtained inverse relationship between the data acquisition delay and the LoRa transfer rate, the data acquisition delay can be reduced to less than 67 ms.

2. System Architecture

The system architecture is shown in Figure 1. In the perception layer, wireless sensor nodes detect information of soil moisture, temperature, and pH value, which are designed using the STM32F103C8T6 microcontroller unit (MCU) and soil sensors. In the network layer, the LoRa wireless communication technology is used for data transmission between the sensor node and the gateway server node. In the application layer, the gateway server node is designed based on the Raspberry Pi 5 with the Django framework, which receives sensor data and stores them in a MySQL database. The gateway server node uses a 4G/Wi-Fi unit to connect to the Internet, and employs the fast reverse proxy (FRP) network address translation (NAT) traversal service for public network access. Using the designed remote monitoring platform, remote users can conveniently monitor real-time and historical soil information via Internet terminals.

The main objective of this paper is to develop an IoT-based PV-powered soil remote monitoring system with a low data acquisition failure rate and low data acquisition delay. It mainly encompasses the development of both hardware and software systems, primarily involving the design of wireless sensor nodes, a gateway server node, PV power supply systems, and a remote monitoring platform. In addition to enabling remote monitoring of soil temperature, humidity, and pH, the system to be designed is required to maintain a data acquisition failure rate of less than 1% and a data acquisition delay of less than 100 ms.

3. System Hardware Design

The system hardware structure is shown in Figure 2, which mainly includes a wireless sensor node, a gateway server node, and a PV power module. The wireless sensor node communicates with the gateway server node via LoRa unit using the 915 MHz frequency band, while the gateway server node connects with Internet via 4G/Wi-Fi. The PV power module with converters provides energy to loads such as the gateway server node.

3.1. Wireless Sensor Node

As shown in Figure 3, the wireless sensor node is designed using the LoRa communication unit with antenna, STM32F103C8T6 MCU, RS485-TTL conversion unit, soil temperature and humidity sensor, and pH sensor. To improve the node’s endurance capacity, it is powered by both of lithium battery pack and a PV power unit.

Using the RS485-TTL conversion unit, the STM32F103C8T6 MCU collects data from the soil temperature and humidity sensor SN-3000-TR-N01 and the pH sensor ZTS-3000-TR-pH. As shown in

Table 2. Sensor parameters.

Soil Sensor Type	Model	Operating Voltage	Measure Range	Resolution	Accuracy
temperature sensor	SN-3000-TR-N01	4.5–30 V DC	−40~80 °C	0.1 °C	0.5 °C
humidity sensor	SN-3000-TR-N01	4.5–30 V DC	0–100%	0.1%	3%
pH sensor	ZTS-3000-TR-pH	5–30 V DC	3–9 pH	0.1 pH	0.3 pH

, these sensors have the advantages of wide voltage compatibility, high resolution, and high accuracy.

3.2. LoRa Communication Unit

For the wireless communication protocols, including LoRa, Zigbee, Wi-Fi, Bluetooth, and fifth generation,

Table 3. Comparison of wireless communication protocols.

Protocol	Transmission Range	Power Consumption	Data Rate	Cost
LoRa	15 km	Ultra-low	62 kbps	Low
Zigbee	10–100 m	Low	250 kbps	Low
Wi-Fi	100 m	Medium	11–9600 Mbps	Low
Bluetooth	15–30 m	Low	1–2 Mbps	Low
Fifth generation	100–300 m	Medium	<1 × 107 kbps	High

provides a comparison in terms of transmission range, power consumption, data rate, and cost [27,28,29]. Clearly, LoRa has the advantages of long transmission range, low power consumption, and low cost, but it suffers from a low data rate. In the soil remote monitoring system to be designed, wireless sensor nodes need to be deployed in farmland, with requirements for long communication distance, low power consumption, and low cost. These wireless sensor nodes primarily transmit information on soil temperature, humidity, and pH value, with low demands on the data rate. LoRa is adequate to meet these requirements. Therefore, LoRa has been selected as the wireless communication protocol for this application.

