# From Autonomous Flight to AI Detection in Sustainable Renewable Energy Generation in Sarawak - Event: NestGen ’26 - Published: October 2, 2026 - Length: 36:16 - Watch: https://nestgen.org/library/watch/ws7-5TU8ZvM/ > The transcript of a recorded NestGen session, for reading or for use with an AI tool. It comes from the video's > captions; automatic captions can mishear names and terms. ## About this session Renewable energy operators manage generation assets spread across remote, rugged terrain, where manual inspection is slow, costly, and hard to repeat at the frequency critical infrastructure demands. Turning drone flights into timely, actionable insight means going beyond flying to detecting issues automatically. In this NestGen session, Sarawak Energy will share how they moved from autonomous flight to AI detection across sustainable renewable energy generation in Sarawak, and what utility teams should think through when scaling autonomous inspection and AI-driven analysis across generation assets. ## Transcript **[0:00]** So uh it is my first NestGen uh at FlytBase and uh it is also the first session here at NestGen while uh we have just all come off of keynote. This is uh the first session and we're kicking off with renewable energy. How dog drones are creating value in the re renewable energy domain and I'm sure that a lot of attendees who have tuned into the session are coming from uh hydro solar and uh transmission grids uh backgrounds. Um and uh it would be really lovely to also know if you guys uh have deployed dog drones uh and are coming with like a specific set of questions or what would you like to learn from Hardy in this session. Uh you can feel free to put your questions in the chat box or in in the Q&A tab and we'll take it up from there. Um but I would love to introduce uh our speaker today uh Hardy who has been in the drone industry for a few year now. He began his career as the VP of business **[1:01]** development at Aeroline Group. Uh covering the Malaysian Wario market and supporting Aerodine's global subsidiaries in solution development. This has kind of given him the exposure to drone improvements across some of the world's toughest industrial environments especially um in the Bornean market. Uh his notable projects would include uh oil and gas pipeline surveillance in Brazil, uh multisspectral soil movement analysis in Sarowak, methane monitoring for OGMP 2.0 in Libya, rice crops analysis uh in the Gambia, and nested drone operations for ports and biodiversity monitoring. And today he's joining us uh to represent Saravak Energy. He leads the drone management division over there and he runs the end-to-end operations for security surveillance, asset inspection and project progress monitoring across the utilities footprint. Very warm welcome to you Hardy. Welcome to the Nesgen 50 pioneer list. Welcome to the Nesgjen stage. I will just hand over the stage to you. Uh after I just need to make a couple of like housekeeping announcements over here. So first of all uh I **[2:05]** just want to make sure that everyone is aware that this session will be recorded as Hardy talks. So um even when you have all of your questions like and when he answers them those things will also be recorded so you and we will publish them on nextgen YouTube. So you can always come back to the YouTube and watch the session and share the playbook with your peers. Secondly, um if you find the session valuable, I'm pretty sure that people in your network and people um within um your peer network will also find it useful. So, it'll be great if you share a screenshot anytime while Hardy is talking on LinkedIn. And if you're doing that, please be sure to um add NestGen and you can also tag the NestGen LinkedIn account. So, with that, I will pass on the stage to Hardy. >> Thank you, Ki. Um, thank you Rupes Jen for having me here. Let me just share my screen real quick before I begin with the introduction. [sighs and gasps] >> Are you able to see the screen now? >> Yes. >> Okay. I'm going to turn it into uh **[3:10]** presentation mode. >> Absolutely. We can get started. >> Okay. All right. Thank you, Ki again. Thank you for NestGen for having me here today. I'm very excited. uh this is my first uh NestGen uh that I attend and um and also presenting. So perhaps I'll start with um uh after the introduction that Kani made earlier uh about my experience prior to joining Strawa Energy um last exposure that I had in the last three and a half to four years has been um various industry also globally. So with that experience and now um I'm I'm leading the drone management division at Sarawa Energy. Um Sarawa Energy is a state utility uh company. It's the largest power uh largest utility company in East Malaysia uh developing majority of our power through hydro. So renewable energy. Um we have been a you know long um company that been **[4:13]** here for a while now. uh even before the formation of the country really. Uh so over the years now we're a little bit over 100 years. We are one of the largest uh employers in the state. Um holding about 858,000 accounts across across the state. Um our mission is really to provide and becomes the uh the the one of the main