The LoRa unit A39C-T900A30D1a works in fixed-point transmission mode with a frequency band of 915 MHz and maximum transmission power of 30 dBm. The circuit connection between the LoRa unit and the STM32F103C8T6 MCU is shown in Figure 4. VCC and GND provide a 5 V DC power supply. RXD and TXD are used for serial communication with the STM32F103C8T6 MCU. The logic levels in MD1 and MD2 are used to adjust LoRa’s operating mode [30]. When transmitting data, the LoRa unit switches to running mode with an operating current of approximately 6.4 mA. During idle periods, it switches to standby mode with an operating current of approximately 2.5 μA, which helps to reduce energy consumption.

3.3. Gateway Server Node

As shown in Figure 5, the gateway server node is designed using the Raspberry Pi 5, LoRa unit, 4G/Wi-Fi router unit, and power supply unit. The Raspberry Pi 5 is equipped with a 2.4 GHz quad-core 64-bit Arm Cortex-A76 processor and Debian-based operating system [31]. By deploying a Django Web server and MySQL database, the Raspberry Pi 5 can perform data collection, storage, and push functions. By configuring the FRP client to achieve NAT traversal, the gateway server can be accessible from the public Internet. The LoRa unit is connected to the Raspberry Pi 5 using a 40-pin ribbon cable with communication baud rate of 115,200. Using this LoRa unit with broadcast address 0x02, the Raspberry Pi 5 communicates wirelessly with the wireless sensor node. The 4G/Wi-Fi industrial router unit AR300 is connected to the Raspberry Pi 5 via RJ45 ethernet interface, which enables Internet users to access the server.

The power supply unit is connected to a 12 V power source, which can be PV power, a battery, or utility power. In addition to powering the LoRa and router units, the power supply unit also supplies 5 V voltage to the Raspberry Pi 5 through a power delivery (PD) protocol-based voltage conversion unit.

3.4. PV Power Module

As shown in Figure 6, the PV power module is designed to provide power to the gateway node and other loads, such as the irrigation pump, which contains PV panels, a lithium battery, and an inverter-controller unit. Four LR5-72HGD-580M PV panels connect in series with a peak voltage of 164.28 V, peak current of 10.75 A, and peak power of 1766 Wp. Inverter-controller unit GROWATT-SPF 3500 ES has the Maximum Power Point Tracking function with operating voltage 120 V–430 V DC, rated power 3500 W, and maximum PV charging current 80 A. Lithium battery LiFePO4 has the parameters of capacity 100 AH, rated voltage 51.2 V, maximum charging current 50 A, and maximum discharging current 100 A.

PV panels generate electricity, supply power to the loads, and store the surplus electricity in the battery. If the output of PV panels is insufficient, the inverter-controller unit controls the battery to supply power to the loads.

4. System Software Design

As shown in Figure 7, system software mainly contains software in the gateway server and wireless sensor nodes and the remote monitoring platform. The gateway server periodically sends a data acquisition command to the wireless sensor node via LoRa. The wireless sensor node collects soil information and transmits sensor data to the gateway server via LoRa. The gateway server receives sensor data, appends the timestamp, and stores it in the MySQL database [32]. Using a remote monitoring platform, Internet users can send web requests to the gateway server via the HTTP protocol. The server responds to web requests, retrieves data from the MySQL database, and sends data back to user terminals.

4.1. Software Design for Gateway Server

The gateway server mainly handles the reception, processing, and storage of data transmitted by wireless sensor nodes, as well as HTTP template management and web application programming interface (API) processing. The gateway server software is developed based on the Django’s MVC (Model-View-Controller) design pattern [33]. The Controller handles web requests and interacts with the Model when retrieving sensor data. The Model establishes connections to the database, executes data query operations, and sends the results to the View. The View, represented by HTML files, dynamically displays the data through charts.

As shown in Figure 8, using the preset period, the gateway server sends a data acquisition command 0x00 0x05 0x41 0xDA 0xDA to its LoRa unit via the USART interface where 0x00 0x05 indicates the LoRa address of wireless sensor node, 0x41 denotes the communication channel with a frequency band of 915 MHz, and 0xDA 0xDA indicates that the objective is to obtain sensor data. Using LoRa communication, the wireless sensor node receives the data acquisition instruction 0xDA 0xDA, drives sensors to collect soil data, and transmits sensor data back to the gateway server. The gateway server receives sensor data and stores them with the timestamp in dat_soildata table in the MySQL database, which includes fields of id, Soiltime, Soiltemp, Soilhum, and Soilph.

Using a remote monitoring platform, the gateway server can respond to web requests initiated by Internet users and retrieve the stored sensor data from the database for display, which will be comprehensively introduced in Section 4.3.

4.2. Software Design for Wireless Sensor Node

Wireless sensor node receives data detection instructions from the gateway server via LoRa, uses sensors to collect soil temperature, humidity, and pH value, and then returns sensor data to the gateway server.

As shown in the flowchart in Figure 9, the serial ports of the STM32F103C8T6 board are initialized. USART1 serves as a debug interface using a baud rate of 9600, USART2 communicates with the LoRa unit using a baud rate of 9600, and USART3 interfaces with sensors via the RS485-TTL conversion unit using a baud rate of 115,200. In addition, interrupt handler functions USART2_IRQHandler of USART2 and USART3_IRQHandler of USART3 are initialized. Direct Memory Access (DMA), including DMA1 and DMA2, is also configured, where the sixth channel and third channel of DMA1 are used to store the received data of USART2 and USART3, respectively.

After receiving data from the gateway server, the LoRa unit sends them to the STM32F103C8T6 MCU via USART2. If it is the data acquisition command 0xDA 0xDA, the STM32F103C8T6 MCU sends data query frames to soil sensors through USART3. First, the STM32F103C8T6 MCU sends a query frame 0x01 0x03 0x00 0x00 0x00 0x02 0xC4 0x0B via USART3, the soil temperature and humidity sensor with address 0x01 returns response frame 0x01 0x03 0x01 0xE6 0x00 0xFC 0x80 0xA5 with humidity 0x01 0xE6 (i.e., 48.6%) and temperature 0x00 0xFC (i.e., 25.2 degrees Celsius). Then, the STM32F103C8T6 MCU sends another query frame 0x02 0x03 0x00 0x00 0x00 0x01 0x84 0x39 via USART3, and soil pH sensor with address 0x02 returns response frame 0x02 0x03 0x02 0x00 0x42 0xE3 0x8E with pH value 0x00 0x42 (i.e., 6.6).

After receiving soil sensors’ response frames, the STM32F103C8T6 MCU generates a data frame 0x00 0x02 0x41 0x05 0x00 0xFC 0x01 0xE6 0x00 0x42 0x05, and transmits it to its LoRa unit via USART2. Using LoRa communication, the gateway server receives data 0x05 0x00 0xFC 0x01 0xE6 0x00 0x42 0x05, where the check bit 0x05 is used to verify the LoRa address 0x00 0x05 of the wireless sensor node, and 0x00 0xFC, 0x01 0xE6 and 0x00 0x42 indicate soil temperature, humidity and pH value, respectively.

4.3. Remote Monitoring Platform

The remote monitoring platform has functions of real-time and historical monitoring of sensor data, and remote configuration of system parameters, which is developed using HTML, CSS, and JavaScript technologies [34].

As shown in Figure 10, an Internet user logs in web browser via uniform resource locator (URL). Web requests are sent to the backend API via Ajax, which are then forwarded to the Django-based gateway server via NAT traversal technology. The gateway server returns the HTML template with data in Json format. The browser processes the information received from the gateway server, and generates HTML pages for display.

5. System Test and Performance Analysis

5.1. System Test

As shown in Figure 11, sensor probes are inserted in the soil to detect soil temperature, humidity, and pH value. Using the LoRa unit with antenna, the sensor node wirelessly sends soil data to the gateway server. To improve the power sustainability, the sensor node is powered by a PV power unit (5 V, 18 W) and a lithium battery (5 V, 30,000 mAh).

As shown in Figure 12, the gateway server receives soil information from the wireless sensor node via the LoRa unit with antenna, and stores data with timestamp in the MySQL database. Using the remote monitoring platform, Internet users can monitor real-time and historical soil data retrieved from the database in the gateway server.