power producer in the region providing power both for residential as well as commercial. um clients of ours in Sraawa and beyond as well. So that being talk about um our facilities and all that where where do uh we come in with drones um the vast amount of facilities that we have uh we also uh basically facing a few challenges here. So covering the entire state we covering various different type of geographical uh location. Um the scale of our operation is also big. So the challen some of these **[5:19]** challenges here are really focusing on these three uh main uh types of challenges that I'm focusing on here. Geographical of course uh in terms of accessibility to some of the sites for maintenance uh and operations. Some of our sites are with no road access. Um so um and it takes when I say no road access meaning they're not pave roads so it takes a while to get there and when it rains really muddy um roads will make it dangerous to access. uh the scale of our facilities as well. Uh the the the hydro for instance are covering more than 700 square kilometer in area and this is just one of it and we have four um and of course uh the number of poles and towers across the state uh distribution alone is more than 25,000 kilometer. So we need manpower to cover such a large area. Um and a lot of time it takes a lot of you know **[6:20]** plenty of time to it's quite time consuming to to arrive at these places um operating to do the inspection um and there's this maintenance that will be involved in it as well. Um so ever since the introduction of drones uh we have been using it utilizing it uh over 10 years now and we grow with the development of drones as well uh from the very early days in 2017 2016 with the phantom uh with a very simple flying and just grabbing images which you know at that point of time um we don't we don't require anyone to climb the poles anymore. So that was a very big significant change from the conventional way of doing inspection um in terms of time as well as safety. So we the next few minutes uh the next few slides I will show I'll be sharing what are our use cases um across the company across our organization **[7:23]** um and how we utilize drone um and the agnostic nature of our operation when it comes to the drones as well as the uh data side of things. So drone will be you know the hardware selection um and then we'll also be covering a little bit of how do we manage the data um to improve efficiency of operations overall. So the bread and butter of utility company of course um we are um producing electricity and the the the power that we generate through our hydros um and some of our other generations um are being transported with transmission and distribution lines. So uh one of the biggest assets in in the organization is of course our overhead lines and this has been um drone has been very uh uh efficient in helping our operations. um where [snorts] we have improved the time of inspection significantly **[8:28]** um as well as improving the safety of doing the inspection very significantly and uh the overhead lines inspections covering the entire span of our assets over 25,000 kilometer and paired with a software that's like um also helping in validating the data that we have an example here you can see on the screen here an example of how we are mapping our asset across the state. And uh previously uh the maps or the GIS data that we have on record might have not been updated uh due to you know from the day they were designed um and then over the years they may be operational change uh maintenance that's uh that were done to the assets and the record is not updated on our GIS. So drone operation uh not only is helping to do improve the operation and maintenance of these poles and towers but they also giving an added value in providing validation of our GIS data um uh you know giving the update to the **[9:33]** records. So this is one of the um you know uh day-to-day example that we have uh one of the mo biggest operate drone operations in our organizations uh with very high frequency as well both in transmission and distribution uh where uh the the number of pilots in the companies as well are majority are of them are in uh transmission and distribution due to the the size of the asset the quantity of the assets we have in this area. Secondly, um a very increasingly popular or one of the most straightforward um use case is of course uh security surveillance where we deploy the docking um or drone in the box uh the docking station or we call it NASA drone here. Um it's very straightforward of helping surveillance and patrol giving that uh um overhead or or yeah overhead view um aerial view. Uh patrol does not need to be um **[10:34]** just physical anymore. Um now with the support of a drone flying above it gives our patrol men um a better view of what's going on in the surrounding. And uh this can help not just the security side of things, it also help with the safety side of things on various type of sites that we have. Uh in this case here is an example of how a a hydroelect electric power plant that's under construction giving them you know the the health uh and safety um parameters as well as uh security surveillance that help with our uh patrol men. And we are slowly increasing the deployment for security and surveillance. And this could also be applicable not just in uh hydrop power plant or a site that's under construction. It can also be deployed in operational sites such as um our substations for both the distribution and also transmission and uh other type of facilities that we have as well. **[11:38]** The third one um which I uh during the keynote I saw some of these had already been mentioned as well in other use cases there's a small part of our generation that's using solar PV while you know majority of our generation is using is hydroelectric renewable energy um we deploy offrid uh places with solar PV so they are scattered around the state um where these places the villages that are not yet connected to our grid. Um we had been deploying um solar uh PV for generation of that uh population over there and um and these are very remote location. So it takes a long time uh you know like significant amount of time uh to get to these places. For that reason over the years they have not uh um able [snorts] to I mean the operational maintenance may be limited due to the time they record to get there. So when we bring in drones and **[12:40]** there uh it really helps with improving their operational efficiency. Um the the traveling and or the operation now has been made easier because we're flying and then with thermal drones we'll be able to identify hotspots which they did not before. Um so this as you know can really improve um the on andm of it. Uh there's a lot of uh costsaving that could be done when you know ahead of time if the panels are starting to have problems and how severe they are uh before it gets really bad. And then because prior to this uh if there's any problem they just replacing the parts without uh any preventive maintenance. um it's just more of like repair replacement. So uh drone thermal drone has been very uh effective and we have proven that it works and very beneficial to the team and uh of course uh with the assistance of AI enabled uh inspection platform uh that we use **[13:44]** it further improve the time that they need to do the inspection and come up with reports and analysis for the management to make decision uh moving forward. In addition to that, the fourth use case is in confined space inspection. Um, uh, as I mentioned earlier, we have hydroelect electric dams that, uh, there are four operating ones currently for us. Um and um prior to prior to drones um you can imagine how you know engineers u maintenance engineers have to uh climb down or into a confined space a pen stock um off a hydroelectric dam and it's it's it's dark and dangerous and slippery in there. And uh basically it's just using human vision to um analyze the condition of it. Uh which obviously not as efficient when we're bl you know flying down a highdefinition camera drone **[14:47]** and be able to get up close to the structure and give you detailed view of what's going on. So during shutdowns like this um now we are we having drones available to us in the market um as well as platforms for the uh analytics providing big provided to the engineers. We're now able to do um the inspection of pentock uh or any confined space uh like a boiler a lot more efficiently um in time in productivity as well. uh this really elevated us significantly for uh inspection of uh facilities or construction or sites um that is GPS denied. Um this is something fairly uh new to us. Um and uh we're looking forward to expanding this program as well to all our confined space uh facilities um in the future moving forward. So again agnostic type of approach to the drones uh **[15:51]** where we are we are working with um different type of confined space drone whether is tethered system um or cage system um and u continuously bringing uh the different type of payloads as well both visual there's lidar as well as uh uh UT payloads for um thickness measurement for the assets drone that Um another um payload that is also increasingly uh popular in use for our um organization is LAR. Um there are a few case use cases for LAR. Um vegetation is being one of it. Um as you are allware aware that point clouds provide you with measurement analytics. So um anything that require precise measurement um uh point cloud can you provide you that. So in this case your vegetation is one of the uh application that we deploy **[16:52]** lidar for uh where some of these places are quite remote as I mentioned earlier you know uh we have a lot of uh assets uh that's located in hard to access area for that reason they are seldom being checked on and uh vegetation can grow very quickly in this region and uh uh you know for that reason it will affect uh some of these assets and could cause like tripping Uh, for example, so flying a drone over an area that's not reachable um by vehicle right off the spot. Um, and then when you can scan uh a few kilometer of your assets, come back and provide you with this data, we can easily identify and focus on an area that we want to fix. Uh so this has been also a use case that we have proved the value for and we will continue to expand this throughout the