Figure 13 shows the soil information monitoring webpage. The horizontal axis represents time, while the vertical axes represent soil temperature, soil moisture, and soil pH, respectively. By selecting the sensor number and specifying the start and end times for the database query, users can view the soil information detected by that sensor during the selected time period. For example, Figure 13 displays the soil information detected by Sensor 01 between 15:54 and 15:56 on 14 November 2025. As shown in partially enlarged views, the soil temperature at 15:55:55 was 21.5 °C, the soil moisture at 15:55:15 was 14.9%, and the soil pH at 15:54:15 was 6.4. In addition, by clicking the Monitor button in the menu bar, as shown in Figure 14, users can remotely view video information of the agricultural planting environment.

5.2. Data Acquisition Failure Rate

When the wireless sensor node receives a data acquisition command from the gateway server, the STM32F103C8T6 MCU sends a query frame to the soil sensors. If the soil sensors do not return a response frame within the predefined waiting time, the wireless sensor node cannot send sensor data to the gateway server, resulting in data acquisition failure. Data acquisition failure has a negative impact on the performance of networked systems [35].

The data acquisition failure rate $r$ indicates the failure rate of the gateway server acquiring data from the wireless sensor node, which is defined as

$$r=(1-{\displaystyle \frac{n_r}{n_s}})\times 100\%$$

(1)

where $n_s$ denotes the number of data acquisition commands sent by the gateway server, and $n_r$ indicates the number of data packets received by the gateway server from the wireless sensor node.

To analyze the data acquisition failure rate, as shown in Figure 15, experiments have been conducted on the Henan Agricultural University (HAU) campus in Zhengzhou, China. The red Point A at the center of the campus represents the location of the gateway server, and this location remains unchanged throughout the experiment. The red Points 1–9, distributed in different directions across the campus, represent the 1st–9th placement locations where the wireless sensor node can effectively communicate with the gateway server at Point A.

During the experiment, the weather was sunny with a gentle breeze, but the numerous buildings and trees on campus introduced interference in system performance testing. Each testing point underwent three repeated tests, with each test lasting 600 s. Using a sampling period of 2 s and a LoRa transfer rate of 62.5 kbps, the data acquisition failure rates are obtained in

Table 4. Data acquisition failure rate on the HAU campus.

Sensor Location	Distance from Point A	Data Acquisition Failure Rate Under the Traditional Method	Data Acquisition Failure Rate Under the Proposed QADA Method
Point 1	304 m	24%	0
Point 2	174 m	23%	0.33%
Point 3	198 m	25%	0
Point 4	315 m	21%	0.33%
Point 5	171 m	25%	0
Point 6	164 m	25%	0
Point 7	145 m	25%	0.33%
Point 8	266 m	23%	0
Point 9	307 m	22%	0

. Using the traditional method, the data acquisition failure rate is high, ranging from 21% to 25%. For example, when the wireless sensor node is placed at Point 1, which is 304 m away from Point A (where the gateway server is located), the data acquisition failure rate using the traditional method reaches as high as 24%.

To reduce the data acquisition failure rate, inspired by the time-out retransmission idea of the automatic repeat request (ARQ) scheme [36], a novel QADA method is proposed, which enables the STM32F103C8T6 MCU to dynamically adjust the number of query attempts to the soil sensors. The QADA works as follows: The MCU sends a query frame to the soil sensor and waits for the sensor to return a response frame. If the MCU receives the sensor’s response frame in a timely manner, it will then transmit the sensor data to the gateway server. If the MCU does not receive the sensor’s response frame within a preset time limit, it will initiate a time-out retransmission by sending the query frame to the sensor again. If the MCU receives a response frame from the sensor in this attempt, it discontinues retransmission. Otherwise, it continues retransmission. To avoid the issue of infinite query attempts caused by reasons such as sensor failures, the maximum number of query attempts is set to three. By dynamically changing the amount of query efforts, the QADA method can improve the packet transmission success rate of the wireless sensor node, thereby effectively reducing the data acquisition failure rate of the gateway server. As shown in Table 4, by employing the QADA method, the data acquisition failure rate has been significantly reduced to no more than 0.33%, which demonstrates the effectiveness of the proposed method.

To further assess data acquisition performance in a long-distance scenario, as shown in Figure 16, experiments are conducted in Longhu Park, Zhengzhou City. The red Point B represents the location of the gateway server, and the red Points 1–2 indicate the first and second placement positions of the wireless sensor node, respectively. As shown in

Table 5. Data acquisition failure rate in Longhu Park.