organization uh for vegetation encouragement and uh of course uh LAR providing um the the digital **[17:53]** map for uh slope monitoring you know elevation contour is very useful for our um engineering team um not to especially in the small focus area prior Prior to drone um previously helicopters were deployed with LAR sensors. Uh but for a smaller area uh it might not be economically feasible to you know get a helicopter to fly. Um so this is where drones comes in. uh it's complementing or supplementing uh previous data that been done with helicopter and then um drone can do a smaller more focus and giving a higher accuracy um data for the engineering team for their assessment. So the next one um this is a video a short uh clip uh taken from our uh docking docking or nested drone. uh just to show I think a lot of you are aware of this uh uh **[18:56]** functionality from uh the Nessa drone you know that's deployed with AI um with uh object identification quantification and uh an interesting um use case on the right side of the screen um night vision is for security uh where uh we were able to identify movement um of intruders at our facility at night, which may not be um possible had it not been for this drone. If patrolman are going on foot, uh the suspects or the intruders uh might be aware of them and be hiding somewhere. But in this case here, we were flying a drone above um and turning off the lights and everything and high enough that they cannot hear us. And then we from that command center deploy our patrolman to go to a specific site where we are monitoring them using this drone. So they can efficiently effectively apprehend these intruders. Um and that's proving that drone is very helpful uh for **[20:00]** security felons um in previously which was not possible to do at night. So uh just sharing a bit of like uh some exciting findings that we have and operations uh that we conducted for um security and surveillance. uh you know again this give the uh uh decision maker an informed um data for them to make that decision. Um >> other than inspecting and um surveillance we also use um nested drone for or or combination of LAR and terrestrial and aerial or drone based LAR to create a digital twin for the purpose of flight planning. So this is a a use case an example of how um at a substation the drone can do security surveillance and also inspection. So more use cases and application per drone will increase the value in the use of the drone. So uh we continuously **[21:01]** trying to find how um you know any technology that we bring in or testing can give the utmost value to our operations. In this case here, uh drone based LAR and terrestrial LAR being consolidated to provide um a digital twin per se, uh for for the substation. And with that um we are planning flights based on the digital twin uh for the drone to do daily inspection or scheduled inspection autonomously uh without having to previously someone flying a pilot a piloted drone. Um so with this this is some of the uh you know uh show showing how we integrating different use cases into one for uh our facility and um this will be uh the last use case I would like to share before uh concluding the in the next slide. So what what what have we done over the years in terms of like you know using drones **[22:01]** for in in many different areas? Um we have gone as far as into uh developing guidelines uh on how to fly uh or how to do data capture for the purpose of improving um operational workflow um and also uh AI models that we are training in house as well. So with this consistency with this guideline it provide consistency of data capture and uh it's it's a cycle the more data we capture the better the uh machine learning is and it helps to improve the analysis on the platform when we deploy these AI models on it. So um and we are expanding this or using this uh workflow in all the different different different areas that we have um not just in overhead lines but uh once we are maturing in other the application as well we will also like come up with a guideline uh to provide consistency of data capture that will be utilized for uh data analysis and **[23:01]** AI and machine learning in future. So what do we get out from uh deploying drones as mentioned throughout this conversation or my presentation? There's definitely significant saving monetary um savings from our operation. um we have at least one of the use cases achieving more than 80% efficiency when AI is paired with our operations and of course uh being a renewable energy uh producer uh sustainability goals is um important to us and uh drone operations definitely provide us with that value um for uh for for our organizations. So this is um the last slide ki from from me um and uh I think I'm happy happy to hear a question if there's any coming from you. >> Absolutely. There are a few questions quite a few questions in the uh Q&A tab over here which I will surface one by one. Um so thank you so much Hardy. First of all I think this was an incredible **[24:04]** presentation. uh especially the part where you explained how AI um is now like completely changing your workflow and like how you guys are like headon uh you know going head first into adopting the technology and figuring out what are the best use cases. I think that's super inspiring and of course like tying it back to what we were tying what we communicating in the keynote as well that uh you know like AI is like a big unlock in terms of where the drone industry is going. So really inspiring to see how uh you at Sarowak have sort of like um completely given a playbook for like how um AI can be leveraged and uh from I think our attendees we have like a bunch of questions. So let me just pull them up um over here. >> Um so you mentioned AI detection. What is AI actually detecting on your renewable assets and how accurate has it been compared to what your maintenance teams were catching manually? Maybe on a high level we can answer this Hardy. >> Yeah, sure. **[25:07]** Um, our main use case in AI detection right now that we built in house is overhead line uh inspection defects. Uh, so identifying things like beehives, um, cracks, uh, fleshover marks on on insulators. um you know the percentage accuracy has uh gone up to more than 70%. Um and this is something that we consistent like I said hence why we come up with a guideline right to keep consistency on the data and it will continuously train them. So uh we expect that the AI models will improve over time um with more data that we capture. Uh so that's one of them. Yep. >> Understood. >> Uh let me pull up the next one. Uh so they're saying that Sarak's renewable sites are spread across a dense tropical terrain. How far apart are your dogs and can one dog cover multiple generation **[26:07]** assets or does each site need its own? [snorts] >> Um very far apart actually. So when we talk about if I take example of our um hydro u electric plant uh or hps they are a few hundred kilometers apart. Uh so for deploying uh for security surveillance at these hydro plants they the docks will be localized. So each of these power warehouses have their own docks. They might have multiple at that particular powerhouses for different use cases. uh but we do not cover a large area given the limitation of the docks and also the the the how far apart our our assets are um not even substations are quite far apart. So we have not yet uh reached to a point where one dock um covering a few assets at this point of time unless I think in the future you know the direction of fixed wing uh with a dock that's possible. So we're we're we look forward to that in the future. Amazing. **[27:12]** Um, uh, maybe I'll I'll just pull up the next question over here. >> Do you use any additional payload with the drone? If yes, how are you managing multiple use cases? Since there's only one PSDK port, does the hardware support swapping of payloads? >> Right. Uh, we do, but not with different not with single drone. So right now our mode of operation is uh each any payload that a drone um so thermal has its own drone [snorts] uh RGB has its own lidar carry its own so we're we're not swapping payloads per drone at this point of time um while that's possible with some of the hardware out there uh at at this current of moment of time our operational method that we have a particular drone for a particular payload uh for ease of operation I think as we bit more experience in operation um and you know also gaining more technical knowledge internally as well um this **[28:18]** will be what we're looking at uh moving forward >> right >> I think um that answers the question uh so another question over here um takes a while to load on the screen tropical environments uh of course Like I mean given especially like the kind of environments that Sarabak works in they mean constant rain. You mentioned extreme humidity and heavy vegetation growths. How often does weather prevent the drone from flying and how does that affect your inspection schedule? >> Yeah. So very interesting question that is true. Vegetation growth and also rain affects our uh operation for sure. um weather when it rains obviously we cannot fly. Um however for in terms of inspection uh schedule they will have to look at the weather. So they will uh typically if it's not monsoon **[29:19]** raining the whole day um heavy rain typically occurs in the afternoon. So the team might want to schedule everything in the morning. Um so this is this is uh something that the team is um uh aware and um they will have to fit that into their schedule operational schedule. Um so looking at the forecast is one way to help it. Uh so planning your time of schedule throughout the day. Um unless it rains um you know like monsoon it rains typically the whole day but on a normal day um heavy rain typically occurs in the afternoon. So the team will try to um finish up in the morning or before afternoon. So that's that's also why we come up with a guideline and then uh to have a sufficient enough uh images that cover our um area of interest. Uh with that they can you know work around the schedule based on the weather. Um but yes you know the rain here being a tropical rainforest it tends to **[30:20]** be heavy and when it does rain there's no operation. But uh so my question to that as a followup could be Hardy that uh um so when uh let's say the drone is