Sensor Location	Distance from Point B	Data Acquisition Failure Rate
Point 1	1780 m	0
Point 2	1696 m	0.33%

, although the distance between the gateway server and wireless sensor node reaches up to 1696 m, the data acquisition failure rate still does not exceed 0.33%, which demonstrates the effectiveness of the proposed method in a long-distance scenario.

5.3. Data Acquisition Delay

Delay is an important performance metric for networked systems [37]. Data acquisition delay $\tau$ is defined as

$$\tau =t_r-t_s$$

(2)

where $t_s$ indicates the instant when the gateway server issues a data acquisition command, and $t_r$ refers to the instant when the gateway server receives data returned by the wireless sensor node. To analyze data acquisition delay, experiments have been conducted at the HAU campus and Longhu Park. Using a sampling period of 1 s and different LoRa transfer rates, data acquisition delays are obtained in Figure 17.

For the experiment on the HAU campus, as shown in Figure 15, the gateway server is placed at Point A, and the wireless sensor node is placed at Point 9. As shown in Table 4, the distance between Point A and Point 9 is 307 m. As shown in Figure 17, when the LoRa transfer rates (in kbps) are set to 1.2, 2.4, 4.8, 9.6, 19.2, 38.4, 50, and 62.5, respectively, the corresponding data acquisition delays (in ms) are obtained as 577.157, 306.69, 175.81, 116.62, 86.262, 68.395, 66.86, and 66.188, respectively.

For the experiment in Longhu Park, as shown in Figure 16, the gateway server is placed at Point B, and the wireless sensor node is placed at Point 2. As shown in Table 5, the distance between Point B and Point 2 is 1696 m. As shown in Figure 17, when the LoRa transfer rates (in kbps) are set to 1.2, 2.4, 4.8, 9.6, 19.2, 38.4, 50, and 62.5, respectively, the corresponding data acquisition delays (in ms) are obtained as 577.463, 306.38, 175.997, 116.693, 86.313, 69.58, 67.08, and 66.213, respectively.

The experimental results show that when the LoRa transmission rate gradually increases from 1.2 kbps to 62.5 kbps, in the campus experiment, the data acquisition delay gradually decreases from 577.157 ms to 66.188 ms, and in the park experiment, it gradually decreases from 577.463 ms to 66.213 ms. That is to say, the data acquisition delay exhibits an inverse proportional variation relationship with the LoRa transfer rate. In the experiments, when the LoRa transmission rate is set to 62.5 kbps, the data acquisition delay can be reduced to below 67 ms.

5.4. Comparison with the Zigbee Method

To compare the performance of Zigbee communication in [38,39] with that of LoRa communication in this paper, an experiment has been conducted on the HAU campus. As shown in Figure 18, Point A indicates the location of the gateway server, and Points 1–9 and Points I–VIII represent the placement positions of the wireless sensor nodes, respectively.

Using the Zigbee wireless communication component with chips CC2530 and CC2591, the experimental results regarding communication distance and data acquisition failure rate are shown in

Table 6. Zigbee experimental data on the HAU campus.

Sensor Location	Distance from Point A	Data Acquisition Failure Rate	Sensor Location	Distance from Point A	Data Acquisition Failure Rate
Point 1	37 m	3.5%	Point I	77 m	49.5%
Point 2	54 m	4.0%	Point II	81 m	33.0%
Point 3	47 m	3.5%	Point III	66 m	39.5%
Point 4	60 m	3.0%	Point IV	87 m	27.0%
Point 5	47 m	2.0%	Point V	69 m	32.0%
Point 6	56 m	3.5%	Point VI	86 m	42.0%
Point 7	52 m	3.5%	Point VII	85 m	33.5%
Point 8	68 m	3.0%	Point VIII	83 m	42.0%

. For example, when the wireless sensor node is placed at Point 1, its distance from the gateway server at Point A is 37 m, and the data acquisition failure rate is 3.5%. Similarly, when the wireless sensor nodes are placed at Points 2–8, the data acquisition failure rates range between 2.0% and 4.0%. When the communication distance is increased, the data acquisition failure rate of the Zigbee method rises rapidly. For example, when the wireless sensor node is placed at Point I, its distance from the gateway server at Point A is 77 m, and the data acquisition failure rate reaches as high as 49.5%. Similarly, when the wireless sensor nodes are placed at Points II–VIII, the data acquisition failure rates increase to a range between 27.0% and 42.0%.