already in flight and in case the weather gets bad I'm pretty sure like that wouldn't affect the flight itself like does it actually hamper anything? Well, we have the pre-flight checks, right? Um, if we know the rain, we would know because uh there's there's radar coming in, wind speed, all this indicate if the rain is going to come and the drone will also give you warning if the the wind speed is at such level that it's it might not recommend you to take off. So, there will be on a prerogative of the pilot of the operation whether they would want to still take off and take that risk. Um so sometimes this is uh it depends on the operation uh the need to take off in that higher risk situation and uh we expect that they're all trained um of course we expect that they can uh handle those situ and make that calculated risk um before they take off. Um so u I think with that you **[31:27]** know the pilot will not still go out and knowing that it's going to rain heavily then they they probably will you know return to home uh for the operation. >> Understood. Understood. All right. Uh just I I think we're running uh over time by a couple of minutes, but I think there are a lot of interesting questions. Um I'm looking at a specific one where a person said person has said, "I'm a drone and robotics manager from a similar company from Netherlands." Uh well, hi there. Um how can you summarize on a high level what were the key ROI metrics you used to convince your hires? I'm making a case. However, my team has been constantly asking me for KPI metrics. Would be glad to connect after the session as well. >> Very well. I think um there are uh different use cases and different um ROI and KPIs for sure. Uh you can use security surveillance for instance. You can you can calculate how long does it take your patrolman to patrol before a drone existed and **[32:28]** then you know you can um compare that to a drone. When you deploy a drone a perimeter of like two square kilometer the drone can fly in however many minutes. Meanwhile the patrolman if they use a conventional way whether it's on a vehicle it will take them versus a few uh this you know amount of time. So that that is one way of looking at it through time. Um [snorts] you can also um look at how much services costs out there uh compared to if you were to do it on on your own when you purchase a drone and the payload and the processing time involved if you have the resources for it. So that's that's another way. Um I think that on a high level those are the two example that uh I I can give but of course it depends on different organization will have a different way of doing things. Um I think these are just parameters that I can I can mention and share the calculation will come down to how you operate. Um I think it's different from one organization to another on you know what really what **[33:33]** are the things you were trying to achieve. makes complete sense and I think we have uh just one last question. Uh somebody in the chat has said um based on the use cases looks like you have various you have tested various drone platforms and uh are your pilots drone specific? >> Yes. Um, [snorts] are our pilots drone specific meaning one pilot pilot to one drone or um, if I could answer I mean uh >> I think probably they they're probably asking about like the brand of drones or or like the drone hardware. >> Yeah, >> it would be specific for you. >> Yeah. Yeah. So, it's quite agnostic. Uh, we we take the agnostic approach. Um however our operation our drone pilots are specific to their operation meaning the transmission team will be working deploying drones for transmission related um use cases. The security surveillance will only do security **[34:38]** surveillance scope. So they will not do inspection from transmission um vice versa. So uh the pilots that are that are you know that are certified in in the organization are all within their own um respective units and business unit and operations. >> Perfect. I think that answers all the audience's questions. Thank you so much uh for taking the time uh and thank you. I mean even though we ran out of time we are over running by 5 minutes but it's incredible like that people have like people were quite curious about like specific things and thank you so much for taking the time out for answering each one of them and like uh um our friend from Netherlands says that they would be glad to connect after the session as well. If other attendees also feel the same way I am just dropping Hardy's LinkedIn uh in the chat box. You can always uh connect with him. Uh I'm speaking for for you Hardy over here. I hope that's not a problem. >> No problem. Happy happy to uh connect after this session and uh thank you so much **[35:41]** Khi for uh moderating this. Thank you to your team as well for having me. It's good to see some of the familiar faces and names in there. I I I haven't really got in touch with them but uh um yeah thank you so much and great grateful for this opportunity. >> It was wonderful to work with you on this Hari. I will see you soon and I hope to see you in other sessions at NestGen today as well. >> Look forward to it. Thank you. Have a good day everyone.