Compared with the aforementioned Zigbee experimental results, the LoRa method adopted in this paper (as depicted in Figure 15 and Table 4) offers a longer communication distance and a lower data acquisition failure rate. For example, when using the Zigbee method, as shown in Table 6, with a communication distance of Points I–VIII ranging from 66 to 87 m, the data acquisition failure rate can be as high as 27.0% to 49.5%. Using the same HAU campus experimental environment, when employing the LoRa method in this paper, as shown in Table 4, with a communication distance ranging from 145 to 315 m, the data acquisition failure rate drops to as low as 0.33% or even less, which verifies the advantages of the LoRa method in this paper in terms of long communication distance and low data acquisition failure rate.

5.5. Multi-Node Soil Detecting System

As shown in Figure 19, the experiment of the multi-node soil detecting system has been conducted on the HAU campus, where the gateway server is placed at Point A, and three wireless sensor nodes are positioned at Points 1–3, respectively. The gateway server is capable of communicating with the three wireless sensor nodes via LoRa and addresses 0x05,0x06, and 0x07, collecting and storing the soil information detected by these sensor nodes.

As illustrated in Figure 20, users can view soil temperature, humidity, and pH value information from Points 1–3 in a unified platform. For example, at 11:00 a.m., the soil data detected by the three sensor nodes at Points 1–3 were as follows: soil temperatures were 20.1 °C (at Point 1), 20.1 °C (at Point 2), and 19.9 °C (at Point 3); soil humidity levels were 21.7% (at Point 1), 21.6% (at Point 2), and 22.4% (at Point 3); and soil pH values were 6.4 (at Point 1), 6.6 (at Point 2), and 7.3 (at Point 3). The multi-node soil detecting system offers a multi-point and multi-dimensional measurement platform for gaining a comprehensive understanding of farmland soil information.

6. Conclusions

This paper develops an IoT-based PV-powered soil remote monitoring system with a low data acquisition failure rate and delay, which mainly encompasses the design of the gateway server node, wireless sensor nodes, a remote monitoring platform, and PV power modules. For the gateway server with the Django framework, the Raspberry Pi 5 sends data acquisition commands to the wireless sensor node via LoRa, receives sensor data, and stores them with timestamp in a MySQL database. Using a remote monitoring platform based on HTML, CSS, and JavaScript, together with a 4G/Wi-Fi router unit and a NAT traversal technique, Internet users can remotely monitor soil status. For the wireless sensor node, the STM32F103C8T6 MCU receives data acquisition commands from the gateway server via LoRa, drives sensors to detect soil temperature, humidity, and pH value, and sends them back to the gateway server. The PV power modules are designed to provide energy to the gateway server and wireless sensor nodes, respectively. Experiments have been conducted at the HAU campus and Longhu park, which confirm the advantages of the proposed IoT-based PV-powered soil remote monitoring system as follows. By using the IoT-based system architecture, users can remotely view real-time and historical information on soil temperature, humidity, and pH value via the Internet. By applying the designed improved QADA method, the data acquisition failure rate can be significantly reduced from 21–25% to no more than 0.33%. Based on the obtained inverse relationship between the data acquisition delay and the LoRa transfer rate, by setting a high LoRa transfer rate of 62.5 kbps, the data acquisition delay can be reduced to below 67 ms. By utilizing both a PV new energy system and lithium batteries for power supply, the system significantly enhances its endurance capacity.

Although the proposed monitoring system has the aforementioned advantages, it still has the following limitations: sensing accuracy can be affected by soil environments, the cost for system development and maintenance is relatively high, and data security is not given enough attention. In the future, we will strive to enhance system stability, reduce system cost, and strengthen data security. Based on the soil data acquired by the proposed system, machine learning and deep learning can be applied to soil information prediction and cloud data analysis. In addition, by leveraging distributed collaboration and privacy protection mechanisms, federated learning can effectively address challenges such as data silos and privacy breaches, which exhibits promising application prospects in smart agriculture.